diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,139642 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9999946280747557,
+  "eval_steps": 500,
+  "global_step": 23269,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.430615164520744e-08,
+      "loss": 2.0497,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.861230329041488e-08,
+      "loss": 1.9261,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.291845493562232e-08,
+      "loss": 1.9504,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.722460658082976e-08,
+      "loss": 1.8377,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.15307582260372e-08,
+      "loss": 2.1877,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.583690987124465e-08,
+      "loss": 1.7979,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0014306151645208e-07,
+      "loss": 1.8225,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1444921316165952e-07,
+      "loss": 2.0478,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2875536480686695e-07,
+      "loss": 1.744,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.430615164520744e-07,
+      "loss": 1.9221,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5736766809728184e-07,
+      "loss": 1.5679,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.716738197424893e-07,
+      "loss": 1.6196,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8597997138769672e-07,
+      "loss": 1.909,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0028612303290415e-07,
+      "loss": 1.9246,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.145922746781116e-07,
+      "loss": 1.9066,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2889842632331904e-07,
+      "loss": 2.3913,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4320457796852647e-07,
+      "loss": 1.7268,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.575107296137339e-07,
+      "loss": 2.3374,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.718168812589414e-07,
+      "loss": 2.1937,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.861230329041488e-07,
+      "loss": 2.0396,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0042918454935624e-07,
+      "loss": 1.6594,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1473533619456367e-07,
+      "loss": 1.7744,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2904148783977116e-07,
+      "loss": 1.9218,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.433476394849786e-07,
+      "loss": 1.6821,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.57653791130186e-07,
+      "loss": 2.0334,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.7195994277539345e-07,
+      "loss": 2.3941,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.8626609442060093e-07,
+      "loss": 2.0397,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.005722460658083e-07,
+      "loss": 1.8525,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.148783977110158e-07,
+      "loss": 1.6923,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.291845493562232e-07,
+      "loss": 2.1885,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.434907010014307e-07,
+      "loss": 2.4603,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.577968526466381e-07,
+      "loss": 2.2375,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.7210300429184556e-07,
+      "loss": 1.6545,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.864091559370529e-07,
+      "loss": 1.713,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.007153075822605e-07,
+      "loss": 2.297,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.150214592274678e-07,
+      "loss": 2.2765,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.293276108726753e-07,
+      "loss": 2.0982,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.436337625178828e-07,
+      "loss": 1.5208,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.579399141630902e-07,
+      "loss": 1.9758,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.722460658082976e-07,
+      "loss": 1.8141,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.865522174535051e-07,
+      "loss": 2.0282,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.008583690987125e-07,
+      "loss": 2.2176,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.151645207439199e-07,
+      "loss": 2.1856,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.294706723891273e-07,
+      "loss": 2.1039,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.437768240343348e-07,
+      "loss": 1.8223,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.580829756795423e-07,
+      "loss": 1.5534,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.723891273247497e-07,
+      "loss": 1.7059,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.866952789699572e-07,
+      "loss": 2.3174,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.010014306151645e-07,
+      "loss": 2.081,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.15307582260372e-07,
+      "loss": 1.9285,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.296137339055795e-07,
+      "loss": 1.854,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.439198855507869e-07,
+      "loss": 1.7468,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.582260371959943e-07,
+      "loss": 1.7557,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.725321888412019e-07,
+      "loss": 1.6152,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.868383404864093e-07,
+      "loss": 1.75,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.011444921316166e-07,
+      "loss": 1.7709,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.15450643776824e-07,
+      "loss": 1.8085,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.297567954220316e-07,
+      "loss": 1.7696,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.44062947067239e-07,
+      "loss": 1.7375,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.583690987124464e-07,
+      "loss": 1.9221,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.726752503576539e-07,
+      "loss": 2.0312,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.869814020028614e-07,
+      "loss": 1.7697,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.012875536480687e-07,
+      "loss": 1.7217,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.155937052932762e-07,
+      "loss": 1.5616,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.298998569384836e-07,
+      "loss": 1.6039,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.442060085836911e-07,
+      "loss": 1.6388,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.585121602288984e-07,
+      "loss": 1.8032,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.728183118741059e-07,
+      "loss": 1.8773,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.871244635193133e-07,
+      "loss": 1.9401,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.001430615164521e-06,
+      "loss": 1.9572,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0157367668097284e-06,
+      "loss": 1.4213,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0300429184549356e-06,
+      "loss": 1.6139,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.044349070100143e-06,
+      "loss": 1.7494,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0586552217453507e-06,
+      "loss": 1.6376,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.072961373390558e-06,
+      "loss": 1.4781,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0872675250357655e-06,
+      "loss": 1.5486,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.101573676680973e-06,
+      "loss": 1.5408,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1158798283261804e-06,
+      "loss": 1.2952,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1301859799713878e-06,
+      "loss": 1.834,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1444921316165953e-06,
+      "loss": 1.869,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1587982832618027e-06,
+      "loss": 1.5534,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1731044349070101e-06,
+      "loss": 1.2054,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1874105865522175e-06,
+      "loss": 1.3365,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.201716738197425e-06,
+      "loss": 1.6891,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2160228898426324e-06,
+      "loss": 1.5585,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2303290414878398e-06,
+      "loss": 1.2728,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2446351931330473e-06,
+      "loss": 1.6392,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2589413447782547e-06,
+      "loss": 1.4046,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2732474964234623e-06,
+      "loss": 1.1687,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2875536480686696e-06,
+      "loss": 1.1732,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3018597997138772e-06,
+      "loss": 1.5055,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3161659513590846e-06,
+      "loss": 1.4395,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3304721030042918e-06,
+      "loss": 1.5057,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3447782546494995e-06,
+      "loss": 1.5278,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3590844062947067e-06,
+      "loss": 1.2806,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3733905579399143e-06,
+      "loss": 1.5583,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3876967095851218e-06,
+      "loss": 1.108,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.402002861230329e-06,
+      "loss": 1.3959,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4163090128755366e-06,
+      "loss": 1.4329,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.430615164520744e-06,
+      "loss": 1.2841,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4449213161659515e-06,
+      "loss": 1.3221,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.459227467811159e-06,
+      "loss": 1.4202,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4735336194563661e-06,
+      "loss": 1.5131,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4878397711015738e-06,
+      "loss": 1.3882,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5021459227467814e-06,
+      "loss": 1.3056,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5164520743919886e-06,
+      "loss": 1.3164,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.530758226037196e-06,
+      "loss": 1.5115,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5450643776824037e-06,
+      "loss": 1.6479,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.559370529327611e-06,
+      "loss": 1.1483,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5736766809728186e-06,
+      "loss": 1.2829,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5879828326180258e-06,
+      "loss": 1.1747,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6022889842632332e-06,
+      "loss": 1.4052,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6165951359084409e-06,
+      "loss": 1.3513,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.630901287553648e-06,
+      "loss": 1.3603,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6452074391988557e-06,
+      "loss": 1.3609,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6595135908440632e-06,
+      "loss": 1.3587,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6738197424892704e-06,
+      "loss": 1.5572,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.688125894134478e-06,
+      "loss": 1.5184,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7024320457796852e-06,
+      "loss": 1.348,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7167381974248929e-06,
+      "loss": 1.3546,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7310443490701003e-06,
+      "loss": 1.5148,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7453505007153077e-06,
+      "loss": 1.2792,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7596566523605152e-06,
+      "loss": 1.3309,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7739628040057228e-06,
+      "loss": 1.3809,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.78826895565093e-06,
+      "loss": 1.1866,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8025751072961375e-06,
+      "loss": 1.2106,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8168812589413449e-06,
+      "loss": 1.2661,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8311874105865523e-06,
+      "loss": 1.3189,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.84549356223176e-06,
+      "loss": 1.2286,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8597997138769672e-06,
+      "loss": 1.2588,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8741058655221748e-06,
+      "loss": 1.1688,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8884120171673823e-06,
+      "loss": 1.2301,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9027181688125895e-06,
+      "loss": 1.2122,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.917024320457797e-06,
+      "loss": 1.398,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9313304721030043e-06,
+      "loss": 1.4308,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9456366237482118e-06,
+      "loss": 1.3401,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9599427753934196e-06,
+      "loss": 1.225,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9742489270386266e-06,
+      "loss": 1.3688,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.988555078683834e-06,
+      "loss": 1.1973,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.002861230329042e-06,
+      "loss": 1.5082,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.017167381974249e-06,
+      "loss": 1.5748,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0314735336194568e-06,
+      "loss": 1.3493,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0457796852646638e-06,
+      "loss": 1.2889,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.060085836909871e-06,
+      "loss": 1.0909,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.074391988555079e-06,
+      "loss": 1.3441,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.088698140200286e-06,
+      "loss": 1.364,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.103004291845494e-06,
+      "loss": 1.1839,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1173104434907013e-06,
+      "loss": 1.3246,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1316165951359088e-06,
+      "loss": 1.3216,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.145922746781116e-06,
+      "loss": 1.2168,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.160228898426323e-06,
+      "loss": 1.3711,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.174535050071531e-06,
+      "loss": 1.3449,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1888412017167385e-06,
+      "loss": 1.2941,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.203147353361946e-06,
+      "loss": 1.2318,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2174535050071534e-06,
+      "loss": 1.1925,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2317596566523608e-06,
+      "loss": 1.3632,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2460658082975682e-06,
+      "loss": 1.2163,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2603719599427756e-06,
+      "loss": 1.2795,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.274678111587983e-06,
+      "loss": 1.5887,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2889842632331905e-06,
+      "loss": 1.2111,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.303290414878398e-06,
+      "loss": 1.4725,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3175965665236054e-06,
+      "loss": 1.3898,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.331902718168813e-06,
+      "loss": 1.3348,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3462088698140202e-06,
+      "loss": 1.1825,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3605150214592277e-06,
+      "loss": 1.5072,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.374821173104435e-06,
+      "loss": 1.129,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3891273247496425e-06,
+      "loss": 1.4599,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.40343347639485e-06,
+      "loss": 1.4239,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4177396280400574e-06,
+      "loss": 1.2629,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.432045779685265e-06,
+      "loss": 1.3854,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4463519313304722e-06,
+      "loss": 1.0879,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4606580829756797e-06,
+      "loss": 1.4066,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.474964234620887e-06,
+      "loss": 1.1046,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4892703862660945e-06,
+      "loss": 1.4145,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.503576537911302e-06,
+      "loss": 1.2776,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5178826895565094e-06,
+      "loss": 1.2101,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5321888412017172e-06,
+      "loss": 1.4481,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5464949928469247e-06,
+      "loss": 1.1414,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5608011444921317e-06,
+      "loss": 1.2607,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.575107296137339e-06,
+      "loss": 1.373,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5894134477825465e-06,
+      "loss": 1.4124,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6037195994277544e-06,
+      "loss": 1.1762,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.618025751072962e-06,
+      "loss": 1.3323,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6323319027181693e-06,
+      "loss": 1.3503,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6466380543633763e-06,
+      "loss": 1.4091,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6609442060085837e-06,
+      "loss": 1.6058,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6752503576537915e-06,
+      "loss": 1.3394,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.689556509298999e-06,
+      "loss": 1.2537,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7038626609442064e-06,
+      "loss": 1.3345,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7181688125894134e-06,
+      "loss": 1.183,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.732474964234621e-06,
+      "loss": 1.3832,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7467811158798287e-06,
+      "loss": 1.2604,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.761087267525036e-06,
+      "loss": 1.2473,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7753934191702436e-06,
+      "loss": 1.3494,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7896995708154506e-06,
+      "loss": 1.3265,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.804005722460658e-06,
+      "loss": 1.2688,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.818311874105866e-06,
+      "loss": 1.1935,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8326180257510733e-06,
+      "loss": 1.1691,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8469241773962807e-06,
+      "loss": 1.3672,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.861230329041488e-06,
+      "loss": 1.2157,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.875536480686695e-06,
+      "loss": 1.0263,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.889842632331903e-06,
+      "loss": 1.3012,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9041487839771104e-06,
+      "loss": 1.3322,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.918454935622318e-06,
+      "loss": 1.219,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9327610872675257e-06,
+      "loss": 1.0381,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9470672389127323e-06,
+      "loss": 1.1157,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.96137339055794e-06,
+      "loss": 1.3362,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9756795422031476e-06,
+      "loss": 1.188,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.989985693848355e-06,
+      "loss": 1.0444,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.004291845493563e-06,
+      "loss": 1.1709,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.01859799713877e-06,
+      "loss": 1.35,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0329041487839773e-06,
+      "loss": 1.4312,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0472103004291847e-06,
+      "loss": 1.2292,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.061516452074392e-06,
+      "loss": 1.4418,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0758226037196e-06,
+      "loss": 1.4781,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0901287553648074e-06,
+      "loss": 1.3424,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1044349070100144e-06,
+      "loss": 1.1475,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.118741058655222e-06,
+      "loss": 1.2749,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1330472103004293e-06,
+      "loss": 1.4067,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.147353361945637e-06,
+      "loss": 1.2197,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1616595135908446e-06,
+      "loss": 1.3152,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1759656652360516e-06,
+      "loss": 1.1865,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.190271816881259e-06,
+      "loss": 1.192,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2045779685264665e-06,
+      "loss": 1.1509,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2188841201716743e-06,
+      "loss": 1.0049,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2331902718168817e-06,
+      "loss": 1.3605,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2474964234620887e-06,
+      "loss": 1.3905,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.261802575107296e-06,
+      "loss": 1.38,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2761087267525036e-06,
+      "loss": 1.3629,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2904148783977115e-06,
+      "loss": 1.3254,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.304721030042919e-06,
+      "loss": 1.1595,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3190271816881263e-06,
+      "loss": 1.2282,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 1.2776,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3476394849785408e-06,
+      "loss": 1.3127,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3619456366237486e-06,
+      "loss": 1.2307,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.376251788268956e-06,
+      "loss": 1.2275,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3905579399141635e-06,
+      "loss": 1.1637,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4048640915593705e-06,
+      "loss": 1.0742,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4191702432045783e-06,
+      "loss": 1.1034,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4334763948497858e-06,
+      "loss": 1.3241,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.447782546494993e-06,
+      "loss": 1.2875,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4620886981402006e-06,
+      "loss": 1.4233,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4763948497854076e-06,
+      "loss": 1.1385,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4907010014306155e-06,
+      "loss": 1.1569,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.505007153075823e-06,
+      "loss": 1.2197,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5193133047210303e-06,
+      "loss": 1.2402,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5336194563662378e-06,
+      "loss": 1.3282,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5479256080114456e-06,
+      "loss": 1.4254,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5622317596566526e-06,
+      "loss": 1.2784,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.57653791130186e-06,
+      "loss": 1.3613,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5908440629470675e-06,
+      "loss": 1.1117,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.605150214592275e-06,
+      "loss": 1.028,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6194563662374828e-06,
+      "loss": 1.2798,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6337625178826898e-06,
+      "loss": 1.2516,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.648068669527897e-06,
+      "loss": 1.1781,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6623748211731046e-06,
+      "loss": 1.1467,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.676680972818312e-06,
+      "loss": 1.1073,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.69098712446352e-06,
+      "loss": 1.2672,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.705293276108727e-06,
+      "loss": 1.2434,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7195994277539344e-06,
+      "loss": 1.3531,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.733905579399142e-06,
+      "loss": 1.1796,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7482117310443496e-06,
+      "loss": 1.1428,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.762517882689557e-06,
+      "loss": 1.1986,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7768240343347645e-06,
+      "loss": 1.2472,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7911301859799715e-06,
+      "loss": 1.2375,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.805436337625179e-06,
+      "loss": 1.2034,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.819742489270387e-06,
+      "loss": 1.1227,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.834048640915594e-06,
+      "loss": 1.2983,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.848354792560802e-06,
+      "loss": 1.1039,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.862660944206009e-06,
+      "loss": 1.2123,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8769670958512165e-06,
+      "loss": 1.1721,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8912732474964235e-06,
+      "loss": 1.1222,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.905579399141631e-06,
+      "loss": 1.1241,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.919885550786839e-06,
+      "loss": 1.2062,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.934191702432046e-06,
+      "loss": 1.3181,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.948497854077253e-06,
+      "loss": 1.3121,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.962804005722461e-06,
+      "loss": 1.1426,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.977110157367668e-06,
+      "loss": 1.2065,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.991416309012876e-06,
+      "loss": 1.3692,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.005722460658084e-06,
+      "loss": 1.1284,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.020028612303291e-06,
+      "loss": 1.3334,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.034334763948498e-06,
+      "loss": 1.1074,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.048640915593706e-06,
+      "loss": 1.1469,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0629470672389135e-06,
+      "loss": 1.2514,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0772532188841205e-06,
+      "loss": 1.2777,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0915593705293275e-06,
+      "loss": 1.2255,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.105865522174535e-06,
+      "loss": 1.181,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.120171673819742e-06,
+      "loss": 1.1603,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.13447782546495e-06,
+      "loss": 1.3536,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.148783977110158e-06,
+      "loss": 1.2061,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.163090128755365e-06,
+      "loss": 1.0824,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.177396280400572e-06,
+      "loss": 1.0552,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.19170243204578e-06,
+      "loss": 1.3327,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.206008583690988e-06,
+      "loss": 1.3376,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.220314735336195e-06,
+      "loss": 1.3303,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.234620886981403e-06,
+      "loss": 1.1341,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.24892703862661e-06,
+      "loss": 1.2745,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2632331902718175e-06,
+      "loss": 1.3829,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2775393419170246e-06,
+      "loss": 1.1843,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.291845493562232e-06,
+      "loss": 1.2352,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.306151645207439e-06,
+      "loss": 1.3126,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.320457796852646e-06,
+      "loss": 1.1466,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.334763948497854e-06,
+      "loss": 1.2636,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.349070100143062e-06,
+      "loss": 1.0143,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.363376251788269e-06,
+      "loss": 1.1118,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.377682403433477e-06,
+      "loss": 1.1667,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.391988555078684e-06,
+      "loss": 1.0844,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.406294706723892e-06,
+      "loss": 1.4327,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.420600858369099e-06,
+      "loss": 1.1325,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.434907010014307e-06,
+      "loss": 1.351,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.449213161659514e-06,
+      "loss": 1.2001,
+      "step": 311
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4635193133047216e-06,
+      "loss": 1.0559,
+      "step": 312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4778254649499286e-06,
+      "loss": 1.0691,
+      "step": 313
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4921316165951364e-06,
+      "loss": 1.0759,
+      "step": 314
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5064377682403434e-06,
+      "loss": 1.0589,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.520743919885551e-06,
+      "loss": 1.0647,
+      "step": 316
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.535050071530759e-06,
+      "loss": 1.1535,
+      "step": 317
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.549356223175966e-06,
+      "loss": 1.3021,
+      "step": 318
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.563662374821173e-06,
+      "loss": 1.2881,
+      "step": 319
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.577968526466381e-06,
+      "loss": 1.2313,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.592274678111589e-06,
+      "loss": 1.1491,
+      "step": 321
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.606580829756796e-06,
+      "loss": 0.9778,
+      "step": 322
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.620886981402003e-06,
+      "loss": 1.201,
+      "step": 323
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.635193133047211e-06,
+      "loss": 1.1437,
+      "step": 324
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.649499284692418e-06,
+      "loss": 1.1593,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.663805436337626e-06,
+      "loss": 1.1305,
+      "step": 326
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6781115879828334e-06,
+      "loss": 1.0823,
+      "step": 327
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6924177396280405e-06,
+      "loss": 1.2416,
+      "step": 328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7067238912732475e-06,
+      "loss": 1.1407,
+      "step": 329
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.721030042918455e-06,
+      "loss": 1.168,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.735336194563663e-06,
+      "loss": 1.2663,
+      "step": 331
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.74964234620887e-06,
+      "loss": 1.1117,
+      "step": 332
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.763948497854078e-06,
+      "loss": 1.1866,
+      "step": 333
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.778254649499285e-06,
+      "loss": 1.196,
+      "step": 334
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.792560801144492e-06,
+      "loss": 1.1528,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8068669527897e-06,
+      "loss": 1.3451,
+      "step": 336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.821173104434908e-06,
+      "loss": 1.1555,
+      "step": 337
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.835479256080115e-06,
+      "loss": 1.2187,
+      "step": 338
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.849785407725322e-06,
+      "loss": 1.1914,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.86409155937053e-06,
+      "loss": 1.0504,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8783977110157375e-06,
+      "loss": 1.307,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8927038626609445e-06,
+      "loss": 1.1035,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.907010014306152e-06,
+      "loss": 1.1348,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.921316165951359e-06,
+      "loss": 1.3473,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.935622317596566e-06,
+      "loss": 1.2409,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.949928469241774e-06,
+      "loss": 1.106,
+      "step": 346
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.964234620886982e-06,
+      "loss": 1.1269,
+      "step": 347
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.978540772532189e-06,
+      "loss": 1.1805,
+      "step": 348
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.992846924177397e-06,
+      "loss": 1.2541,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.007153075822604e-06,
+      "loss": 1.2474,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.021459227467812e-06,
+      "loss": 1.1284,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.035765379113019e-06,
+      "loss": 1.0382,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.050071530758226e-06,
+      "loss": 1.0785,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.0643776824034345e-06,
+      "loss": 1.2018,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.0786838340486415e-06,
+      "loss": 1.1935,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.092989985693849e-06,
+      "loss": 1.2368,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.107296137339056e-06,
+      "loss": 1.2517,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.121602288984263e-06,
+      "loss": 1.1175,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.135908440629471e-06,
+      "loss": 1.1028,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.150214592274678e-06,
+      "loss": 1.2183,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.164520743919886e-06,
+      "loss": 1.172,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.178826895565093e-06,
+      "loss": 1.2755,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.1931330472103e-06,
+      "loss": 1.3312,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.207439198855509e-06,
+      "loss": 1.231,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.221745350500716e-06,
+      "loss": 1.2179,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.236051502145924e-06,
+      "loss": 1.1716,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.250357653791131e-06,
+      "loss": 1.015,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2646638054363385e-06,
+      "loss": 1.0967,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2789699570815455e-06,
+      "loss": 0.975,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2932761087267525e-06,
+      "loss": 1.2233,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.30758226037196e-06,
+      "loss": 1.352,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.321888412017167e-06,
+      "loss": 1.0201,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.336194563662376e-06,
+      "loss": 1.0811,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.350500715307583e-06,
+      "loss": 1.1839,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.36480686695279e-06,
+      "loss": 1.3005,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.379113018597998e-06,
+      "loss": 1.2394,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.393419170243205e-06,
+      "loss": 1.0952,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.407725321888413e-06,
+      "loss": 1.1709,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.42203147353362e-06,
+      "loss": 1.2562,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.436337625178827e-06,
+      "loss": 1.2239,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.450643776824035e-06,
+      "loss": 1.1699,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.464949928469242e-06,
+      "loss": 1.1611,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.47925608011445e-06,
+      "loss": 1.0375,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.493562231759657e-06,
+      "loss": 0.9122,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.507868383404864e-06,
+      "loss": 1.0397,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.522174535050072e-06,
+      "loss": 1.0665,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.536480686695279e-06,
+      "loss": 1.0806,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.550786838340487e-06,
+      "loss": 1.2058,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.565092989985694e-06,
+      "loss": 1.1407,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.579399141630901e-06,
+      "loss": 0.9694,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.593705293276109e-06,
+      "loss": 1.0959,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.608011444921316e-06,
+      "loss": 1.2486,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.622317596566525e-06,
+      "loss": 1.2196,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.636623748211732e-06,
+      "loss": 1.0687,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.650929899856939e-06,
+      "loss": 1.0641,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.6652360515021465e-06,
+      "loss": 1.0538,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.6795422031473535e-06,
+      "loss": 0.9961,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.693848354792561e-06,
+      "loss": 1.2169,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.708154506437768e-06,
+      "loss": 1.1727,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.722460658082976e-06,
+      "loss": 1.155,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.736766809728183e-06,
+      "loss": 1.2198,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.75107296137339e-06,
+      "loss": 1.1951,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.765379113018599e-06,
+      "loss": 1.2429,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.779685264663806e-06,
+      "loss": 1.1921,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.793991416309014e-06,
+      "loss": 1.0997,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.808297567954221e-06,
+      "loss": 1.162,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.822603719599428e-06,
+      "loss": 1.0876,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.836909871244636e-06,
+      "loss": 1.1202,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.851216022889843e-06,
+      "loss": 1.1639,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.865522174535051e-06,
+      "loss": 1.2464,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.879828326180258e-06,
+      "loss": 1.1973,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.894134477825465e-06,
+      "loss": 1.3039,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.908440629470673e-06,
+      "loss": 1.1779,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.92274678111588e-06,
+      "loss": 1.0957,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.937052932761088e-06,
+      "loss": 1.0114,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.951359084406295e-06,
+      "loss": 1.0288,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.965665236051502e-06,
+      "loss": 0.9689,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.97997138769671e-06,
+      "loss": 1.0632,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.994277539341917e-06,
+      "loss": 1.0,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.008583690987126e-06,
+      "loss": 0.9836,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.022889842632333e-06,
+      "loss": 0.9834,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.03719599427754e-06,
+      "loss": 1.217,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.051502145922748e-06,
+      "loss": 1.1212,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.065808297567955e-06,
+      "loss": 1.1704,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.0801144492131624e-06,
+      "loss": 1.0115,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.0944206008583694e-06,
+      "loss": 1.1563,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.1087267525035765e-06,
+      "loss": 1.2324,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.123032904148784e-06,
+      "loss": 1.1746,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.137339055793991e-06,
+      "loss": 1.2489,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.1516452074392e-06,
+      "loss": 1.0774,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.165951359084407e-06,
+      "loss": 1.1808,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.180257510729615e-06,
+      "loss": 1.0182,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.194563662374822e-06,
+      "loss": 0.9358,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.208869814020029e-06,
+      "loss": 1.257,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.223175965665237e-06,
+      "loss": 1.2269,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.237482117310444e-06,
+      "loss": 1.241,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.251788268955652e-06,
+      "loss": 1.2468,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.266094420600859e-06,
+      "loss": 1.1161,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.280400572246066e-06,
+      "loss": 1.2309,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.294706723891274e-06,
+      "loss": 1.1545,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.309012875536481e-06,
+      "loss": 1.1431,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.323319027181689e-06,
+      "loss": 1.1992,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.337625178826896e-06,
+      "loss": 1.1952,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.351931330472103e-06,
+      "loss": 1.0284,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.366237482117311e-06,
+      "loss": 1.1076,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.380543633762518e-06,
+      "loss": 1.128,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.394849785407726e-06,
+      "loss": 1.2658,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.409155937052933e-06,
+      "loss": 1.2337,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.42346208869814e-06,
+      "loss": 1.0494,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.437768240343349e-06,
+      "loss": 1.0093,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.452074391988556e-06,
+      "loss": 1.2169,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.4663805436337635e-06,
+      "loss": 0.9092,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.4806866952789705e-06,
+      "loss": 0.9525,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.4949928469241775e-06,
+      "loss": 1.334,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.509298998569385e-06,
+      "loss": 1.0662,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.523605150214592e-06,
+      "loss": 1.1299,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.5379113018598e-06,
+      "loss": 1.1904,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.552217453505007e-06,
+      "loss": 0.9884,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.566523605150214e-06,
+      "loss": 0.9777,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.580829756795423e-06,
+      "loss": 1.2505,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.59513590844063e-06,
+      "loss": 1.1095,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.609442060085838e-06,
+      "loss": 1.1983,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.623748211731045e-06,
+      "loss": 1.0058,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.638054363376253e-06,
+      "loss": 1.1968,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.65236051502146e-06,
+      "loss": 1.0927,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.0666,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.680972818311875e-06,
+      "loss": 1.1467,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.6952789699570815e-06,
+      "loss": 1.0953,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.70958512160229e-06,
+      "loss": 1.1672,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.723891273247497e-06,
+      "loss": 1.2535,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.738197424892704e-06,
+      "loss": 1.0799,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.752503576537912e-06,
+      "loss": 1.3051,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.766809728183119e-06,
+      "loss": 0.9367,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.781115879828327e-06,
+      "loss": 1.0057,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.795422031473534e-06,
+      "loss": 1.1898,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.809728183118741e-06,
+      "loss": 1.1397,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.82403433476395e-06,
+      "loss": 0.999,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.838340486409157e-06,
+      "loss": 1.1575,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.8526466380543645e-06,
+      "loss": 1.0279,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.8669527896995715e-06,
+      "loss": 1.3642,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.8812589413447785e-06,
+      "loss": 1.0767,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.895565092989986e-06,
+      "loss": 1.0343,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.909871244635193e-06,
+      "loss": 1.1108,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.924177396280401e-06,
+      "loss": 0.9931,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.938483547925608e-06,
+      "loss": 1.2092,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.952789699570815e-06,
+      "loss": 1.1677,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.967095851216024e-06,
+      "loss": 1.1557,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.981402002861231e-06,
+      "loss": 1.1603,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.995708154506439e-06,
+      "loss": 1.0857,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.010014306151646e-06,
+      "loss": 1.1163,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.024320457796853e-06,
+      "loss": 1.0541,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.038626609442061e-06,
+      "loss": 0.9209,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.052932761087268e-06,
+      "loss": 1.1641,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.0672389127324755e-06,
+      "loss": 0.984,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.0815450643776825e-06,
+      "loss": 1.2088,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.095851216022891e-06,
+      "loss": 1.2288,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.110157367668098e-06,
+      "loss": 1.0257,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.124463519313305e-06,
+      "loss": 1.1271,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.138769670958513e-06,
+      "loss": 1.0706,
+      "step": 499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.15307582260372e-06,
+      "loss": 1.1534,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.167381974248928e-06,
+      "loss": 1.0459,
+      "step": 501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.181688125894135e-06,
+      "loss": 1.0811,
+      "step": 502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.195994277539342e-06,
+      "loss": 1.3096,
+      "step": 503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.21030042918455e-06,
+      "loss": 1.0522,
+      "step": 504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.224606580829757e-06,
+      "loss": 1.1897,
+      "step": 505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.2389127324749655e-06,
+      "loss": 1.2312,
+      "step": 506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.2532188841201726e-06,
+      "loss": 1.0771,
+      "step": 507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.2675250357653796e-06,
+      "loss": 1.0239,
+      "step": 508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.281831187410587e-06,
+      "loss": 1.1095,
+      "step": 509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.296137339055794e-06,
+      "loss": 1.0359,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.310443490701002e-06,
+      "loss": 1.3219,
+      "step": 511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.324749642346209e-06,
+      "loss": 1.2068,
+      "step": 512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.339055793991416e-06,
+      "loss": 1.1618,
+      "step": 513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.353361945636624e-06,
+      "loss": 0.946,
+      "step": 514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.367668097281831e-06,
+      "loss": 0.8616,
+      "step": 515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.38197424892704e-06,
+      "loss": 1.3015,
+      "step": 516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.396280400572247e-06,
+      "loss": 1.1551,
+      "step": 517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.410586552217454e-06,
+      "loss": 1.2152,
+      "step": 518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.424892703862662e-06,
+      "loss": 1.3144,
+      "step": 519
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.439198855507869e-06,
+      "loss": 1.1897,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.4535050071530766e-06,
+      "loss": 1.1864,
+      "step": 521
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.467811158798284e-06,
+      "loss": 1.0178,
+      "step": 522
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.482117310443491e-06,
+      "loss": 1.2431,
+      "step": 523
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.496423462088699e-06,
+      "loss": 1.1749,
+      "step": 524
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.5107296137339054e-06,
+      "loss": 1.2131,
+      "step": 525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.525035765379114e-06,
+      "loss": 1.1985,
+      "step": 526
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.539341917024321e-06,
+      "loss": 1.2513,
+      "step": 527
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.553648068669529e-06,
+      "loss": 1.1296,
+      "step": 528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.567954220314736e-06,
+      "loss": 1.222,
+      "step": 529
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.582260371959943e-06,
+      "loss": 1.2454,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.596566523605151e-06,
+      "loss": 1.1235,
+      "step": 531
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.610872675250358e-06,
+      "loss": 1.1607,
+      "step": 532
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.625178826895567e-06,
+      "loss": 1.1864,
+      "step": 533
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.639484978540774e-06,
+      "loss": 0.9554,
+      "step": 534
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.65379113018598e-06,
+      "loss": 1.1247,
+      "step": 535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.668097281831188e-06,
+      "loss": 1.1302,
+      "step": 536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.682403433476395e-06,
+      "loss": 1.0292,
+      "step": 537
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.696709585121603e-06,
+      "loss": 1.2815,
+      "step": 538
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.71101573676681e-06,
+      "loss": 1.1369,
+      "step": 539
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.725321888412017e-06,
+      "loss": 1.111,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.739628040057226e-06,
+      "loss": 1.0282,
+      "step": 541
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.753934191702433e-06,
+      "loss": 1.1047,
+      "step": 542
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.76824034334764e-06,
+      "loss": 1.28,
+      "step": 543
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.782546494992847e-06,
+      "loss": 0.9848,
+      "step": 544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.796852646638054e-06,
+      "loss": 1.1417,
+      "step": 545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.811158798283263e-06,
+      "loss": 1.1212,
+      "step": 546
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.82546494992847e-06,
+      "loss": 0.9772,
+      "step": 547
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.839771101573678e-06,
+      "loss": 1.0936,
+      "step": 548
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.854077253218885e-06,
+      "loss": 1.1022,
+      "step": 549
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.868383404864092e-06,
+      "loss": 1.0623,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.8826895565093e-06,
+      "loss": 0.9137,
+      "step": 551
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.896995708154506e-06,
+      "loss": 1.0506,
+      "step": 552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.911301859799715e-06,
+      "loss": 1.1111,
+      "step": 553
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.925608011444922e-06,
+      "loss": 1.1823,
+      "step": 554
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.93991416309013e-06,
+      "loss": 1.2501,
+      "step": 555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.954220314735336e-06,
+      "loss": 1.1277,
+      "step": 556
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.968526466380543e-06,
+      "loss": 1.0336,
+      "step": 557
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.982832618025752e-06,
+      "loss": 0.9283,
+      "step": 558
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.997138769670959e-06,
+      "loss": 1.0602,
+      "step": 559
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.011444921316168e-06,
+      "loss": 0.902,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.025751072961375e-06,
+      "loss": 1.2411,
+      "step": 561
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.040057224606582e-06,
+      "loss": 1.1942,
+      "step": 562
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.054363376251789e-06,
+      "loss": 1.1897,
+      "step": 563
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.068669527896996e-06,
+      "loss": 0.8762,
+      "step": 564
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.082975679542204e-06,
+      "loss": 1.1122,
+      "step": 565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.097281831187411e-06,
+      "loss": 0.8586,
+      "step": 566
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.111587982832618e-06,
+      "loss": 1.0942,
+      "step": 567
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.125894134477827e-06,
+      "loss": 1.0473,
+      "step": 568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.140200286123034e-06,
+      "loss": 1.1124,
+      "step": 569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.154506437768241e-06,
+      "loss": 1.2449,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.168812589413448e-06,
+      "loss": 0.9646,
+      "step": 571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.183118741058655e-06,
+      "loss": 1.036,
+      "step": 572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.197424892703864e-06,
+      "loss": 1.0462,
+      "step": 573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.21173104434907e-06,
+      "loss": 1.2152,
+      "step": 574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.226037195994278e-06,
+      "loss": 0.9467,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.240343347639485e-06,
+      "loss": 1.2501,
+      "step": 576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.254649499284692e-06,
+      "loss": 1.2219,
+      "step": 577
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.2689556509299e-06,
+      "loss": 1.0924,
+      "step": 578
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.283261802575108e-06,
+      "loss": 1.0708,
+      "step": 579
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.297567954220316e-06,
+      "loss": 1.0855,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.311874105865523e-06,
+      "loss": 0.9688,
+      "step": 581
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.32618025751073e-06,
+      "loss": 1.147,
+      "step": 582
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.340486409155937e-06,
+      "loss": 1.0324,
+      "step": 583
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.354792560801144e-06,
+      "loss": 1.1422,
+      "step": 584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.369098712446353e-06,
+      "loss": 1.2019,
+      "step": 585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.38340486409156e-06,
+      "loss": 1.0396,
+      "step": 586
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.397711015736767e-06,
+      "loss": 1.082,
+      "step": 587
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.412017167381976e-06,
+      "loss": 1.2775,
+      "step": 588
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.426323319027183e-06,
+      "loss": 1.1122,
+      "step": 589
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.44062947067239e-06,
+      "loss": 1.0735,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.454935622317597e-06,
+      "loss": 1.013,
+      "step": 591
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.469241773962805e-06,
+      "loss": 1.0821,
+      "step": 592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.483547925608012e-06,
+      "loss": 1.1735,
+      "step": 593
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.49785407725322e-06,
+      "loss": 1.0631,
+      "step": 594
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.512160228898428e-06,
+      "loss": 1.2443,
+      "step": 595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.526466380543635e-06,
+      "loss": 1.1332,
+      "step": 596
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.540772532188842e-06,
+      "loss": 1.0711,
+      "step": 597
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.555078683834049e-06,
+      "loss": 1.0077,
+      "step": 598
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.569384835479256e-06,
+      "loss": 1.0306,
+      "step": 599
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.583690987124465e-06,
+      "loss": 0.9192,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.597997138769672e-06,
+      "loss": 0.9153,
+      "step": 601
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.612303290414879e-06,
+      "loss": 1.0914,
+      "step": 602
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.626609442060086e-06,
+      "loss": 1.0089,
+      "step": 603
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.640915593705293e-06,
+      "loss": 1.0843,
+      "step": 604
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.655221745350502e-06,
+      "loss": 0.9383,
+      "step": 605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.669527896995709e-06,
+      "loss": 1.0482,
+      "step": 606
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.683834048640917e-06,
+      "loss": 1.0715,
+      "step": 607
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.698140200286124e-06,
+      "loss": 1.3819,
+      "step": 608
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.712446351931331e-06,
+      "loss": 0.961,
+      "step": 609
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.726752503576538e-06,
+      "loss": 1.2277,
+      "step": 610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.741058655221745e-06,
+      "loss": 1.13,
+      "step": 611
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.755364806866954e-06,
+      "loss": 1.2119,
+      "step": 612
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.769670958512161e-06,
+      "loss": 1.2716,
+      "step": 613
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.783977110157368e-06,
+      "loss": 1.0675,
+      "step": 614
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.798283261802577e-06,
+      "loss": 1.1215,
+      "step": 615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.812589413447784e-06,
+      "loss": 1.1553,
+      "step": 616
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.82689556509299e-06,
+      "loss": 1.2452,
+      "step": 617
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.841201716738198e-06,
+      "loss": 1.0021,
+      "step": 618
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.855507868383405e-06,
+      "loss": 1.0943,
+      "step": 619
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.869814020028613e-06,
+      "loss": 1.0827,
+      "step": 620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.88412017167382e-06,
+      "loss": 1.0211,
+      "step": 621
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.898426323319027e-06,
+      "loss": 1.1055,
+      "step": 622
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.912732474964234e-06,
+      "loss": 1.2088,
+      "step": 623
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.927038626609443e-06,
+      "loss": 1.0547,
+      "step": 624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.94134477825465e-06,
+      "loss": 1.0274,
+      "step": 625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.955650929899857e-06,
+      "loss": 1.1682,
+      "step": 626
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.969957081545066e-06,
+      "loss": 1.3117,
+      "step": 627
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.984263233190273e-06,
+      "loss": 1.0674,
+      "step": 628
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.99856938483548e-06,
+      "loss": 1.0762,
+      "step": 629
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.012875536480687e-06,
+      "loss": 1.0628,
+      "step": 630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.027181688125894e-06,
+      "loss": 1.0775,
+      "step": 631
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.041487839771103e-06,
+      "loss": 1.0086,
+      "step": 632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.05579399141631e-06,
+      "loss": 1.0484,
+      "step": 633
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.070100143061518e-06,
+      "loss": 0.9527,
+      "step": 634
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.084406294706725e-06,
+      "loss": 1.0102,
+      "step": 635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.098712446351932e-06,
+      "loss": 1.0469,
+      "step": 636
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.11301859799714e-06,
+      "loss": 1.1015,
+      "step": 637
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.127324749642346e-06,
+      "loss": 1.0637,
+      "step": 638
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.141630901287555e-06,
+      "loss": 1.1243,
+      "step": 639
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.155937052932762e-06,
+      "loss": 1.2887,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.170243204577969e-06,
+      "loss": 1.0977,
+      "step": 641
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.184549356223178e-06,
+      "loss": 1.0112,
+      "step": 642
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.198855507868383e-06,
+      "loss": 0.9385,
+      "step": 643
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.213161659513592e-06,
+      "loss": 1.0492,
+      "step": 644
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.227467811158799e-06,
+      "loss": 1.0027,
+      "step": 645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.241773962804006e-06,
+      "loss": 1.1273,
+      "step": 646
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.256080114449214e-06,
+      "loss": 1.2264,
+      "step": 647
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.270386266094421e-06,
+      "loss": 0.9571,
+      "step": 648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.284692417739628e-06,
+      "loss": 1.1433,
+      "step": 649
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.298998569384835e-06,
+      "loss": 1.1228,
+      "step": 650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.313304721030042e-06,
+      "loss": 1.1138,
+      "step": 651
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.327610872675251e-06,
+      "loss": 0.9578,
+      "step": 652
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.341917024320458e-06,
+      "loss": 1.0844,
+      "step": 653
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.356223175965667e-06,
+      "loss": 1.0749,
+      "step": 654
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.370529327610874e-06,
+      "loss": 1.0496,
+      "step": 655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.384835479256081e-06,
+      "loss": 1.1549,
+      "step": 656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.399141630901288e-06,
+      "loss": 0.9982,
+      "step": 657
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.413447782546495e-06,
+      "loss": 0.9739,
+      "step": 658
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.427753934191704e-06,
+      "loss": 1.1428,
+      "step": 659
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.44206008583691e-06,
+      "loss": 1.1055,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.45636623748212e-06,
+      "loss": 1.2437,
+      "step": 661
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.470672389127326e-06,
+      "loss": 1.0401,
+      "step": 662
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.484978540772533e-06,
+      "loss": 1.0464,
+      "step": 663
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.49928469241774e-06,
+      "loss": 0.8645,
+      "step": 664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.513590844062947e-06,
+      "loss": 0.9789,
+      "step": 665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.527896995708156e-06,
+      "loss": 1.0025,
+      "step": 666
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.542203147353363e-06,
+      "loss": 1.0798,
+      "step": 667
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.55650929899857e-06,
+      "loss": 1.0357,
+      "step": 668
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.570815450643777e-06,
+      "loss": 0.9981,
+      "step": 669
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.585121602288984e-06,
+      "loss": 1.131,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.599427753934193e-06,
+      "loss": 1.142,
+      "step": 671
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.6137339055794e-06,
+      "loss": 1.0019,
+      "step": 672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.628040057224607e-06,
+      "loss": 1.0512,
+      "step": 673
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.642346208869815e-06,
+      "loss": 1.0105,
+      "step": 674
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.656652360515022e-06,
+      "loss": 1.2167,
+      "step": 675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.67095851216023e-06,
+      "loss": 1.3028,
+      "step": 676
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.685264663805437e-06,
+      "loss": 0.955,
+      "step": 677
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.699570815450644e-06,
+      "loss": 1.1857,
+      "step": 678
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.713876967095852e-06,
+      "loss": 0.8891,
+      "step": 679
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.72818311874106e-06,
+      "loss": 0.9331,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.742489270386268e-06,
+      "loss": 1.0914,
+      "step": 681
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.756795422031475e-06,
+      "loss": 1.178,
+      "step": 682
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.771101573676682e-06,
+      "loss": 1.1687,
+      "step": 683
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.785407725321889e-06,
+      "loss": 0.8766,
+      "step": 684
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.799713876967096e-06,
+      "loss": 1.1011,
+      "step": 685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.814020028612305e-06,
+      "loss": 0.8963,
+      "step": 686
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.828326180257512e-06,
+      "loss": 1.1008,
+      "step": 687
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.842632331902719e-06,
+      "loss": 1.0244,
+      "step": 688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.856938483547926e-06,
+      "loss": 1.133,
+      "step": 689
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.871244635193133e-06,
+      "loss": 1.146,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.885550786838341e-06,
+      "loss": 0.9737,
+      "step": 691
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.899856938483548e-06,
+      "loss": 1.1642,
+      "step": 692
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.914163090128757e-06,
+      "loss": 1.0828,
+      "step": 693
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.928469241773964e-06,
+      "loss": 1.0829,
+      "step": 694
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.942775393419171e-06,
+      "loss": 1.028,
+      "step": 695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.957081545064378e-06,
+      "loss": 1.1097,
+      "step": 696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971387696709585e-06,
+      "loss": 1.1684,
+      "step": 697
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985693848354794e-06,
+      "loss": 1.0985,
+      "step": 698
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-05,
+      "loss": 0.91,
+      "step": 699
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999999951563069e-06,
+      "loss": 0.9201,
+      "step": 700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999980625227e-06,
+      "loss": 1.0077,
+      "step": 701
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999956406761e-06,
+      "loss": 1.1464,
+      "step": 702
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999999225009095e-06,
+      "loss": 0.9894,
+      "step": 703
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999998789076726e-06,
+      "loss": 1.2068,
+      "step": 704
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999998256270515e-06,
+      "loss": 1.1267,
+      "step": 705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999997626590472e-06,
+      "loss": 1.3655,
+      "step": 706
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999690003661e-06,
+      "loss": 1.085,
+      "step": 707
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999996076608943e-06,
+      "loss": 0.9903,
+      "step": 708
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999995156307485e-06,
+      "loss": 1.0007,
+      "step": 709
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999994139132254e-06,
+      "loss": 1.0073,
+      "step": 710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999993025083271e-06,
+      "loss": 0.9686,
+      "step": 711
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999181416056e-06,
+      "loss": 0.997,
+      "step": 712
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999990506364141e-06,
+      "loss": 0.9826,
+      "step": 713
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99998910169404e-06,
+      "loss": 0.8399,
+      "step": 714
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999987600150282e-06,
+      "loss": 1.0727,
+      "step": 715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9999860017329e-06,
+      "loss": 1.131,
+      "step": 716
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999984306441923e-06,
+      "loss": 1.1339,
+      "step": 717
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999982514277385e-06,
+      "loss": 1.0368,
+      "step": 718
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999980625239321e-06,
+      "loss": 1.1248,
+      "step": 719
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999978639327764e-06,
+      "loss": 0.9502,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999976556542756e-06,
+      "loss": 1.0659,
+      "step": 721
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999974376884338e-06,
+      "loss": 1.1168,
+      "step": 722
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999972100352548e-06,
+      "loss": 0.7582,
+      "step": 723
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999969726947436e-06,
+      "loss": 1.1287,
+      "step": 724
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99996725666904e-06,
+      "loss": 1.1716,
+      "step": 725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999964689517417e-06,
+      "loss": 1.076,
+      "step": 726
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99996202549261e-06,
+      "loss": 1.1684,
+      "step": 727
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999959264594673e-06,
+      "loss": 1.0829,
+      "step": 728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99995640682366e-06,
+      "loss": 1.1004,
+      "step": 729
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999953452179627e-06,
+      "loss": 1.0275,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99995040066263e-06,
+      "loss": 1.0159,
+      "step": 731
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999947252272726e-06,
+      "loss": 0.8174,
+      "step": 732
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99994400700998e-06,
+      "loss": 1.2808,
+      "step": 733
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999940664874453e-06,
+      "loss": 1.036,
+      "step": 734
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999937225866208e-06,
+      "loss": 0.9984,
+      "step": 735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999933689985315e-06,
+      "loss": 0.9556,
+      "step": 736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999930057231839e-06,
+      "loss": 1.024,
+      "step": 737
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999926327605854e-06,
+      "loss": 0.9492,
+      "step": 738
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99992250110743e-06,
+      "loss": 1.1455,
+      "step": 739
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999918577736643e-06,
+      "loss": 1.1726,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999914557493566e-06,
+      "loss": 1.2151,
+      "step": 741
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99991044037828e-06,
+      "loss": 0.8916,
+      "step": 742
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999906226390863e-06,
+      "loss": 0.8705,
+      "step": 743
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999901915531395e-06,
+      "loss": 0.9393,
+      "step": 744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999897507799964e-06,
+      "loss": 0.9328,
+      "step": 745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999893003196653e-06,
+      "loss": 1.0585,
+      "step": 746
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99988840172155e-06,
+      "loss": 1.0769,
+      "step": 747
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99988370337474e-06,
+      "loss": 1.1848,
+      "step": 748
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999878908156319e-06,
+      "loss": 0.8864,
+      "step": 749
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99987401606638e-06,
+      "loss": 1.0304,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999869027105014e-06,
+      "loss": 1.0503,
+      "step": 751
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999863941272321e-06,
+      "loss": 1.0966,
+      "step": 752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999858758568397e-06,
+      "loss": 1.0888,
+      "step": 753
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999853478993344e-06,
+      "loss": 0.8743,
+      "step": 754
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999848102547265e-06,
+      "loss": 0.9116,
+      "step": 755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999842629230263e-06,
+      "loss": 1.0889,
+      "step": 756
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999837059042445e-06,
+      "loss": 1.0594,
+      "step": 757
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999831391983916e-06,
+      "loss": 1.0344,
+      "step": 758
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999825628054788e-06,
+      "loss": 0.9932,
+      "step": 759
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999819767255175e-06,
+      "loss": 1.0365,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999813809585186e-06,
+      "loss": 1.0395,
+      "step": 761
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999807755044938e-06,
+      "loss": 0.9264,
+      "step": 762
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999801603634552e-06,
+      "loss": 1.2452,
+      "step": 763
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999795355354143e-06,
+      "loss": 0.9685,
+      "step": 764
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99978901020383e-06,
+      "loss": 1.1324,
+      "step": 765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999782568183742e-06,
+      "loss": 1.0543,
+      "step": 766
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999776029294002e-06,
+      "loss": 1.025,
+      "step": 767
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999769393534733e-06,
+      "loss": 1.227,
+      "step": 768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999762660906068e-06,
+      "loss": 0.9277,
+      "step": 769
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999755831408133e-06,
+      "loss": 1.3328,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999748905041065e-06,
+      "loss": 1.1791,
+      "step": 771
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999741881804995e-06,
+      "loss": 1.1965,
+      "step": 772
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999734761700061e-06,
+      "loss": 0.9967,
+      "step": 773
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999727544726401e-06,
+      "loss": 1.224,
+      "step": 774
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999720230884154e-06,
+      "loss": 1.2902,
+      "step": 775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99971282017346e-06,
+      "loss": 1.218,
+      "step": 776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999705312594464e-06,
+      "loss": 1.0631,
+      "step": 777
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999697708147312e-06,
+      "loss": 0.9525,
+      "step": 778
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999690006832152e-06,
+      "loss": 0.9316,
+      "step": 779
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999682208649132e-06,
+      "loss": 0.9285,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999674313598401e-06,
+      "loss": 0.9224,
+      "step": 781
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999666321680119e-06,
+      "loss": 1.1218,
+      "step": 782
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999658232894433e-06,
+      "loss": 0.9456,
+      "step": 783
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999650047241503e-06,
+      "loss": 0.8248,
+      "step": 784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999641764721488e-06,
+      "loss": 0.9486,
+      "step": 785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999633385334548e-06,
+      "loss": 0.8922,
+      "step": 786
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999624909080845e-06,
+      "loss": 1.1786,
+      "step": 787
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999616335960542e-06,
+      "loss": 0.9945,
+      "step": 788
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99960766597381e-06,
+      "loss": 1.1058,
+      "step": 789
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999598899120811e-06,
+      "loss": 1.0413,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999590035401718e-06,
+      "loss": 1.0525,
+      "step": 791
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999581074816702e-06,
+      "loss": 1.1551,
+      "step": 792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999572017365937e-06,
+      "loss": 1.082,
+      "step": 793
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999562863049598e-06,
+      "loss": 0.967,
+      "step": 794
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999553611867863e-06,
+      "loss": 1.1071,
+      "step": 795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99954426382091e-06,
+      "loss": 1.1368,
+      "step": 796
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999534818908922e-06,
+      "loss": 0.9847,
+      "step": 797
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99952527713208e-06,
+      "loss": 1.0699,
+      "step": 798
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999515638490572e-06,
+      "loss": 1.0932,
+      "step": 799
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999505902984581e-06,
+      "loss": 1.0606,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999496070614296e-06,
+      "loss": 1.1916,
+      "step": 801
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99948614137991e-06,
+      "loss": 1.0988,
+      "step": 802
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999476115281615e-06,
+      "loss": 1.1711,
+      "step": 803
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999465992319603e-06,
+      "loss": 0.9603,
+      "step": 804
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999455772494072e-06,
+      "loss": 1.2187,
+      "step": 805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99944545580522e-06,
+      "loss": 1.0094,
+      "step": 806
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999435042253244e-06,
+      "loss": 1.0083,
+      "step": 807
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99942453183835e-06,
+      "loss": 1.0737,
+      "step": 808
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99941392456074e-06,
+      "loss": 1.1846,
+      "step": 809
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999403220420619e-06,
+      "loss": 1.1655,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999392419418194e-06,
+      "loss": 1.1215,
+      "step": 811
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999381521553675e-06,
+      "loss": 1.0712,
+      "step": 812
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999370526827274e-06,
+      "loss": 1.0816,
+      "step": 813
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999359435239202e-06,
+      "loss": 0.9871,
+      "step": 814
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999348246789677e-06,
+      "loss": 1.0284,
+      "step": 815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999336961478912e-06,
+      "loss": 1.2113,
+      "step": 816
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999325579307128e-06,
+      "loss": 1.0467,
+      "step": 817
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999314100274545e-06,
+      "loss": 0.9368,
+      "step": 818
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999302524381385e-06,
+      "loss": 1.0743,
+      "step": 819
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999290851627873e-06,
+      "loss": 1.0373,
+      "step": 820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999279082014233e-06,
+      "loss": 1.001,
+      "step": 821
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999267215540697e-06,
+      "loss": 0.8555,
+      "step": 822
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999255252207492e-06,
+      "loss": 0.9716,
+      "step": 823
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99924319201485e-06,
+      "loss": 0.8371,
+      "step": 824
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999231034963007e-06,
+      "loss": 0.9572,
+      "step": 825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999218781052193e-06,
+      "loss": 1.1052,
+      "step": 826
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999206430282653e-06,
+      "loss": 1.1511,
+      "step": 827
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99919398265462e-06,
+      "loss": 1.0024,
+      "step": 828
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99918143816834e-06,
+      "loss": 0.8674,
+      "step": 829
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999168796824051e-06,
+      "loss": 1.0354,
+      "step": 830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999156058622002e-06,
+      "loss": 1.0265,
+      "step": 831
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999143223562437e-06,
+      "loss": 0.9943,
+      "step": 832
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999130291645606e-06,
+      "loss": 1.115,
+      "step": 833
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999117262871763e-06,
+      "loss": 1.0513,
+      "step": 834
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999104137241153e-06,
+      "loss": 1.1388,
+      "step": 835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999090914754038e-06,
+      "loss": 1.0063,
+      "step": 836
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999077595410666e-06,
+      "loss": 0.882,
+      "step": 837
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999064179211305e-06,
+      "loss": 0.9746,
+      "step": 838
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999050666156204e-06,
+      "loss": 0.9234,
+      "step": 839
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999037056245634e-06,
+      "loss": 1.1693,
+      "step": 840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999023349479855e-06,
+      "loss": 0.9837,
+      "step": 841
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99900954585913e-06,
+      "loss": 1.0279,
+      "step": 842
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998995645383732e-06,
+      "loss": 1.1946,
+      "step": 843
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998981648053925e-06,
+      "loss": 1.098,
+      "step": 844
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998967553869982e-06,
+      "loss": 1.1589,
+      "step": 845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99895336283218e-06,
+      "loss": 1.1038,
+      "step": 846
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998939074940788e-06,
+      "loss": 0.9141,
+      "step": 847
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998924690196085e-06,
+      "loss": 1.0618,
+      "step": 848
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99891020859835e-06,
+      "loss": 1.2091,
+      "step": 849
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998895630147867e-06,
+      "loss": 1.1092,
+      "step": 850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998880954844912e-06,
+      "loss": 1.101,
+      "step": 851
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998866182689772e-06,
+      "loss": 1.101,
+      "step": 852
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998851313682736e-06,
+      "loss": 1.1245,
+      "step": 853
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998836347824088e-06,
+      "loss": 1.0018,
+      "step": 854
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998821285114122e-06,
+      "loss": 0.9816,
+      "step": 855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998806125553125e-06,
+      "loss": 0.9995,
+      "step": 856
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998790869141394e-06,
+      "loss": 1.0825,
+      "step": 857
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998775515879223e-06,
+      "loss": 1.1061,
+      "step": 858
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998760065766911e-06,
+      "loss": 0.8498,
+      "step": 859
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998744518804758e-06,
+      "loss": 1.0949,
+      "step": 860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998728874993062e-06,
+      "loss": 1.0616,
+      "step": 861
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998713134332128e-06,
+      "loss": 1.036,
+      "step": 862
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998697296822262e-06,
+      "loss": 1.2632,
+      "step": 863
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99868136246377e-06,
+      "loss": 1.1252,
+      "step": 864
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998665331256958e-06,
+      "loss": 0.9504,
+      "step": 865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998649203202142e-06,
+      "loss": 1.0354,
+      "step": 866
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998632978299629e-06,
+      "loss": 1.2934,
+      "step": 867
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998616656549739e-06,
+      "loss": 0.9405,
+      "step": 868
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998600237952783e-06,
+      "loss": 1.128,
+      "step": 869
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998583722509082e-06,
+      "loss": 0.8151,
+      "step": 870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998567110218954e-06,
+      "loss": 0.982,
+      "step": 871
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998550401082722e-06,
+      "loss": 0.9594,
+      "step": 872
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998533595100713e-06,
+      "loss": 0.935,
+      "step": 873
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998516692273246e-06,
+      "loss": 1.0422,
+      "step": 874
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998499692600653e-06,
+      "loss": 1.0996,
+      "step": 875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998482596083263e-06,
+      "loss": 1.0225,
+      "step": 876
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998465402721407e-06,
+      "loss": 1.0648,
+      "step": 877
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998448112515417e-06,
+      "loss": 1.0883,
+      "step": 878
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99843072546563e-06,
+      "loss": 0.9023,
+      "step": 879
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998413241572378e-06,
+      "loss": 1.1853,
+      "step": 880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998395660836005e-06,
+      "loss": 0.8912,
+      "step": 881
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998377983256851e-06,
+      "loss": 1.0702,
+      "step": 882
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998360208835256e-06,
+      "loss": 1.0774,
+      "step": 883
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998342337571566e-06,
+      "loss": 1.0246,
+      "step": 884
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998324369466128e-06,
+      "loss": 0.8844,
+      "step": 885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998306304519287e-06,
+      "loss": 1.1256,
+      "step": 886
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998288142731399e-06,
+      "loss": 1.0005,
+      "step": 887
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99826988410281e-06,
+      "loss": 0.9858,
+      "step": 888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998251528633874e-06,
+      "loss": 1.086,
+      "step": 889
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99823307632495e-06,
+      "loss": 1.0428,
+      "step": 890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998214527176393e-06,
+      "loss": 1.0447,
+      "step": 891
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998195881188565e-06,
+      "loss": 1.2454,
+      "step": 892
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998177138361825e-06,
+      "loss": 0.9693,
+      "step": 893
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998158298696537e-06,
+      "loss": 1.0828,
+      "step": 894
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998139362193064e-06,
+      "loss": 0.8902,
+      "step": 895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998120328851776e-06,
+      "loss": 1.01,
+      "step": 896
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998101198673041e-06,
+      "loss": 0.981,
+      "step": 897
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998081971657228e-06,
+      "loss": 1.2828,
+      "step": 898
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99806264780471e-06,
+      "loss": 0.9871,
+      "step": 899
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998043227115863e-06,
+      "loss": 1.2039,
+      "step": 900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998023709591063e-06,
+      "loss": 0.9744,
+      "step": 901
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998004095230685e-06,
+      "loss": 1.0893,
+      "step": 902
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997984384035114e-06,
+      "loss": 1.1543,
+      "step": 903
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997964576004727e-06,
+      "loss": 0.8936,
+      "step": 904
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997944671139912e-06,
+      "loss": 0.9328,
+      "step": 905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997924669441052e-06,
+      "loss": 0.9855,
+      "step": 906
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997904570908535e-06,
+      "loss": 1.0085,
+      "step": 907
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997884375542751e-06,
+      "loss": 0.9974,
+      "step": 908
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99786408334409e-06,
+      "loss": 1.0509,
+      "step": 909
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997843694312946e-06,
+      "loss": 1.2458,
+      "step": 910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997823208449715e-06,
+      "loss": 1.1134,
+      "step": 911
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997802625754794e-06,
+      "loss": 1.0924,
+      "step": 912
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99778194622858e-06,
+      "loss": 0.9943,
+      "step": 913
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997761169871474e-06,
+      "loss": 1.0056,
+      "step": 914
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997740296683878e-06,
+      "loss": 1.007,
+      "step": 915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997719326666199e-06,
+      "loss": 1.0131,
+      "step": 916
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997698259818843e-06,
+      "loss": 0.9103,
+      "step": 917
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997677096142213e-06,
+      "loss": 1.0823,
+      "step": 918
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997655835636727e-06,
+      "loss": 1.0114,
+      "step": 919
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99763447830279e-06,
+      "loss": 1.1933,
+      "step": 920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99761302414082e-06,
+      "loss": 1.1428,
+      "step": 921
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997591473151231e-06,
+      "loss": 0.9327,
+      "step": 922
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99756982533444e-06,
+      "loss": 1.0648,
+      "step": 923
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99754808069087e-06,
+      "loss": 1.0215,
+      "step": 924
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997526239220935e-06,
+      "loss": 0.9659,
+      "step": 925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997504300925068e-06,
+      "loss": 1.0353,
+      "step": 926
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997482265803683e-06,
+      "loss": 1.1897,
+      "step": 927
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997460133857217e-06,
+      "loss": 1.2347,
+      "step": 928
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997437905086093e-06,
+      "loss": 0.9543,
+      "step": 929
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997415579490745e-06,
+      "loss": 0.9684,
+      "step": 930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997393157071602e-06,
+      "loss": 1.1024,
+      "step": 931
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9973706378291e-06,
+      "loss": 0.8343,
+      "step": 932
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997348021763676e-06,
+      "loss": 0.9554,
+      "step": 933
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997325308875768e-06,
+      "loss": 1.0409,
+      "step": 934
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997302499165815e-06,
+      "loss": 0.9438,
+      "step": 935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997279592634261e-06,
+      "loss": 0.8907,
+      "step": 936
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997256589281548e-06,
+      "loss": 0.9369,
+      "step": 937
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997233489108122e-06,
+      "loss": 1.0443,
+      "step": 938
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99721029211443e-06,
+      "loss": 1.0013,
+      "step": 939
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997186998300922e-06,
+      "loss": 1.1749,
+      "step": 940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99716360766805e-06,
+      "loss": 1.3702,
+      "step": 941
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997140120216266e-06,
+      "loss": 1.0255,
+      "step": 942
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997116535946028e-06,
+      "loss": 1.2921,
+      "step": 943
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997092854857788e-06,
+      "loss": 1.2125,
+      "step": 944
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997069076952008e-06,
+      "loss": 0.9447,
+      "step": 945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997045202229149e-06,
+      "loss": 1.0199,
+      "step": 946
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997021230689672e-06,
+      "loss": 0.9925,
+      "step": 947
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996997162334043e-06,
+      "loss": 0.976,
+      "step": 948
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996972997162726e-06,
+      "loss": 1.1553,
+      "step": 949
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996948735176192e-06,
+      "loss": 1.0129,
+      "step": 950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99692437637491e-06,
+      "loss": 1.0563,
+      "step": 951
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99689992075935e-06,
+      "loss": 0.898,
+      "step": 952
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996875368329991e-06,
+      "loss": 1.1361,
+      "step": 953
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996850719087302e-06,
+      "loss": 1.1381,
+      "step": 954
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996825973031765e-06,
+      "loss": 0.9511,
+      "step": 955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996801130163858e-06,
+      "loss": 1.0301,
+      "step": 956
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996776190484064e-06,
+      "loss": 1.0621,
+      "step": 957
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996751153992861e-06,
+      "loss": 1.0717,
+      "step": 958
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996726020690741e-06,
+      "loss": 0.7437,
+      "step": 959
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996700790578186e-06,
+      "loss": 1.1402,
+      "step": 960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996675463655688e-06,
+      "loss": 1.1305,
+      "step": 961
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996650039923736e-06,
+      "loss": 1.0098,
+      "step": 962
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996624519382823e-06,
+      "loss": 1.048,
+      "step": 963
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996598902033442e-06,
+      "loss": 0.9208,
+      "step": 964
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996573187876093e-06,
+      "loss": 1.0285,
+      "step": 965
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996547376911271e-06,
+      "loss": 0.9138,
+      "step": 966
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996521469139476e-06,
+      "loss": 1.0341,
+      "step": 967
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996495464561211e-06,
+      "loss": 0.9861,
+      "step": 968
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996469363176982e-06,
+      "loss": 1.1339,
+      "step": 969
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99644316498729e-06,
+      "loss": 1.2277,
+      "step": 970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996416869992646e-06,
+      "loss": 1.0189,
+      "step": 971
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996390478193559e-06,
+      "loss": 1.0993,
+      "step": 972
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996363989590539e-06,
+      "loss": 1.0201,
+      "step": 973
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9963374041841e-06,
+      "loss": 0.998,
+      "step": 974
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996310721974758e-06,
+      "loss": 1.0474,
+      "step": 975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996283942963028e-06,
+      "loss": 0.9567,
+      "step": 976
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996257067149432e-06,
+      "loss": 1.1504,
+      "step": 977
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996230094534487e-06,
+      "loss": 1.0393,
+      "step": 978
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996203025118717e-06,
+      "loss": 0.9421,
+      "step": 979
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996175858902648e-06,
+      "loss": 1.1491,
+      "step": 980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996148595886803e-06,
+      "loss": 1.2052,
+      "step": 981
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996121236071714e-06,
+      "loss": 0.9983,
+      "step": 982
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996093779457909e-06,
+      "loss": 0.9323,
+      "step": 983
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99606622604592e-06,
+      "loss": 1.0397,
+      "step": 984
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996038575836282e-06,
+      "loss": 1.0772,
+      "step": 985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99601082882953e-06,
+      "loss": 1.1049,
+      "step": 986
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9959829850262e-06,
+      "loss": 1.0335,
+      "step": 987
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995955044426833e-06,
+      "loss": 1.1037,
+      "step": 988
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995927007031972e-06,
+      "loss": 1.135,
+      "step": 989
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995898872842158e-06,
+      "loss": 0.9817,
+      "step": 990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995870641857935e-06,
+      "loss": 0.998,
+      "step": 991
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995842314079854e-06,
+      "loss": 0.9996,
+      "step": 992
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995813889508459e-06,
+      "loss": 0.9097,
+      "step": 993
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995785368144304e-06,
+      "loss": 0.9941,
+      "step": 994
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995756749987942e-06,
+      "loss": 1.0614,
+      "step": 995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995728035039925e-06,
+      "loss": 1.0326,
+      "step": 996
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99569922330081e-06,
+      "loss": 1.2343,
+      "step": 997
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995670314771158e-06,
+      "loss": 1.0189,
+      "step": 998
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995641309451526e-06,
+      "loss": 1.1111,
+      "step": 999
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995612207342474e-06,
+      "loss": 1.1079,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995583008444571e-06,
+      "loss": 1.0603,
+      "step": 1001
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99555371275838e-06,
+      "loss": 1.1191,
+      "step": 1002
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995524320284469e-06,
+      "loss": 1.0743,
+      "step": 1003
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99549483102341e-06,
+      "loss": 0.8542,
+      "step": 1004
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995465244975768e-06,
+      "loss": 1.1396,
+      "step": 1005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995435562142121e-06,
+      "loss": 1.0939,
+      "step": 1006
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995405782523045e-06,
+      "loss": 1.0698,
+      "step": 1007
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995375906119112e-06,
+      "loss": 0.8862,
+      "step": 1008
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995345932930906e-06,
+      "loss": 0.9569,
+      "step": 1009
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995315862959005e-06,
+      "loss": 1.0162,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995285696203994e-06,
+      "loss": 0.9612,
+      "step": 1011
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995255432666455e-06,
+      "loss": 1.1442,
+      "step": 1012
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995225072346973e-06,
+      "loss": 1.0258,
+      "step": 1013
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99519461524614e-06,
+      "loss": 1.0517,
+      "step": 1014
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995164061364546e-06,
+      "loss": 0.952,
+      "step": 1015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995133410702778e-06,
+      "loss": 1.0726,
+      "step": 1016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995102663261435e-06,
+      "loss": 0.8607,
+      "step": 1017
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995071819041111e-06,
+      "loss": 0.9587,
+      "step": 1018
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995040878042403e-06,
+      "loss": 1.1226,
+      "step": 1019
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995009840265913e-06,
+      "loss": 1.0481,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994978705712239e-06,
+      "loss": 1.2307,
+      "step": 1021
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994947474381985e-06,
+      "loss": 1.0669,
+      "step": 1022
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994916146275756e-06,
+      "loss": 1.1077,
+      "step": 1023
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994884721394162e-06,
+      "loss": 1.2259,
+      "step": 1024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994853199737809e-06,
+      "loss": 0.8862,
+      "step": 1025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994821581307308e-06,
+      "loss": 1.0249,
+      "step": 1026
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994789866103272e-06,
+      "loss": 1.0057,
+      "step": 1027
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994758054126315e-06,
+      "loss": 0.9639,
+      "step": 1028
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994726145377055e-06,
+      "loss": 1.0124,
+      "step": 1029
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994694139856107e-06,
+      "loss": 1.1384,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994662037564094e-06,
+      "loss": 1.045,
+      "step": 1031
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994629838501637e-06,
+      "loss": 0.9715,
+      "step": 1032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99459754266936e-06,
+      "loss": 1.1229,
+      "step": 1033
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994565150067887e-06,
+      "loss": 0.8444,
+      "step": 1034
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994532660697849e-06,
+      "loss": 0.9509,
+      "step": 1035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994500074559873e-06,
+      "loss": 1.0116,
+      "step": 1036
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99446739165459e-06,
+      "loss": 1.0142,
+      "step": 1037
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994434611982636e-06,
+      "loss": 0.9076,
+      "step": 1038
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994401735544643e-06,
+      "loss": 1.119,
+      "step": 1039
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994368762341251e-06,
+      "loss": 1.0275,
+      "step": 1040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994335692373096e-06,
+      "loss": 0.9594,
+      "step": 1041
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99430252564082e-06,
+      "loss": 0.8456,
+      "step": 1042
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994269262145064e-06,
+      "loss": 1.0464,
+      "step": 1043
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994235901886475e-06,
+      "loss": 0.9886,
+      "step": 1044
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994202444865699e-06,
+      "loss": 1.0384,
+      "step": 1045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994168891083383e-06,
+      "loss": 1.3945,
+      "step": 1046
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994135240540178e-06,
+      "loss": 1.0598,
+      "step": 1047
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.994101493236736e-06,
+      "loss": 1.0855,
+      "step": 1048
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.994067649173709e-06,
+      "loss": 1.1032,
+      "step": 1049
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.994033708351754e-06,
+      "loss": 1.0814,
+      "step": 1050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99399967077153e-06,
+      "loss": 0.886,
+      "step": 1051
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993965536433694e-06,
+      "loss": 0.8104,
+      "step": 1052
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99393130533891e-06,
+      "loss": 1.1343,
+      "step": 1053
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993896977487839e-06,
+      "loss": 1.1467,
+      "step": 1054
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993862552881148e-06,
+      "loss": 1.4293,
+      "step": 1055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993828031519501e-06,
+      "loss": 1.0477,
+      "step": 1056
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99379341340357e-06,
+      "loss": 1.0159,
+      "step": 1057
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993758698534024e-06,
+      "loss": 0.948,
+      "step": 1058
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993723886911535e-06,
+      "loss": 0.9214,
+      "step": 1059
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99368897853678e-06,
+      "loss": 0.8231,
+      "step": 1060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993653973410433e-06,
+      "loss": 0.9327,
+      "step": 1061
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993618871533173e-06,
+      "loss": 0.9811,
+      "step": 1062
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99358367290568e-06,
+      "loss": 0.9826,
+      "step": 1063
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993548377528636e-06,
+      "loss": 0.9358,
+      "step": 1064
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993512985402724e-06,
+      "loss": 1.0564,
+      "step": 1065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993477496528632e-06,
+      "loss": 1.0474,
+      "step": 1066
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993441910907047e-06,
+      "loss": 1.0365,
+      "step": 1067
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993406228538657e-06,
+      "loss": 1.055,
+      "step": 1068
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993370449424153e-06,
+      "loss": 1.1523,
+      "step": 1069
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993334573564231e-06,
+      "loss": 1.0388,
+      "step": 1070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993298600959582e-06,
+      "loss": 1.0016,
+      "step": 1071
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993262531610908e-06,
+      "loss": 0.959,
+      "step": 1072
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993226365518902e-06,
+      "loss": 0.9752,
+      "step": 1073
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99319010268427e-06,
+      "loss": 1.0606,
+      "step": 1074
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993153743107711e-06,
+      "loss": 1.1511,
+      "step": 1075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993117286789932e-06,
+      "loss": 1.0342,
+      "step": 1076
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99308073373164e-06,
+      "loss": 0.962,
+      "step": 1077
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99304408393354e-06,
+      "loss": 0.8748,
+      "step": 1078
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993007337396342e-06,
+      "loss": 0.9965,
+      "step": 1079
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99297049412076e-06,
+      "loss": 0.9058,
+      "step": 1080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992933554107506e-06,
+      "loss": 0.9449,
+      "step": 1081
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9928965173573e-06,
+      "loss": 0.9351,
+      "step": 1082
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992859383870854e-06,
+      "loss": 1.0024,
+      "step": 1083
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99282215364889e-06,
+      "loss": 1.1384,
+      "step": 1084
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992784826692132e-06,
+      "loss": 0.9256,
+      "step": 1085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992747403001299e-06,
+      "loss": 1.0202,
+      "step": 1086
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992709882577118e-06,
+      "loss": 0.9734,
+      "step": 1087
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992672265420315e-06,
+      "loss": 1.0211,
+      "step": 1088
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992634551531619e-06,
+      "loss": 1.1071,
+      "step": 1089
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992596740911763e-06,
+      "loss": 0.8544,
+      "step": 1090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992558833561474e-06,
+      "loss": 1.2578,
+      "step": 1091
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992520829481494e-06,
+      "loss": 1.0131,
+      "step": 1092
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992482728672553e-06,
+      "loss": 1.0016,
+      "step": 1093
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992444531135392e-06,
+      "loss": 1.1869,
+      "step": 1094
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99240623687075e-06,
+      "loss": 1.1379,
+      "step": 1095
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992367845879372e-06,
+      "loss": 1.011,
+      "step": 1096
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992329358161998e-06,
+      "loss": 1.1824,
+      "step": 1097
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992290773719376e-06,
+      "loss": 0.9925,
+      "step": 1098
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992252092552252e-06,
+      "loss": 1.0217,
+      "step": 1099
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992213314661376e-06,
+      "loss": 0.9361,
+      "step": 1100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9921744400475e-06,
+      "loss": 0.8313,
+      "step": 1101
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992135468711375e-06,
+      "loss": 1.0076,
+      "step": 1102
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99209640065376e-06,
+      "loss": 1.0821,
+      "step": 1103
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992057235875407e-06,
+      "loss": 0.9654,
+      "step": 1104
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99201797437708e-06,
+      "loss": 1.0409,
+      "step": 1105
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991978616159535e-06,
+      "loss": 1.0716,
+      "step": 1106
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991939161223538e-06,
+      "loss": 0.9844,
+      "step": 1107
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991899609569853e-06,
+      "loss": 0.8209,
+      "step": 1108
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991859961199244e-06,
+      "loss": 0.934,
+      "step": 1109
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99182021611248e-06,
+      "loss": 1.1997,
+      "step": 1110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991780374310334e-06,
+      "loss": 0.9978,
+      "step": 1111
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991740435793573e-06,
+      "loss": 1.0448,
+      "step": 1112
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991700400562975e-06,
+      "loss": 0.8346,
+      "step": 1113
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991660268619313e-06,
+      "loss": 1.0388,
+      "step": 1114
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991620039963368e-06,
+      "loss": 0.8855,
+      "step": 1115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991579714595914e-06,
+      "loss": 0.8438,
+      "step": 1116
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991539292517735e-06,
+      "loss": 1.163,
+      "step": 1117
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991498773729617e-06,
+      "loss": 1.0825,
+      "step": 1118
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99145815823234e-06,
+      "loss": 0.9513,
+      "step": 1119
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991417446026697e-06,
+      "loss": 0.9312,
+      "step": 1120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99137663711347e-06,
+      "loss": 1.175,
+      "step": 1121
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991335731493455e-06,
+      "loss": 1.0098,
+      "step": 1122
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99129472916744e-06,
+      "loss": 1.0132,
+      "step": 1123
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991253630136224e-06,
+      "loss": 1.085,
+      "step": 1124
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991212434400601e-06,
+      "loss": 0.9091,
+      "step": 1125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99117114196137e-06,
+      "loss": 1.0548,
+      "step": 1126
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991129752819327e-06,
+      "loss": 1.1541,
+      "step": 1127
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99108826697528e-06,
+      "loss": 1.0838,
+      "step": 1128
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991046684430029e-06,
+      "loss": 1.188,
+      "step": 1129
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99100500518438e-06,
+      "loss": 1.0814,
+      "step": 1130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990963229239144e-06,
+      "loss": 1.0456,
+      "step": 1131
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990921356595125e-06,
+      "loss": 1.0923,
+      "step": 1132
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990879387253137e-06,
+      "loss": 1.0487,
+      "step": 1133
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990837321213993e-06,
+      "loss": 1.0069,
+      "step": 1134
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990795158478508e-06,
+      "loss": 1.1203,
+      "step": 1135
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9907528990475e-06,
+      "loss": 1.0261,
+      "step": 1136
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990710542921786e-06,
+      "loss": 1.0695,
+      "step": 1137
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990668090102186e-06,
+      "loss": 1.0829,
+      "step": 1138
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990625540589528e-06,
+      "loss": 1.0165,
+      "step": 1139
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990582894384628e-06,
+      "loss": 1.0577,
+      "step": 1140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990540151488317e-06,
+      "loss": 1.2029,
+      "step": 1141
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990497311901423e-06,
+      "loss": 1.1312,
+      "step": 1142
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990454375624778e-06,
+      "loss": 1.0425,
+      "step": 1143
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990411342659208e-06,
+      "loss": 1.1479,
+      "step": 1144
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990368213005553e-06,
+      "loss": 1.1496,
+      "step": 1145
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990324986664645e-06,
+      "loss": 1.0982,
+      "step": 1146
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990281663637322e-06,
+      "loss": 1.0792,
+      "step": 1147
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990238243924423e-06,
+      "loss": 0.8962,
+      "step": 1148
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990194727526792e-06,
+      "loss": 1.0875,
+      "step": 1149
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990151114445268e-06,
+      "loss": 1.0095,
+      "step": 1150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990107404680699e-06,
+      "loss": 1.0055,
+      "step": 1151
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99006359823393e-06,
+      "loss": 1.1252,
+      "step": 1152
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990019695105812e-06,
+      "loss": 0.9168,
+      "step": 1153
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989975695297196e-06,
+      "loss": 1.2295,
+      "step": 1154
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989931598808931e-06,
+      "loss": 1.09,
+      "step": 1155
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989887405641871e-06,
+      "loss": 0.9398,
+      "step": 1156
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989843115796878e-06,
+      "loss": 1.053,
+      "step": 1157
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989798729274805e-06,
+      "loss": 0.9338,
+      "step": 1158
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989754246076513e-06,
+      "loss": 1.0431,
+      "step": 1159
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989709666202865e-06,
+      "loss": 1.0963,
+      "step": 1160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989664989654724e-06,
+      "loss": 0.9838,
+      "step": 1161
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989620216432955e-06,
+      "loss": 1.0753,
+      "step": 1162
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989575346538428e-06,
+      "loss": 1.1183,
+      "step": 1163
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989530379972008e-06,
+      "loss": 0.9973,
+      "step": 1164
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989485316734571e-06,
+      "loss": 1.2052,
+      "step": 1165
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989440156826985e-06,
+      "loss": 1.1476,
+      "step": 1166
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989394900250132e-06,
+      "loss": 0.9637,
+      "step": 1167
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98934954700488e-06,
+      "loss": 0.9295,
+      "step": 1168
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989304097092116e-06,
+      "loss": 0.9559,
+      "step": 1169
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989258550512714e-06,
+      "loss": 1.0926,
+      "step": 1170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989212907267561e-06,
+      "loss": 1.0406,
+      "step": 1171
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989167167357538e-06,
+      "loss": 0.9558,
+      "step": 1172
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989121330783535e-06,
+      "loss": 0.9706,
+      "step": 1173
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989075397546437e-06,
+      "loss": 1.1449,
+      "step": 1174
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989029367647136e-06,
+      "loss": 0.9831,
+      "step": 1175
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988983241086521e-06,
+      "loss": 1.1985,
+      "step": 1176
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988937017865488e-06,
+      "loss": 0.9381,
+      "step": 1177
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988890697984931e-06,
+      "loss": 1.0219,
+      "step": 1178
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98884428144575e-06,
+      "loss": 0.9285,
+      "step": 1179
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988797768248844e-06,
+      "loss": 1.01,
+      "step": 1180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98875115839511e-06,
+      "loss": 1.0129,
+      "step": 1181
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988704451885454e-06,
+      "loss": 1.1123,
+      "step": 1182
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988657648720783e-06,
+      "loss": 0.8861,
+      "step": 1183
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988610748902e-06,
+      "loss": 0.9338,
+      "step": 1184
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988563752430016e-06,
+      "loss": 1.0241,
+      "step": 1185
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98851665930574e-06,
+      "loss": 0.9633,
+      "step": 1186
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988469469530086e-06,
+      "loss": 0.9393,
+      "step": 1187
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988422183103969e-06,
+      "loss": 1.1925,
+      "step": 1188
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988374800028302e-06,
+      "loss": 0.9194,
+      "step": 1189
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988327320304006e-06,
+      "loss": 1.0979,
+      "step": 1190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988279743931998e-06,
+      "loss": 1.0654,
+      "step": 1191
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988232070913201e-06,
+      "loss": 1.0013,
+      "step": 1192
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988184301248543e-06,
+      "loss": 1.2075,
+      "step": 1193
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988136434938942e-06,
+      "loss": 1.2122,
+      "step": 1194
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988088471985331e-06,
+      "loss": 0.9624,
+      "step": 1195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988040412388635e-06,
+      "loss": 1.0627,
+      "step": 1196
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98799225614979e-06,
+      "loss": 1.1087,
+      "step": 1197
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987944003269725e-06,
+      "loss": 0.8375,
+      "step": 1198
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987895653749378e-06,
+      "loss": 0.9391,
+      "step": 1199
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987847207589682e-06,
+      "loss": 1.1325,
+      "step": 1200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98779866479158e-06,
+      "loss": 0.9196,
+      "step": 1201
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987750025356009e-06,
+      "loss": 1.0385,
+      "step": 1202
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987701289283914e-06,
+      "loss": 1.0152,
+      "step": 1203
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987652456576235e-06,
+      "loss": 1.0564,
+      "step": 1204
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987603527233924e-06,
+      "loss": 0.9224,
+      "step": 1205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987554501257925e-06,
+      "loss": 1.1054,
+      "step": 1206
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98750537864919e-06,
+      "loss": 1.0544,
+      "step": 1207
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987456159408667e-06,
+      "loss": 1.1567,
+      "step": 1208
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987406843537313e-06,
+      "loss": 1.092,
+      "step": 1209
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987357431036084e-06,
+      "loss": 1.1202,
+      "step": 1210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987307921905936e-06,
+      "loss": 0.8314,
+      "step": 1211
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987258316147828e-06,
+      "loss": 0.8815,
+      "step": 1212
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987208613762721e-06,
+      "loss": 1.107,
+      "step": 1213
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987158814751578e-06,
+      "loss": 0.9078,
+      "step": 1214
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987108919115366e-06,
+      "loss": 0.9025,
+      "step": 1215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987058926855048e-06,
+      "loss": 1.0378,
+      "step": 1216
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987008837971595e-06,
+      "loss": 1.0747,
+      "step": 1217
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986958652465979e-06,
+      "loss": 0.7557,
+      "step": 1218
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986908370339169e-06,
+      "loss": 1.0543,
+      "step": 1219
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986857991592139e-06,
+      "loss": 1.0091,
+      "step": 1220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98680751622587e-06,
+      "loss": 1.0364,
+      "step": 1221
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986756944241334e-06,
+      "loss": 0.9308,
+      "step": 1222
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986706275639512e-06,
+      "loss": 0.9602,
+      "step": 1223
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98665551042139e-06,
+      "loss": 1.003,
+      "step": 1224
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986604648587947e-06,
+      "loss": 1.0448,
+      "step": 1225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986553690140172e-06,
+      "loss": 1.0887,
+      "step": 1226
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986502635079048e-06,
+      "loss": 1.1535,
+      "step": 1227
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986451483405568e-06,
+      "loss": 1.1125,
+      "step": 1228
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986400235120722e-06,
+      "loss": 1.1911,
+      "step": 1229
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986348890225501e-06,
+      "loss": 1.1032,
+      "step": 1230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986297448720903e-06,
+      "loss": 0.8746,
+      "step": 1231
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98624591060792e-06,
+      "loss": 0.9845,
+      "step": 1232
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986194275887557e-06,
+      "loss": 0.9033,
+      "step": 1233
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986142544560811e-06,
+      "loss": 1.0156,
+      "step": 1234
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986090716628683e-06,
+      "loss": 1.0911,
+      "step": 1235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986038792092179e-06,
+      "loss": 1.1742,
+      "step": 1236
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985986770952303e-06,
+      "loss": 1.0805,
+      "step": 1237
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985934653210066e-06,
+      "loss": 1.0859,
+      "step": 1238
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985882438866474e-06,
+      "loss": 1.0956,
+      "step": 1239
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985830127922542e-06,
+      "loss": 0.9538,
+      "step": 1240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985777720379281e-06,
+      "loss": 1.0677,
+      "step": 1241
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98572521623771e-06,
+      "loss": 0.9234,
+      "step": 1242
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985672615498843e-06,
+      "loss": 0.8179,
+      "step": 1243
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985619918163701e-06,
+      "loss": 1.0348,
+      "step": 1244
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985567124233302e-06,
+      "loss": 0.8514,
+      "step": 1245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985514233708672e-06,
+      "loss": 0.9633,
+      "step": 1246
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985461246590833e-06,
+      "loss": 1.0533,
+      "step": 1247
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985408162880813e-06,
+      "loss": 0.953,
+      "step": 1248
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985354982579646e-06,
+      "loss": 1.0537,
+      "step": 1249
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985301705688351e-06,
+      "loss": 1.1345,
+      "step": 1250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985248332207969e-06,
+      "loss": 0.9451,
+      "step": 1251
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985194862139532e-06,
+      "loss": 0.9997,
+      "step": 1252
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985141295484073e-06,
+      "loss": 1.0781,
+      "step": 1253
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985087632242634e-06,
+      "loss": 1.1724,
+      "step": 1254
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985033872416252e-06,
+      "loss": 0.9696,
+      "step": 1255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98498001600597e-06,
+      "loss": 1.0287,
+      "step": 1256
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984926063012832e-06,
+      "loss": 1.1196,
+      "step": 1257
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984872013437881e-06,
+      "loss": 1.0062,
+      "step": 1258
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984817867282166e-06,
+      "loss": 0.9938,
+      "step": 1259
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984763624546737e-06,
+      "loss": 1.1263,
+      "step": 1260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984709285232642e-06,
+      "loss": 1.0297,
+      "step": 1261
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984654849340933e-06,
+      "loss": 0.9684,
+      "step": 1262
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98460031687267e-06,
+      "loss": 1.045,
+      "step": 1263
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984545687828906e-06,
+      "loss": 0.8292,
+      "step": 1264
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9844909622107e-06,
+      "loss": 1.0621,
+      "step": 1265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98443614001911e-06,
+      "loss": 1.0483,
+      "step": 1266
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984381221255204e-06,
+      "loss": 1.2381,
+      "step": 1267
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98432620592004e-06,
+      "loss": 1.0209,
+      "step": 1268
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984271094014688e-06,
+      "loss": 1.1693,
+      "step": 1269
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984215885540213e-06,
+      "loss": 1.1167,
+      "step": 1270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984160580497688e-06,
+      "loss": 1.0522,
+      "step": 1271
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98410517888818e-06,
+      "loss": 1.0579,
+      "step": 1272
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984049680712767e-06,
+      "loss": 0.7756,
+      "step": 1273
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98399408597252e-06,
+      "loss": 1.0793,
+      "step": 1274
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983938394668517e-06,
+      "loss": 1.1383,
+      "step": 1275
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98388260680184e-06,
+      "loss": 1.0703,
+      "step": 1276
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983826722373569e-06,
+      "loss": 1.3018,
+      "step": 1277
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983770741384785e-06,
+      "loss": 1.0689,
+      "step": 1278
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983714663836572e-06,
+      "loss": 1.0697,
+      "step": 1279
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983658489730019e-06,
+      "loss": 1.1347,
+      "step": 1280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983602219066213e-06,
+      "loss": 0.9966,
+      "step": 1281
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983545851846244e-06,
+      "loss": 0.8994,
+      "step": 1282
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983489388071206e-06,
+      "loss": 0.861,
+      "step": 1283
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98343282774219e-06,
+      "loss": 1.0073,
+      "step": 1284
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983376170860296e-06,
+      "loss": 0.9207,
+      "step": 1285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983319417426615e-06,
+      "loss": 0.9494,
+      "step": 1286
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983262567442253e-06,
+      "loss": 0.7411,
+      "step": 1287
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98320562090831e-06,
+      "loss": 1.0677,
+      "step": 1288
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983148577825887e-06,
+      "loss": 1.0511,
+      "step": 1289
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983091438196091e-06,
+      "loss": 1.1935,
+      "step": 1290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98303420202003e-06,
+      "loss": 0.9624,
+      "step": 1291
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982976869298811e-06,
+      "loss": 1.0394,
+      "step": 1292
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982919440033544e-06,
+      "loss": 1.1311,
+      "step": 1293
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982861914225344e-06,
+      "loss": 0.9983,
+      "step": 1294
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982804291875325e-06,
+      "loss": 1.2567,
+      "step": 1295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982746572984603e-06,
+      "loss": 0.9722,
+      "step": 1296
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982688757554298e-06,
+      "loss": 0.9949,
+      "step": 1297
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982630845585526e-06,
+      "loss": 0.9912,
+      "step": 1298
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982572837079412e-06,
+      "loss": 1.0871,
+      "step": 1299
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98251473203708e-06,
+      "loss": 1.177,
+      "step": 1300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982456530459656e-06,
+      "loss": 0.9569,
+      "step": 1301
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982398232348267e-06,
+      "loss": 1.2235,
+      "step": 1302
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982339837704042e-06,
+      "loss": 0.9165,
+      "step": 1303
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982281346528113e-06,
+      "loss": 1.1108,
+      "step": 1304
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982222758821612e-06,
+      "loss": 0.9479,
+      "step": 1305
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982164074585677e-06,
+      "loss": 0.9742,
+      "step": 1306
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982105293821443e-06,
+      "loss": 1.1076,
+      "step": 1307
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982046416530047e-06,
+      "loss": 1.2262,
+      "step": 1308
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981987442712634e-06,
+      "loss": 1.1426,
+      "step": 1309
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981928372370342e-06,
+      "loss": 0.9518,
+      "step": 1310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98186920550432e-06,
+      "loss": 1.0611,
+      "step": 1311
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981809942115712e-06,
+      "loss": 1.0123,
+      "step": 1312
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981750582205667e-06,
+      "loss": 0.9549,
+      "step": 1313
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981691125775333e-06,
+      "loss": 1.0426,
+      "step": 1314
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981631572825862e-06,
+      "loss": 1.159,
+      "step": 1315
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981571923358414e-06,
+      "loss": 1.1206,
+      "step": 1316
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981512177374136e-06,
+      "loss": 0.9871,
+      "step": 1317
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98145233487419e-06,
+      "loss": 1.1013,
+      "step": 1318
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981392395859737e-06,
+      "loss": 1.0848,
+      "step": 1319
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981332360331934e-06,
+      "loss": 0.909,
+      "step": 1320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98127222829195e-06,
+      "loss": 1.0606,
+      "step": 1321
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981211999740943e-06,
+      "loss": 0.9193,
+      "step": 1322
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981151674680085e-06,
+      "loss": 0.9692,
+      "step": 1323
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981091253110542e-06,
+      "loss": 1.0356,
+      "step": 1324
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981030735033487e-06,
+      "loss": 0.888,
+      "step": 1325
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98097012045009e-06,
+      "loss": 0.9117,
+      "step": 1326
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980909409361529e-06,
+      "loss": 1.1996,
+      "step": 1327
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980848601768976e-06,
+      "loss": 0.8925,
+      "step": 1328
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980787697673612e-06,
+      "loss": 0.728,
+      "step": 1329
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980726697076616e-06,
+      "loss": 1.2137,
+      "step": 1330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980665599979171e-06,
+      "loss": 0.9533,
+      "step": 1331
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980604406382459e-06,
+      "loss": 0.9246,
+      "step": 1332
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980543116287666e-06,
+      "loss": 0.9504,
+      "step": 1333
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98048172969598e-06,
+      "loss": 0.9069,
+      "step": 1334
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980420246608592e-06,
+      "loss": 1.1804,
+      "step": 1335
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980358667026691e-06,
+      "loss": 1.029,
+      "step": 1336
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98029699095147e-06,
+      "loss": 1.0158,
+      "step": 1337
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980235218384124e-06,
+      "loss": 0.9654,
+      "step": 1338
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98017334932585e-06,
+      "loss": 1.0448,
+      "step": 1339
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980111383777848e-06,
+      "loss": 1.0758,
+      "step": 1340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980049321741318e-06,
+      "loss": 1.0167,
+      "step": 1341
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97998716321746e-06,
+      "loss": 1.0028,
+      "step": 1342
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979924908207483e-06,
+      "loss": 1.0162,
+      "step": 1343
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979862556712588e-06,
+      "loss": 1.0964,
+      "step": 1344
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979800108733988e-06,
+      "loss": 0.8911,
+      "step": 1345
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979737564272888e-06,
+      "loss": 0.9857,
+      "step": 1346
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979674923330505e-06,
+      "loss": 1.0102,
+      "step": 1347
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979612185908048e-06,
+      "loss": 1.1424,
+      "step": 1348
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979549352006734e-06,
+      "loss": 1.0649,
+      "step": 1349
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979486421627783e-06,
+      "loss": 1.0293,
+      "step": 1350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97942339477241e-06,
+      "loss": 0.998,
+      "step": 1351
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97936027144184e-06,
+      "loss": 1.0061,
+      "step": 1352
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979297051637293e-06,
+      "loss": 0.9027,
+      "step": 1353
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979233735359996e-06,
+      "loss": 0.8909,
+      "step": 1354
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979170322611174e-06,
+      "loss": 0.846,
+      "step": 1355
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979106813392058e-06,
+      "loss": 0.9046,
+      "step": 1356
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979043207703876e-06,
+      "loss": 1.0848,
+      "step": 1357
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978979505547862e-06,
+      "loss": 1.0259,
+      "step": 1358
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978915706925247e-06,
+      "loss": 1.1552,
+      "step": 1359
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978851811837271e-06,
+      "loss": 0.9018,
+      "step": 1360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978787820285173e-06,
+      "loss": 0.9607,
+      "step": 1361
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978723732270186e-06,
+      "loss": 0.9256,
+      "step": 1362
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978659547793559e-06,
+      "loss": 1.0503,
+      "step": 1363
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978595266856532e-06,
+      "loss": 1.0428,
+      "step": 1364
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978530889460351e-06,
+      "loss": 1.1035,
+      "step": 1365
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978466415606263e-06,
+      "loss": 1.0783,
+      "step": 1366
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978401845295519e-06,
+      "loss": 0.9523,
+      "step": 1367
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978337178529365e-06,
+      "loss": 1.0462,
+      "step": 1368
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97827241530906e-06,
+      "loss": 0.9305,
+      "step": 1369
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978207555635856e-06,
+      "loss": 1.2199,
+      "step": 1370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97814259951101e-06,
+      "loss": 1.1109,
+      "step": 1371
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97807754693578e-06,
+      "loss": 0.9464,
+      "step": 1372
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978012397911428e-06,
+      "loss": 0.8877,
+      "step": 1373
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977947152439212e-06,
+      "loss": 0.9081,
+      "step": 1374
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977881810520402e-06,
+      "loss": 1.0067,
+      "step": 1375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97781637215626e-06,
+      "loss": 1.1925,
+      "step": 1376
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977750837348055e-06,
+      "loss": 0.957,
+      "step": 1377
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977685206097057e-06,
+      "loss": 1.0478,
+      "step": 1378
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977619478404537e-06,
+      "loss": 1.0465,
+      "step": 1379
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977553654271768e-06,
+      "loss": 1.1151,
+      "step": 1380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977487733700027e-06,
+      "loss": 0.8936,
+      "step": 1381
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97742171669059e-06,
+      "loss": 1.0532,
+      "step": 1382
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977355603244736e-06,
+      "loss": 1.072,
+      "step": 1383
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977289393363746e-06,
+      "loss": 1.1088,
+      "step": 1384
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977223087048905e-06,
+      "loss": 1.1783,
+      "step": 1385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977156684301492e-06,
+      "loss": 1.0266,
+      "step": 1386
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9770901851228e-06,
+      "loss": 1.1985,
+      "step": 1387
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977023589514114e-06,
+      "loss": 0.9867,
+      "step": 1388
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976956897476726e-06,
+      "loss": 0.9972,
+      "step": 1389
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976890109011926e-06,
+      "loss": 0.9812,
+      "step": 1390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976823224121008e-06,
+      "loss": 1.0226,
+      "step": 1391
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976756242805271e-06,
+      "loss": 0.8854,
+      "step": 1392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976689165066009e-06,
+      "loss": 1.0008,
+      "step": 1393
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976621990904525e-06,
+      "loss": 0.9737,
+      "step": 1394
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976554720322119e-06,
+      "loss": 1.0501,
+      "step": 1395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976487353320092e-06,
+      "loss": 1.1041,
+      "step": 1396
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976419889899753e-06,
+      "loss": 0.9799,
+      "step": 1397
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976352330062407e-06,
+      "loss": 0.8935,
+      "step": 1398
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976284673809364e-06,
+      "loss": 0.9916,
+      "step": 1399
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976216921141933e-06,
+      "loss": 1.117,
+      "step": 1400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97614907206143e-06,
+      "loss": 0.899,
+      "step": 1401
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976081126569164e-06,
+      "loss": 1.0395,
+      "step": 1402
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976013084666459e-06,
+      "loss": 1.0697,
+      "step": 1403
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975944946354627e-06,
+      "loss": 1.1433,
+      "step": 1404
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975876711634991e-06,
+      "loss": 0.9365,
+      "step": 1405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975808380508871e-06,
+      "loss": 1.0223,
+      "step": 1406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975739952977594e-06,
+      "loss": 1.0259,
+      "step": 1407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975671429042483e-06,
+      "loss": 0.9837,
+      "step": 1408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975602808704868e-06,
+      "loss": 1.0457,
+      "step": 1409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975534091966076e-06,
+      "loss": 0.8928,
+      "step": 1410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975465278827441e-06,
+      "loss": 1.113,
+      "step": 1411
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975396369290294e-06,
+      "loss": 0.9196,
+      "step": 1412
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975327363355972e-06,
+      "loss": 1.012,
+      "step": 1413
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97525826102581e-06,
+      "loss": 0.9792,
+      "step": 1414
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975189062301147e-06,
+      "loss": 1.0655,
+      "step": 1415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975119767183325e-06,
+      "loss": 0.9127,
+      "step": 1416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975050375673686e-06,
+      "loss": 0.8483,
+      "step": 1417
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974980887773576e-06,
+      "loss": 1.0339,
+      "step": 1418
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974911303484338e-06,
+      "loss": 0.9551,
+      "step": 1419
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974841622807323e-06,
+      "loss": 1.1659,
+      "step": 1420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974771845743879e-06,
+      "loss": 1.0523,
+      "step": 1421
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974701972295358e-06,
+      "loss": 1.1325,
+      "step": 1422
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974632002463118e-06,
+      "loss": 1.2227,
+      "step": 1423
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97456193624851e-06,
+      "loss": 0.9315,
+      "step": 1424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974491773652892e-06,
+      "loss": 0.9503,
+      "step": 1425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974421514677623e-06,
+      "loss": 0.8163,
+      "step": 1426
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97435115932407e-06,
+      "loss": 1.0037,
+      "step": 1427
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974280707593588e-06,
+      "loss": 0.9008,
+      "step": 1428
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974210159487546e-06,
+      "loss": 1.1154,
+      "step": 1429
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974139515007311e-06,
+      "loss": 1.2123,
+      "step": 1430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974068774154252e-06,
+      "loss": 1.0701,
+      "step": 1431
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973997936929738e-06,
+      "loss": 1.1387,
+      "step": 1432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973927003335141e-06,
+      "loss": 1.0068,
+      "step": 1433
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97385597337184e-06,
+      "loss": 0.9557,
+      "step": 1434
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973784847041206e-06,
+      "loss": 1.0224,
+      "step": 1435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973713624344619e-06,
+      "loss": 0.7641,
+      "step": 1436
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973642305283458e-06,
+      "loss": 1.0351,
+      "step": 1437
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973570889859106e-06,
+      "loss": 0.9058,
+      "step": 1438
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973499378072947e-06,
+      "loss": 1.0011,
+      "step": 1439
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973427769926363e-06,
+      "loss": 0.8851,
+      "step": 1440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973356065420747e-06,
+      "loss": 0.9678,
+      "step": 1441
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973284264557484e-06,
+      "loss": 1.1258,
+      "step": 1442
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973212367337968e-06,
+      "loss": 1.0369,
+      "step": 1443
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973140373763588e-06,
+      "loss": 0.9218,
+      "step": 1444
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973068283835743e-06,
+      "loss": 1.1823,
+      "step": 1445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972996097555829e-06,
+      "loss": 0.9402,
+      "step": 1446
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972923814925242e-06,
+      "loss": 0.942,
+      "step": 1447
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972851435945386e-06,
+      "loss": 1.0898,
+      "step": 1448
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97277896061766e-06,
+      "loss": 1.1373,
+      "step": 1449
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972706388943469e-06,
+      "loss": 1.0681,
+      "step": 1450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972633720924221e-06,
+      "loss": 0.9864,
+      "step": 1451
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972560956561321e-06,
+      "loss": 1.1519,
+      "step": 1452
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972488095856181e-06,
+      "loss": 0.9663,
+      "step": 1453
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972415138810213e-06,
+      "loss": 1.0518,
+      "step": 1454
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97234208542483e-06,
+      "loss": 0.9149,
+      "step": 1455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972268935701446e-06,
+      "loss": 0.8628,
+      "step": 1456
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97219568964148e-06,
+      "loss": 1.0578,
+      "step": 1457
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97212234724635e-06,
+      "loss": 1.0603,
+      "step": 1458
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972048908517477e-06,
+      "loss": 0.9707,
+      "step": 1459
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971975373456285e-06,
+      "loss": 1.1344,
+      "step": 1460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971901742064196e-06,
+      "loss": 1.0912,
+      "step": 1461
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971828014342641e-06,
+      "loss": 1.0075,
+      "step": 1462
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971754190293045e-06,
+      "loss": 0.9413,
+      "step": 1463
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971680269916839e-06,
+      "loss": 1.088,
+      "step": 1464
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971606253215457e-06,
+      "loss": 0.9725,
+      "step": 1465
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97153214019033e-06,
+      "loss": 0.9165,
+      "step": 1466
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971457930842897e-06,
+      "loss": 1.0519,
+      "step": 1467
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971383625174594e-06,
+      "loss": 1.0474,
+      "step": 1468
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97130922318686e-06,
+      "loss": 0.9817,
+      "step": 1469
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971234724881137e-06,
+      "loss": 0.9689,
+      "step": 1470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97116013025887e-06,
+      "loss": 1.0246,
+      "step": 1471
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971085439321503e-06,
+      "loss": 1.0399,
+      "step": 1472
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971010652070483e-06,
+      "loss": 0.8817,
+      "step": 1473
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970935768507259e-06,
+      "loss": 0.861,
+      "step": 1474
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970860788633284e-06,
+      "loss": 0.9395,
+      "step": 1475
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970785712450007e-06,
+      "loss": 0.8813,
+      "step": 1476
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970710539958884e-06,
+      "loss": 1.0543,
+      "step": 1477
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970635271161372e-06,
+      "loss": 1.0626,
+      "step": 1478
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97055990605893e-06,
+      "loss": 1.1309,
+      "step": 1479
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970484444653017e-06,
+      "loss": 0.9788,
+      "step": 1480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970408886945094e-06,
+      "loss": 1.0345,
+      "step": 1481
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970333232936629e-06,
+      "loss": 0.9667,
+      "step": 1482
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970257482629083e-06,
+      "loss": 0.934,
+      "step": 1483
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970181636023926e-06,
+      "loss": 0.9922,
+      "step": 1484
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970105693122627e-06,
+      "loss": 0.9033,
+      "step": 1485
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970029653926658e-06,
+      "loss": 0.9903,
+      "step": 1486
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969953518437492e-06,
+      "loss": 1.028,
+      "step": 1487
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969877286656605e-06,
+      "loss": 0.7811,
+      "step": 1488
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969800958585471e-06,
+      "loss": 1.1164,
+      "step": 1489
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969724534225573e-06,
+      "loss": 0.909,
+      "step": 1490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969648013578389e-06,
+      "loss": 0.9435,
+      "step": 1491
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9695713966454e-06,
+      "loss": 0.9646,
+      "step": 1492
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969494683428093e-06,
+      "loss": 0.8881,
+      "step": 1493
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969417873927956e-06,
+      "loss": 0.8383,
+      "step": 1494
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969340968146472e-06,
+      "loss": 1.0542,
+      "step": 1495
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969263966085136e-06,
+      "loss": 1.0855,
+      "step": 1496
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969186867745437e-06,
+      "loss": 1.0254,
+      "step": 1497
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96910967312887e-06,
+      "loss": 1.0021,
+      "step": 1498
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969032382236931e-06,
+      "loss": 1.0812,
+      "step": 1499
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968954995071116e-06,
+      "loss": 1.073,
+      "step": 1500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968877511632924e-06,
+      "loss": 1.1944,
+      "step": 1501
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968799931923859e-06,
+      "loss": 0.9292,
+      "step": 1502
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968722255945422e-06,
+      "loss": 1.072,
+      "step": 1503
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968644483699118e-06,
+      "loss": 0.9798,
+      "step": 1504
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968566615186455e-06,
+      "loss": 1.0532,
+      "step": 1505
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96848865040894e-06,
+      "loss": 0.9676,
+      "step": 1506
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968410589368084e-06,
+      "loss": 1.103,
+      "step": 1507
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968332432065403e-06,
+      "loss": 1.1191,
+      "step": 1508
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968254178502404e-06,
+      "loss": 1.0548,
+      "step": 1509
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968175828680608e-06,
+      "loss": 0.9935,
+      "step": 1510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968097382601534e-06,
+      "loss": 1.0504,
+      "step": 1511
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9680188402667e-06,
+      "loss": 0.9542,
+      "step": 1512
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967940201677628e-06,
+      "loss": 0.9075,
+      "step": 1513
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967861466835839e-06,
+      "loss": 1.0754,
+      "step": 1514
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967782635742864e-06,
+      "loss": 1.0553,
+      "step": 1515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967703708400226e-06,
+      "loss": 1.1681,
+      "step": 1516
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967624684809456e-06,
+      "loss": 0.9416,
+      "step": 1517
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967545564972083e-06,
+      "loss": 1.0385,
+      "step": 1518
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967466348889643e-06,
+      "loss": 1.0161,
+      "step": 1519
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96738703656367e-06,
+      "loss": 0.8231,
+      "step": 1520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9673076279957e-06,
+      "loss": 1.1238,
+      "step": 1521
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96722812318727e-06,
+      "loss": 0.8631,
+      "step": 1522
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967148522139923e-06,
+      "loss": 0.9526,
+      "step": 1523
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967068824855198e-06,
+      "loss": 1.0223,
+      "step": 1524
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966989031334643e-06,
+      "loss": 0.9378,
+      "step": 1525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966909141579801e-06,
+      "loss": 0.9979,
+      "step": 1526
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966829155592224e-06,
+      "loss": 0.9261,
+      "step": 1527
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966749073373456e-06,
+      "loss": 1.0031,
+      "step": 1528
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96666889492505e-06,
+      "loss": 0.8838,
+      "step": 1529
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966588620248565e-06,
+      "loss": 0.8856,
+      "step": 1530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96650824934555e-06,
+      "loss": 0.9572,
+      "step": 1531
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966427782217565e-06,
+      "loss": 0.9929,
+      "step": 1532
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966347218866166e-06,
+      "loss": 1.0127,
+      "step": 1533
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96626655929292e-06,
+      "loss": 1.1015,
+      "step": 1534
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966185803499382e-06,
+      "loss": 0.9714,
+      "step": 1535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966104951487123e-06,
+      "loss": 1.0943,
+      "step": 1536
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966024003257705e-06,
+      "loss": 1.076,
+      "step": 1537
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965942958812701e-06,
+      "loss": 1.0039,
+      "step": 1538
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965861818153676e-06,
+      "loss": 1.1493,
+      "step": 1539
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965780581282207e-06,
+      "loss": 0.9718,
+      "step": 1540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965699248199865e-06,
+      "loss": 0.9485,
+      "step": 1541
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965617818908226e-06,
+      "loss": 1.076,
+      "step": 1542
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965536293408869e-06,
+      "loss": 1.1225,
+      "step": 1543
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965454671703372e-06,
+      "loss": 0.9822,
+      "step": 1544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965372953793318e-06,
+      "loss": 1.0346,
+      "step": 1545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965291139680288e-06,
+      "loss": 1.0287,
+      "step": 1546
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965209229365868e-06,
+      "loss": 0.9934,
+      "step": 1547
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965127222851647e-06,
+      "loss": 1.0801,
+      "step": 1548
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965045120139211e-06,
+      "loss": 0.9303,
+      "step": 1549
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964962921230155e-06,
+      "loss": 0.8626,
+      "step": 1550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964880626126066e-06,
+      "loss": 1.0199,
+      "step": 1551
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96479823482854e-06,
+      "loss": 1.0383,
+      "step": 1552
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964715747339178e-06,
+      "loss": 1.0284,
+      "step": 1553
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964633163659572e-06,
+      "loss": 1.0346,
+      "step": 1554
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964550483791326e-06,
+      "loss": 0.9979,
+      "step": 1555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964467707736039e-06,
+      "loss": 1.0212,
+      "step": 1556
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964384835495317e-06,
+      "loss": 0.8916,
+      "step": 1557
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964301867070765e-06,
+      "loss": 1.0287,
+      "step": 1558
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96421880246399e-06,
+      "loss": 0.9067,
+      "step": 1559
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964135641676603e-06,
+      "loss": 0.9941,
+      "step": 1560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964052384710214e-06,
+      "loss": 0.9513,
+      "step": 1561
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963969031566434e-06,
+      "loss": 0.9447,
+      "step": 1562
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963885582246882e-06,
+      "loss": 0.9066,
+      "step": 1563
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963802036753171e-06,
+      "loss": 0.984,
+      "step": 1564
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963718395086922e-06,
+      "loss": 0.8958,
+      "step": 1565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963634657249757e-06,
+      "loss": 0.974,
+      "step": 1566
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963550823243296e-06,
+      "loss": 1.0253,
+      "step": 1567
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963466893069163e-06,
+      "loss": 1.0019,
+      "step": 1568
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963382866728985e-06,
+      "loss": 1.0783,
+      "step": 1569
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963298744224389e-06,
+      "loss": 0.9111,
+      "step": 1570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963214525557006e-06,
+      "loss": 0.9479,
+      "step": 1571
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963130210728468e-06,
+      "loss": 0.9696,
+      "step": 1572
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963045799740407e-06,
+      "loss": 1.0324,
+      "step": 1573
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962961292594461e-06,
+      "loss": 0.9941,
+      "step": 1574
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962876689292265e-06,
+      "loss": 1.0755,
+      "step": 1575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96279198983546e-06,
+      "loss": 1.0734,
+      "step": 1576
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962707194225683e-06,
+      "loss": 1.1451,
+      "step": 1577
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962622302464581e-06,
+      "loss": 0.9451,
+      "step": 1578
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962537314553799e-06,
+      "loss": 0.9061,
+      "step": 1579
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96245223049498e-06,
+      "loss": 1.1203,
+      "step": 1580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962367050289777e-06,
+      "loss": 0.9753,
+      "step": 1581
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962281773939837e-06,
+      "loss": 0.9001,
+      "step": 1582
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962196401446815e-06,
+      "loss": 0.8145,
+      "step": 1583
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962110932812361e-06,
+      "loss": 1.0995,
+      "step": 1584
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962025368038134e-06,
+      "loss": 1.039,
+      "step": 1585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96193970712579e-06,
+      "loss": 0.9855,
+      "step": 1586
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961853950076992e-06,
+      "loss": 1.1167,
+      "step": 1587
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961768096893399e-06,
+      "loss": 1.0072,
+      "step": 1588
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961682147576672e-06,
+      "loss": 1.013,
+      "step": 1589
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96159610212848e-06,
+      "loss": 1.2079,
+      "step": 1590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961509960550491e-06,
+      "loss": 0.9402,
+      "step": 1591
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96142372284437e-06,
+      "loss": 0.9944,
+      "step": 1592
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961337389011792e-06,
+      "loss": 1.0711,
+      "step": 1593
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961250959054426e-06,
+      "loss": 1.1335,
+      "step": 1594
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961164432973946e-06,
+      "loss": 1.0292,
+      "step": 1595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961077810772033e-06,
+      "loss": 0.9617,
+      "step": 1596
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960991092450363e-06,
+      "loss": 0.975,
+      "step": 1597
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960904278010615e-06,
+      "loss": 1.0156,
+      "step": 1598
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960817367454471e-06,
+      "loss": 1.0858,
+      "step": 1599
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960730360783617e-06,
+      "loss": 0.8522,
+      "step": 1600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960643257999737e-06,
+      "loss": 1.0748,
+      "step": 1601
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96055605910452e-06,
+      "loss": 0.9028,
+      "step": 1602
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960468764099655e-06,
+      "loss": 1.1582,
+      "step": 1603
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960381372986832e-06,
+      "loss": 0.8836,
+      "step": 1604
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960293885767745e-06,
+      "loss": 0.8599,
+      "step": 1605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96020630244409e-06,
+      "loss": 1.023,
+      "step": 1606
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960118623017562e-06,
+      "loss": 0.8972,
+      "step": 1607
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96003084748986e-06,
+      "loss": 1.047,
+      "step": 1608
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959942975862687e-06,
+      "loss": 1.1254,
+      "step": 1609
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959855008137745e-06,
+      "loss": 1.2491,
+      "step": 1610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959766944316736e-06,
+      "loss": 1.001,
+      "step": 1611
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959678784401366e-06,
+      "loss": 1.1649,
+      "step": 1612
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959590528393347e-06,
+      "loss": 0.9222,
+      "step": 1613
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959502176294384e-06,
+      "loss": 1.0687,
+      "step": 1614
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959413728106195e-06,
+      "loss": 1.0317,
+      "step": 1615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959325183830488e-06,
+      "loss": 1.0201,
+      "step": 1616
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95923654346898e-06,
+      "loss": 1.325,
+      "step": 1617
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959147807023388e-06,
+      "loss": 1.0476,
+      "step": 1618
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959058974495435e-06,
+      "loss": 0.9686,
+      "step": 1619
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958970045886838e-06,
+      "loss": 0.9503,
+      "step": 1620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95888102119932e-06,
+      "loss": 0.7946,
+      "step": 1621
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958791900434608e-06,
+      "loss": 0.8935,
+      "step": 1622
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958702683594427e-06,
+      "loss": 1.0306,
+      "step": 1623
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958613370680507e-06,
+      "loss": 1.0056,
+      "step": 1624
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95852396169458e-06,
+      "loss": 0.9569,
+      "step": 1625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958434456638373e-06,
+      "loss": 1.0781,
+      "step": 1626
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958344855513624e-06,
+      "loss": 0.947,
+      "step": 1627
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958255158322068e-06,
+      "loss": 0.959,
+      "step": 1628
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958165365065444e-06,
+      "loss": 1.1103,
+      "step": 1629
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95807547574549e-06,
+      "loss": 1.1059,
+      "step": 1630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957985490363949e-06,
+      "loss": 0.9742,
+      "step": 1631
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957895408922561e-06,
+      "loss": 1.0427,
+      "step": 1632
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957805231423077e-06,
+      "loss": 1.0356,
+      "step": 1633
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957714957867242e-06,
+      "loss": 1.0757,
+      "step": 1634
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957624588256803e-06,
+      "loss": 1.1523,
+      "step": 1635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95753412259351e-06,
+      "loss": 0.9468,
+      "step": 1636
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95744356087912e-06,
+      "loss": 1.1302,
+      "step": 1637
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957352903115388e-06,
+      "loss": 0.9501,
+      "step": 1638
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957262149304065e-06,
+      "loss": 1.0805,
+      "step": 1639
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957171299446914e-06,
+      "loss": 1.1351,
+      "step": 1640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957080353545691e-06,
+      "loss": 1.1212,
+      "step": 1641
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95698931160216e-06,
+      "loss": 0.9917,
+      "step": 1642
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956898173618089e-06,
+      "loss": 0.9151,
+      "step": 1643
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956806939595238e-06,
+      "loss": 0.8852,
+      "step": 1644
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956715609535378e-06,
+      "loss": 1.0297,
+      "step": 1645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956624183440278e-06,
+      "loss": 0.9761,
+      "step": 1646
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956532661311708e-06,
+      "loss": 0.8516,
+      "step": 1647
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956441043151442e-06,
+      "loss": 0.9536,
+      "step": 1648
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956349328961255e-06,
+      "loss": 1.0827,
+      "step": 1649
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956257518742925e-06,
+      "loss": 1.0962,
+      "step": 1650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956165612498227e-06,
+      "loss": 0.8948,
+      "step": 1651
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956073610228948e-06,
+      "loss": 1.0592,
+      "step": 1652
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955981511936865e-06,
+      "loss": 1.0244,
+      "step": 1653
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955889317623766e-06,
+      "loss": 0.9666,
+      "step": 1654
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955797027291434e-06,
+      "loss": 0.907,
+      "step": 1655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955704640941659e-06,
+      "loss": 0.9396,
+      "step": 1656
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955612158576231e-06,
+      "loss": 1.0121,
+      "step": 1657
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955519580196942e-06,
+      "loss": 0.9241,
+      "step": 1658
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955426905805586e-06,
+      "loss": 0.9913,
+      "step": 1659
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955334135403955e-06,
+      "loss": 0.9495,
+      "step": 1660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955241268993852e-06,
+      "loss": 1.01,
+      "step": 1661
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955148306577072e-06,
+      "loss": 1.0569,
+      "step": 1662
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955055248155418e-06,
+      "loss": 0.971,
+      "step": 1663
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95496209373069e-06,
+      "loss": 1.0356,
+      "step": 1664
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954868843304699e-06,
+      "loss": 1.1179,
+      "step": 1665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954775496879247e-06,
+      "loss": 0.9631,
+      "step": 1666
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954682054456143e-06,
+      "loss": 1.0698,
+      "step": 1667
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9545885160372e-06,
+      "loss": 1.0074,
+      "step": 1668
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954494881624227e-06,
+      "loss": 1.1007,
+      "step": 1669
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954401151219039e-06,
+      "loss": 0.8376,
+      "step": 1670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954307324823454e-06,
+      "loss": 0.9881,
+      "step": 1671
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954213402439288e-06,
+      "loss": 0.9115,
+      "step": 1672
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954119384068359e-06,
+      "loss": 0.8936,
+      "step": 1673
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954025269712494e-06,
+      "loss": 0.8403,
+      "step": 1674
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95393105937351e-06,
+      "loss": 0.9338,
+      "step": 1675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953836753053239e-06,
+      "loss": 0.9915,
+      "step": 1676
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953742350753502e-06,
+      "loss": 1.2261,
+      "step": 1677
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953647852476131e-06,
+      "loss": 0.9699,
+      "step": 1678
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953553258222956e-06,
+      "loss": 0.8992,
+      "step": 1679
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953458567995812e-06,
+      "loss": 0.9724,
+      "step": 1680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95336378179653e-06,
+      "loss": 0.9708,
+      "step": 1681
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953268899626952e-06,
+      "loss": 1.0545,
+      "step": 1682
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953173921488908e-06,
+      "loss": 1.0381,
+      "step": 1683
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953078847384245e-06,
+      "loss": 1.0549,
+      "step": 1684
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952983677314802e-06,
+      "loss": 0.9427,
+      "step": 1685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952888411282426e-06,
+      "loss": 0.9494,
+      "step": 1686
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952793049288957e-06,
+      "loss": 1.1326,
+      "step": 1687
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95269759133625e-06,
+      "loss": 0.9694,
+      "step": 1688
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95260203742615e-06,
+      "loss": 0.9824,
+      "step": 1689
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952506387560506e-06,
+      "loss": 1.1976,
+      "step": 1690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952410641741177e-06,
+      "loss": 1.0866,
+      "step": 1691
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952314799970012e-06,
+      "loss": 0.9898,
+      "step": 1692
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952218862248875e-06,
+      "loss": 1.0791,
+      "step": 1693
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952122828579619e-06,
+      "loss": 1.0374,
+      "step": 1694
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952026698964107e-06,
+      "loss": 0.9522,
+      "step": 1695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951930473404201e-06,
+      "loss": 0.9936,
+      "step": 1696
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951834151901764e-06,
+      "loss": 1.0739,
+      "step": 1697
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951737734458665e-06,
+      "loss": 1.0987,
+      "step": 1698
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951641221076772e-06,
+      "loss": 0.8774,
+      "step": 1699
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95154461175795e-06,
+      "loss": 1.1643,
+      "step": 1700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951447906504077e-06,
+      "loss": 0.9199,
+      "step": 1701
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951351105317023e-06,
+      "loss": 1.0409,
+      "step": 1702
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951254208198664e-06,
+      "loss": 1.0822,
+      "step": 1703
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95115721515088e-06,
+      "loss": 0.8414,
+      "step": 1704
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951060126175545e-06,
+      "loss": 0.9593,
+      "step": 1705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950962941274544e-06,
+      "loss": 1.0734,
+      "step": 1706
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95086566044976e-06,
+      "loss": 1.0127,
+      "step": 1707
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950768283703078e-06,
+      "loss": 1.0825,
+      "step": 1708
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950670811036382e-06,
+      "loss": 0.8473,
+      "step": 1709
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950573242451562e-06,
+      "loss": 1.0761,
+      "step": 1710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95047557795051e-06,
+      "loss": 0.9912,
+      "step": 1711
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950377817535114e-06,
+      "loss": 1.0021,
+      "step": 1712
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950279961207272e-06,
+      "loss": 1.0279,
+      "step": 1713
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950182008968879e-06,
+      "loss": 0.9614,
+      "step": 1714
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950083960821832e-06,
+      "loss": 1.088,
+      "step": 1715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94998581676803e-06,
+      "loss": 1.2322,
+      "step": 1716
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949887576809378e-06,
+      "loss": 1.0638,
+      "step": 1717
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949789240947776e-06,
+      "loss": 1.0943,
+      "step": 1718
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949690809185132e-06,
+      "loss": 1.1159,
+      "step": 1719
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949592281523348e-06,
+      "loss": 0.8642,
+      "step": 1720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949493657964337e-06,
+      "loss": 1.1064,
+      "step": 1721
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94939493851001e-06,
+      "loss": 1.0155,
+      "step": 1722
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949296123162279e-06,
+      "loss": 0.9547,
+      "step": 1723
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949197211923059e-06,
+      "loss": 1.0313,
+      "step": 1724
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949098204794263e-06,
+      "loss": 1.0278,
+      "step": 1725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948999101777815e-06,
+      "loss": 0.9546,
+      "step": 1726
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94889990287563e-06,
+      "loss": 0.8652,
+      "step": 1727
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948800608089632e-06,
+      "loss": 1.1512,
+      "step": 1728
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948701217421746e-06,
+      "loss": 1.017,
+      "step": 1729
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948601730873896e-06,
+      "loss": 0.7816,
+      "step": 1730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94850214844801e-06,
+      "loss": 1.1229,
+      "step": 1731
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948402470146017e-06,
+      "loss": 1.274,
+      "step": 1732
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948302695969849e-06,
+      "loss": 1.0088,
+      "step": 1733
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948202825921438e-06,
+      "loss": 0.9601,
+      "step": 1734
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94810286000272e-06,
+      "loss": 1.0882,
+      "step": 1735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948002798215632e-06,
+      "loss": 0.9775,
+      "step": 1736
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947902640562113e-06,
+      "loss": 1.0466,
+      "step": 1737
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947802387044101e-06,
+      "loss": 1.1709,
+      "step": 1738
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94770203766354e-06,
+      "loss": 1.125,
+      "step": 1739
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947601592422376e-06,
+      "loss": 0.9493,
+      "step": 1740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947501051322552e-06,
+      "loss": 1.0047,
+      "step": 1741
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947400414366018e-06,
+      "loss": 1.1389,
+      "step": 1742
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947299681554723e-06,
+      "loss": 1.0431,
+      "step": 1743
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94719885289062e-06,
+      "loss": 1.0256,
+      "step": 1744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947097928375661e-06,
+      "loss": 1.1776,
+      "step": 1745
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946996908011802e-06,
+      "loss": 0.9398,
+      "step": 1746
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946895791801e-06,
+      "loss": 1.0283,
+      "step": 1747
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946794579745212e-06,
+      "loss": 1.1612,
+      "step": 1748
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946693271846403e-06,
+      "loss": 0.9106,
+      "step": 1749
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946591868106535e-06,
+      "loss": 1.1738,
+      "step": 1750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946490368527569e-06,
+      "loss": 0.9282,
+      "step": 1751
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946388773111475e-06,
+      "loss": 0.9937,
+      "step": 1752
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946287081860221e-06,
+      "loss": 0.8718,
+      "step": 1753
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946185294775777e-06,
+      "loss": 0.9507,
+      "step": 1754
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946083411860116e-06,
+      "loss": 0.8512,
+      "step": 1755
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94598143311521e-06,
+      "loss": 0.9553,
+      "step": 1756
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945879358543034e-06,
+      "loss": 1.0056,
+      "step": 1757
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945777188145568e-06,
+      "loss": 1.0174,
+      "step": 1758
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945674921924793e-06,
+      "loss": 0.8363,
+      "step": 1759
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945572559882686e-06,
+      "loss": 1.1212,
+      "step": 1760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945470102021233e-06,
+      "loss": 1.1106,
+      "step": 1761
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945367548342419e-06,
+      "loss": 0.8903,
+      "step": 1762
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94526489884823e-06,
+      "loss": 1.0716,
+      "step": 1763
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945162153540658e-06,
+      "loss": 1.051,
+      "step": 1764
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945059312421687e-06,
+      "loss": 0.9864,
+      "step": 1765
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944956375493316e-06,
+      "loss": 0.9712,
+      "step": 1766
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944853342757537e-06,
+      "loss": 1.2617,
+      "step": 1767
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944750214216345e-06,
+      "loss": 1.0776,
+      "step": 1768
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944646989871739e-06,
+      "loss": 0.968,
+      "step": 1769
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944543669725721e-06,
+      "loss": 1.0478,
+      "step": 1770
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94444025378029e-06,
+      "loss": 1.0004,
+      "step": 1771
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944336742037451e-06,
+      "loss": 0.9745,
+      "step": 1772
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944233134499207e-06,
+      "loss": 1.0927,
+      "step": 1773
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94412943116757e-06,
+      "loss": 0.988,
+      "step": 1774
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944025632044545e-06,
+      "loss": 0.9562,
+      "step": 1775
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943921737132147e-06,
+      "loss": 1.0899,
+      "step": 1776
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943817746432385e-06,
+      "loss": 1.0418,
+      "step": 1777
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943713659947276e-06,
+      "loss": 0.9153,
+      "step": 1778
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943609477678837e-06,
+      "loss": 1.1923,
+      "step": 1779
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943505199629087e-06,
+      "loss": 1.0028,
+      "step": 1780
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943400825800042e-06,
+      "loss": 0.9898,
+      "step": 1781
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94329635619373e-06,
+      "loss": 1.0401,
+      "step": 1782
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94319179081217e-06,
+      "loss": 1.0571,
+      "step": 1783
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943087129657392e-06,
+      "loss": 0.9706,
+      "step": 1784
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942982372731423e-06,
+      "loss": 0.9401,
+      "step": 1785
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94287752003629e-06,
+      "loss": 1.0845,
+      "step": 1786
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942772571574027e-06,
+      "loss": 1.0454,
+      "step": 1787
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942667527346665e-06,
+      "loss": 0.8204,
+      "step": 1788
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942562387356244e-06,
+      "loss": 0.8916,
+      "step": 1789
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942457151604797e-06,
+      "loss": 1.0157,
+      "step": 1790
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942351820094363e-06,
+      "loss": 0.8143,
+      "step": 1791
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942246392826983e-06,
+      "loss": 0.8089,
+      "step": 1792
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9421408698047e-06,
+      "loss": 1.1392,
+      "step": 1793
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94203525102956e-06,
+      "loss": 1.0844,
+      "step": 1794
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941929536503608e-06,
+      "loss": 0.9995,
+      "step": 1795
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941823726228891e-06,
+      "loss": 0.9459,
+      "step": 1796
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941717820207461e-06,
+      "loss": 0.8017,
+      "step": 1797
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94161181844137e-06,
+      "loss": 0.9912,
+      "step": 1798
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941505720932672e-06,
+      "loss": 0.8448,
+      "step": 1799
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94139952768342e-06,
+      "loss": 1.0711,
+      "step": 1800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941293238695673e-06,
+      "loss": 1.0059,
+      "step": 1801
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94118685397149e-06,
+      "loss": 1.1205,
+      "step": 1802
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941080373512933e-06,
+      "loss": 1.0071,
+      "step": 1803
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940973797322064e-06,
+      "loss": 0.9423,
+      "step": 1804
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940867125400949e-06,
+      "loss": 1.0285,
+      "step": 1805
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940760357751653e-06,
+      "loss": 0.9653,
+      "step": 1806
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940653494376249e-06,
+      "loss": 1.0168,
+      "step": 1807
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940546535276801e-06,
+      "loss": 1.0495,
+      "step": 1808
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940439480455386e-06,
+      "loss": 1.0094,
+      "step": 1809
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940332329914078e-06,
+      "loss": 0.9813,
+      "step": 1810
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94022508365495e-06,
+      "loss": 0.9221,
+      "step": 1811
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940117741680082e-06,
+      "loss": 1.1262,
+      "step": 1812
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940010303991553e-06,
+      "loss": 0.7668,
+      "step": 1813
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939902770591446e-06,
+      "loss": 0.9887,
+      "step": 1814
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939795141481843e-06,
+      "loss": 0.9532,
+      "step": 1815
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93968741666483e-06,
+      "loss": 0.9627,
+      "step": 1816
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939579596142493e-06,
+      "loss": 1.0084,
+      "step": 1817
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93947167991692e-06,
+      "loss": 0.9673,
+      "step": 1818
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939363667990207e-06,
+      "loss": 1.0328,
+      "step": 1819
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939255560364441e-06,
+      "loss": 1.0429,
+      "step": 1820
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93914735704172e-06,
+      "loss": 1.0003,
+      "step": 1821
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93903905802414e-06,
+      "loss": 1.1406,
+      "step": 1822
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938930663313798e-06,
+      "loss": 0.9805,
+      "step": 1823
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938822172912795e-06,
+      "loss": 0.9969,
+      "step": 1824
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938713586823232e-06,
+      "loss": 0.937,
+      "step": 1825
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938604905047214e-06,
+      "loss": 0.9606,
+      "step": 1826
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938496127586845e-06,
+      "loss": 1.05,
+      "step": 1827
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938387254444236e-06,
+      "loss": 1.1391,
+      "step": 1828
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938278285621493e-06,
+      "loss": 1.0425,
+      "step": 1829
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938169221120728e-06,
+      "loss": 1.0139,
+      "step": 1830
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938060060944054e-06,
+      "loss": 1.1354,
+      "step": 1831
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937950805093588e-06,
+      "loss": 1.0199,
+      "step": 1832
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937841453571446e-06,
+      "loss": 1.1359,
+      "step": 1833
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937732006379744e-06,
+      "loss": 1.1944,
+      "step": 1834
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937622463520606e-06,
+      "loss": 1.1351,
+      "step": 1835
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937512824996152e-06,
+      "loss": 0.9446,
+      "step": 1836
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937403090808506e-06,
+      "loss": 0.8839,
+      "step": 1837
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937293260959795e-06,
+      "loss": 1.0012,
+      "step": 1838
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93718333545215e-06,
+      "loss": 0.8473,
+      "step": 1839
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937073314287695e-06,
+      "loss": 1.035,
+      "step": 1840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936963197468566e-06,
+      "loss": 0.9504,
+      "step": 1841
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936852984996894e-06,
+      "loss": 0.9624,
+      "step": 1842
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936742676874815e-06,
+      "loss": 0.9895,
+      "step": 1843
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936632273104467e-06,
+      "loss": 1.1576,
+      "step": 1844
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936521773687988e-06,
+      "loss": 1.0455,
+      "step": 1845
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936411178627521e-06,
+      "loss": 1.0383,
+      "step": 1846
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936300487925206e-06,
+      "loss": 0.9582,
+      "step": 1847
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936189701583189e-06,
+      "loss": 1.0096,
+      "step": 1848
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936078819603616e-06,
+      "loss": 0.8237,
+      "step": 1849
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935967841988636e-06,
+      "loss": 1.0731,
+      "step": 1850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935856768740398e-06,
+      "loss": 0.8651,
+      "step": 1851
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935745599861056e-06,
+      "loss": 1.1252,
+      "step": 1852
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935634335352761e-06,
+      "loss": 1.1014,
+      "step": 1853
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93552297521767e-06,
+      "loss": 0.9912,
+      "step": 1854
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935411519457944e-06,
+      "loss": 1.0538,
+      "step": 1855
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935299968075738e-06,
+      "loss": 1.0469,
+      "step": 1856
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935188321073212e-06,
+      "loss": 0.9932,
+      "step": 1857
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935076578452535e-06,
+      "loss": 1.0474,
+      "step": 1858
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934964740215867e-06,
+      "loss": 1.0735,
+      "step": 1859
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934852806365378e-06,
+      "loss": 0.9935,
+      "step": 1860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934740776903235e-06,
+      "loss": 0.8101,
+      "step": 1861
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934628651831609e-06,
+      "loss": 0.8462,
+      "step": 1862
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93451643115267e-06,
+      "loss": 0.8707,
+      "step": 1863
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934404114868596e-06,
+      "loss": 1.136,
+      "step": 1864
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934291702981562e-06,
+      "loss": 1.0092,
+      "step": 1865
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934179195493745e-06,
+      "loss": 0.9378,
+      "step": 1866
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934066592407326e-06,
+      "loss": 1.0583,
+      "step": 1867
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933953893724485e-06,
+      "loss": 0.9692,
+      "step": 1868
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933841099447408e-06,
+      "loss": 1.085,
+      "step": 1869
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933728209578277e-06,
+      "loss": 1.0125,
+      "step": 1870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933615224119282e-06,
+      "loss": 0.8901,
+      "step": 1871
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93350214307261e-06,
+      "loss": 1.0288,
+      "step": 1872
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933388966440457e-06,
+      "loss": 0.9093,
+      "step": 1873
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933275694225007e-06,
+      "loss": 1.1556,
+      "step": 1874
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933162326428462e-06,
+      "loss": 0.9758,
+      "step": 1875
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933048863053016e-06,
+      "loss": 1.0898,
+      "step": 1876
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932935304100866e-06,
+      "loss": 0.9431,
+      "step": 1877
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932821649574214e-06,
+      "loss": 1.0178,
+      "step": 1878
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932707899475262e-06,
+      "loss": 0.9494,
+      "step": 1879
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932594053806213e-06,
+      "loss": 0.877,
+      "step": 1880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932480112569273e-06,
+      "loss": 1.0486,
+      "step": 1881
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93236607576665e-06,
+      "loss": 1.0084,
+      "step": 1882
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932251943400554e-06,
+      "loss": 0.9064,
+      "step": 1883
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932137715473193e-06,
+      "loss": 0.9449,
+      "step": 1884
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932023391986786e-06,
+      "loss": 1.2107,
+      "step": 1885
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931908972943542e-06,
+      "loss": 1.0948,
+      "step": 1886
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931794458345681e-06,
+      "loss": 1.0622,
+      "step": 1887
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931679848195421e-06,
+      "loss": 1.0443,
+      "step": 1888
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931565142494983e-06,
+      "loss": 1.0911,
+      "step": 1889
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931450341246589e-06,
+      "loss": 0.973,
+      "step": 1890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931335444452462e-06,
+      "loss": 1.0355,
+      "step": 1891
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931220452114832e-06,
+      "loss": 1.0901,
+      "step": 1892
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931105364235922e-06,
+      "loss": 1.0686,
+      "step": 1893
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930990180817964e-06,
+      "loss": 0.9775,
+      "step": 1894
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930874901863191e-06,
+      "loss": 1.0184,
+      "step": 1895
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930759527373834e-06,
+      "loss": 0.9274,
+      "step": 1896
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930644057352131e-06,
+      "loss": 0.7739,
+      "step": 1897
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930528491800317e-06,
+      "loss": 1.1816,
+      "step": 1898
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930412830720633e-06,
+      "loss": 0.9904,
+      "step": 1899
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930297074115317e-06,
+      "loss": 1.0268,
+      "step": 1900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930181221986617e-06,
+      "loss": 1.158,
+      "step": 1901
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930065274336771e-06,
+      "loss": 0.7788,
+      "step": 1902
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92994923116803e-06,
+      "loss": 1.1321,
+      "step": 1903
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92983309248264e-06,
+      "loss": 0.9387,
+      "step": 1904
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929716858282852e-06,
+      "loss": 0.9194,
+      "step": 1905
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92960052857092e-06,
+      "loss": 1.0095,
+      "step": 1906
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929484103349094e-06,
+      "loss": 1.1193,
+      "step": 1907
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92936758261963e-06,
+      "loss": 0.8701,
+      "step": 1908
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929250966384792e-06,
+      "loss": 0.9214,
+      "step": 1909
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929134254646831e-06,
+      "loss": 0.9525,
+      "step": 1910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929017447408013e-06,
+      "loss": 0.9112,
+      "step": 1911
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9289005446706e-06,
+      "loss": 0.8717,
+      "step": 1912
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928783546436857e-06,
+      "loss": 1.0114,
+      "step": 1913
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928666452709053e-06,
+      "loss": 1.1596,
+      "step": 1914
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928549263489452e-06,
+      "loss": 0.9918,
+      "step": 1915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928431978780328e-06,
+      "loss": 0.8752,
+      "step": 1916
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928314598583951e-06,
+      "loss": 0.8848,
+      "step": 1917
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928197122902598e-06,
+      "loss": 0.7904,
+      "step": 1918
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928079551738542e-06,
+      "loss": 1.0358,
+      "step": 1919
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927961885094065e-06,
+      "loss": 1.122,
+      "step": 1920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927844122971442e-06,
+      "loss": 1.0334,
+      "step": 1921
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927726265372958e-06,
+      "loss": 0.9852,
+      "step": 1922
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927608312300897e-06,
+      "loss": 1.0423,
+      "step": 1923
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92749026375754e-06,
+      "loss": 1.0162,
+      "step": 1924
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927372119745177e-06,
+      "loss": 1.0449,
+      "step": 1925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927253880266099e-06,
+      "loss": 0.9485,
+      "step": 1926
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927135545322594e-06,
+      "loss": 1.003,
+      "step": 1927
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927017114916955e-06,
+      "loss": 1.0474,
+      "step": 1928
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926898589051476e-06,
+      "loss": 1.1186,
+      "step": 1929
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926779967728458e-06,
+      "loss": 0.8889,
+      "step": 1930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926661250950191e-06,
+      "loss": 1.1296,
+      "step": 1931
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926542438718982e-06,
+      "loss": 0.9612,
+      "step": 1932
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926423531037132e-06,
+      "loss": 1.0035,
+      "step": 1933
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926304527906943e-06,
+      "loss": 0.9139,
+      "step": 1934
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926185429330718e-06,
+      "loss": 1.131,
+      "step": 1935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926066235310772e-06,
+      "loss": 1.0187,
+      "step": 1936
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92594694584941e-06,
+      "loss": 1.0089,
+      "step": 1937
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925827560948941e-06,
+      "loss": 1.0336,
+      "step": 1938
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92570808061168e-06,
+      "loss": 1.1486,
+      "step": 1939
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925588504839945e-06,
+      "loss": 0.92,
+      "step": 1940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925468833636048e-06,
+      "loss": 0.8985,
+      "step": 1941
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92534906700231e-06,
+      "loss": 0.8987,
+      "step": 1942
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925229204941052e-06,
+      "loss": 1.0773,
+      "step": 1943
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925109247454594e-06,
+      "loss": 0.8829,
+      "step": 1944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924989194545262e-06,
+      "loss": 0.9652,
+      "step": 1945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924869046215382e-06,
+      "loss": 1.2348,
+      "step": 1946
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92474880246728e-06,
+      "loss": 0.9527,
+      "step": 1947
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92462846330329e-06,
+      "loss": 1.0382,
+      "step": 1948
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924508028725738e-06,
+      "loss": 1.0691,
+      "step": 1949
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92438749873696e-06,
+      "loss": 1.0006,
+      "step": 1950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924266873339293e-06,
+      "loss": 0.8871,
+      "step": 1951
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924146152535072e-06,
+      "loss": 0.9199,
+      "step": 1952
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924025336326634e-06,
+      "loss": 1.0995,
+      "step": 1953
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923904424716324e-06,
+      "loss": 0.8797,
+      "step": 1954
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923783417706483e-06,
+      "loss": 0.9248,
+      "step": 1955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923662315299455e-06,
+      "loss": 1.0415,
+      "step": 1956
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923541117497586e-06,
+      "loss": 1.0523,
+      "step": 1957
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923419824303224e-06,
+      "loss": 1.0636,
+      "step": 1958
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923298435718722e-06,
+      "loss": 1.1231,
+      "step": 1959
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923176951746428e-06,
+      "loss": 1.1047,
+      "step": 1960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923055372388696e-06,
+      "loss": 1.1497,
+      "step": 1961
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922933697647885e-06,
+      "loss": 0.9487,
+      "step": 1962
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92281192752635e-06,
+      "loss": 0.961,
+      "step": 1963
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922690062026449e-06,
+      "loss": 1.0305,
+      "step": 1964
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922568101150546e-06,
+      "loss": 1.0581,
+      "step": 1965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922446044901004e-06,
+      "loss": 1.0116,
+      "step": 1966
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922323893280185e-06,
+      "loss": 1.1228,
+      "step": 1967
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922201646290458e-06,
+      "loss": 1.0351,
+      "step": 1968
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922079303934189e-06,
+      "loss": 0.843,
+      "step": 1969
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921956866213751e-06,
+      "loss": 1.0872,
+      "step": 1970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921834333131514e-06,
+      "loss": 0.7971,
+      "step": 1971
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921711704689855e-06,
+      "loss": 0.9424,
+      "step": 1972
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921588980891148e-06,
+      "loss": 0.9701,
+      "step": 1973
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92146616173777e-06,
+      "loss": 1.1543,
+      "step": 1974
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921343247232102e-06,
+      "loss": 1.1139,
+      "step": 1975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921220237376525e-06,
+      "loss": 0.9836,
+      "step": 1976
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921097132173422e-06,
+      "loss": 1.0185,
+      "step": 1977
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920973931625177e-06,
+      "loss": 0.8103,
+      "step": 1978
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.92085063573418e-06,
+      "loss": 0.9184,
+      "step": 1979
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.92072724450282e-06,
+      "loss": 1.0127,
+      "step": 1980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920603757933483e-06,
+      "loss": 0.9553,
+      "step": 1981
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920480176028564e-06,
+      "loss": 1.1009,
+      "step": 1982
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.92035649879046e-06,
+      "loss": 1.0833,
+      "step": 1983
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920232726221564e-06,
+      "loss": 0.9547,
+      "step": 1984
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920108858324273e-06,
+      "loss": 0.8895,
+      "step": 1985
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91998489510099e-06,
+      "loss": 1.0562,
+      "step": 1986
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919860836554118e-06,
+      "loss": 0.9102,
+      "step": 1987
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919736682686056e-06,
+      "loss": 1.0932,
+      "step": 1988
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919612433499213e-06,
+      "loss": 0.9115,
+      "step": 1989
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919488088995994e-06,
+      "loss": 1.009,
+      "step": 1990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91936364917881e-06,
+      "loss": 1.0657,
+      "step": 1991
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919239114050072e-06,
+      "loss": 0.9956,
+      "step": 1992
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91911448361219e-06,
+      "loss": 1.0581,
+      "step": 1993
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918989757867584e-06,
+      "loss": 0.9993,
+      "step": 1994
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918864936818664e-06,
+      "loss": 1.0298,
+      "step": 1995
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918740020467851e-06,
+      "loss": 0.981,
+      "step": 1996
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918615008817568e-06,
+      "loss": 1.2462,
+      "step": 1997
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918489901870235e-06,
+      "loss": 0.9378,
+      "step": 1998
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918364699628275e-06,
+      "loss": 0.9591,
+      "step": 1999
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918239402094115e-06,
+      "loss": 1.1187,
+      "step": 2000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918114009270182e-06,
+      "loss": 1.0722,
+      "step": 2001
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917988521158906e-06,
+      "loss": 0.9972,
+      "step": 2002
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917862937762719e-06,
+      "loss": 0.9926,
+      "step": 2003
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91773725908405e-06,
+      "loss": 0.9736,
+      "step": 2004
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91761148512534e-06,
+      "loss": 0.8091,
+      "step": 2005
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917485615889022e-06,
+      "loss": 0.9086,
+      "step": 2006
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917359651377536e-06,
+      "loss": 0.913,
+      "step": 2007
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917233591593322e-06,
+      "loss": 1.0675,
+      "step": 2008
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917107436538822e-06,
+      "loss": 0.9753,
+      "step": 2009
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916981186216482e-06,
+      "loss": 0.8805,
+      "step": 2010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916854840628745e-06,
+      "loss": 1.0843,
+      "step": 2011
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916728399778061e-06,
+      "loss": 1.0495,
+      "step": 2012
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916601863666882e-06,
+      "loss": 1.2244,
+      "step": 2013
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916475232297655e-06,
+      "loss": 1.0995,
+      "step": 2014
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916348505672837e-06,
+      "loss": 1.1053,
+      "step": 2015
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916221683794881e-06,
+      "loss": 1.1,
+      "step": 2016
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916094766666246e-06,
+      "loss": 0.9436,
+      "step": 2017
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915967754289388e-06,
+      "loss": 1.0103,
+      "step": 2018
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915840646666772e-06,
+      "loss": 1.0006,
+      "step": 2019
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915713443800857e-06,
+      "loss": 0.8445,
+      "step": 2020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915586145694111e-06,
+      "loss": 0.8996,
+      "step": 2021
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915458752348996e-06,
+      "loss": 1.0436,
+      "step": 2022
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915331263767984e-06,
+      "loss": 1.1176,
+      "step": 2023
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915203679953545e-06,
+      "loss": 0.99,
+      "step": 2024
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915076000908148e-06,
+      "loss": 0.9655,
+      "step": 2025
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914948226634268e-06,
+      "loss": 1.1182,
+      "step": 2026
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914820357134383e-06,
+      "loss": 1.2027,
+      "step": 2027
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914692392410967e-06,
+      "loss": 0.9898,
+      "step": 2028
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914564332466501e-06,
+      "loss": 0.8768,
+      "step": 2029
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914436177303465e-06,
+      "loss": 0.9226,
+      "step": 2030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914307926924344e-06,
+      "loss": 1.0098,
+      "step": 2031
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914179581331622e-06,
+      "loss": 0.9933,
+      "step": 2032
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914051140527785e-06,
+      "loss": 0.8242,
+      "step": 2033
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91392260451532e-06,
+      "loss": 0.9711,
+      "step": 2034
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91379397329672e-06,
+      "loss": 0.9083,
+      "step": 2035
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91366524687448e-06,
+      "loss": 0.969,
+      "step": 2036
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913536425251086e-06,
+      "loss": 1.199,
+      "step": 2037
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913407508429041e-06,
+      "loss": 1.0111,
+      "step": 2038
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913278496410838e-06,
+      "loss": 1.0732,
+      "step": 2039
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91314938919898e-06,
+      "loss": 1.0792,
+      "step": 2040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913020186795967e-06,
+      "loss": 1.0505,
+      "step": 2041
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912890889204303e-06,
+      "loss": 0.9811,
+      "step": 2042
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91276149642649e-06,
+      "loss": 1.0025,
+      "step": 2043
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91263200846504e-06,
+      "loss": 1.242,
+      "step": 2044
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912502425322457e-06,
+      "loss": 1.0014,
+      "step": 2045
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912372747001255e-06,
+      "loss": 0.9195,
+      "step": 2046
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912242973503948e-06,
+      "loss": 0.9953,
+      "step": 2047
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912113104833043e-06,
+      "loss": 1.1132,
+      "step": 2048
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911983140991065e-06,
+      "loss": 0.9784,
+      "step": 2049
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911853081980525e-06,
+      "loss": 0.9413,
+      "step": 2050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911722927803946e-06,
+      "loss": 1.04,
+      "step": 2051
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91159267846385e-06,
+      "loss": 1.0853,
+      "step": 2052
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91146233396276e-06,
+      "loss": 1.0984,
+      "step": 2053
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.9113318943032e-06,
+      "loss": 1.0146,
+      "step": 2054
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911201359487702e-06,
+      "loss": 0.9471,
+      "step": 2055
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911070729518788e-06,
+      "loss": 1.0703,
+      "step": 2056
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910940004398994e-06,
+      "loss": 0.9984,
+      "step": 2057
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910809184130853e-06,
+      "loss": 1.0203,
+      "step": 2058
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910678268716896e-06,
+      "loss": 1.034,
+      "step": 2059
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910547258159662e-06,
+      "loss": 0.9583,
+      "step": 2060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910416152461688e-06,
+      "loss": 0.9433,
+      "step": 2061
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910284951625518e-06,
+      "loss": 1.0508,
+      "step": 2062
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910153655653686e-06,
+      "loss": 0.9927,
+      "step": 2063
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910022264548744e-06,
+      "loss": 0.9842,
+      "step": 2064
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909890778313234e-06,
+      "loss": 0.809,
+      "step": 2065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909759196949705e-06,
+      "loss": 1.049,
+      "step": 2066
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909627520460704e-06,
+      "loss": 0.9005,
+      "step": 2067
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909495748848783e-06,
+      "loss": 0.9078,
+      "step": 2068
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909363882116495e-06,
+      "loss": 1.0173,
+      "step": 2069
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909231920266397e-06,
+      "loss": 0.8497,
+      "step": 2070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909099863301044e-06,
+      "loss": 0.9653,
+      "step": 2071
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908967711222994e-06,
+      "loss": 1.0196,
+      "step": 2072
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90883546403481e-06,
+      "loss": 1.0341,
+      "step": 2073
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90870312173905e-06,
+      "loss": 0.9524,
+      "step": 2074
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90857068433828e-06,
+      "loss": 0.9609,
+      "step": 2075
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90843815183507e-06,
+      "loss": 0.9312,
+      "step": 2076
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908305524231981e-06,
+      "loss": 0.9891,
+      "step": 2077
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908172801531587e-06,
+      "loss": 0.9693,
+      "step": 2078
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908039983736458e-06,
+      "loss": 1.0393,
+      "step": 2079
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907907070849168e-06,
+      "loss": 0.9948,
+      "step": 2080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907774062872292e-06,
+      "loss": 1.0184,
+      "step": 2081
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907640959808407e-06,
+      "loss": 0.9721,
+      "step": 2082
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907507761660092e-06,
+      "loss": 0.9107,
+      "step": 2083
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907374468429925e-06,
+      "loss": 1.1775,
+      "step": 2084
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907241080120492e-06,
+      "loss": 1.2383,
+      "step": 2085
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907107596734377e-06,
+      "loss": 0.9867,
+      "step": 2086
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906974018274164e-06,
+      "loss": 0.9794,
+      "step": 2087
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906840344742443e-06,
+      "loss": 0.9461,
+      "step": 2088
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906706576141805e-06,
+      "loss": 0.9278,
+      "step": 2089
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90657271247484e-06,
+      "loss": 1.0748,
+      "step": 2090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906438753744138e-06,
+      "loss": 1.1079,
+      "step": 2091
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906304699952303e-06,
+      "loss": 1.0091,
+      "step": 2092
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906170551101924e-06,
+      "loss": 0.9749,
+      "step": 2093
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906036307195606e-06,
+      "loss": 0.9964,
+      "step": 2094
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905901968235944e-06,
+      "loss": 1.086,
+      "step": 2095
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905767534225547e-06,
+      "loss": 0.9802,
+      "step": 2096
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905633005167017e-06,
+      "loss": 0.942,
+      "step": 2097
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905498381062958e-06,
+      "loss": 1.0328,
+      "step": 2098
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905363661915983e-06,
+      "loss": 0.9766,
+      "step": 2099
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905228847728698e-06,
+      "loss": 1.0762,
+      "step": 2100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905093938503719e-06,
+      "loss": 1.103,
+      "step": 2101
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904958934243655e-06,
+      "loss": 1.0801,
+      "step": 2102
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904823834951125e-06,
+      "loss": 1.0824,
+      "step": 2103
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904688640628745e-06,
+      "loss": 1.0002,
+      "step": 2104
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904553351279139e-06,
+      "loss": 0.8978,
+      "step": 2105
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90441796690492e-06,
+      "loss": 0.7929,
+      "step": 2106
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904282487508717e-06,
+      "loss": 1.1159,
+      "step": 2107
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904146913093153e-06,
+      "loss": 0.9416,
+      "step": 2108
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904011243660856e-06,
+      "loss": 0.9884,
+      "step": 2109
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903875479214453e-06,
+      "loss": 1.0305,
+      "step": 2110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903739619756575e-06,
+      "loss": 1.0091,
+      "step": 2111
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903603665289854e-06,
+      "loss": 1.0947,
+      "step": 2112
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903467615816924e-06,
+      "loss": 1.0449,
+      "step": 2113
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903331471340423e-06,
+      "loss": 0.852,
+      "step": 2114
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903195231862986e-06,
+      "loss": 0.7434,
+      "step": 2115
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903058897387252e-06,
+      "loss": 0.9095,
+      "step": 2116
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902922467915865e-06,
+      "loss": 1.1929,
+      "step": 2117
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90278594345147e-06,
+      "loss": 1.0608,
+      "step": 2118
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902649323996706e-06,
+      "loss": 0.9213,
+      "step": 2119
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902512609554225e-06,
+      "loss": 1.1264,
+      "step": 2120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902375800126673e-06,
+      "loss": 1.2617,
+      "step": 2121
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902238895716703e-06,
+      "loss": 1.1039,
+      "step": 2122
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902101896326967e-06,
+      "loss": 1.0117,
+      "step": 2123
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901964801960118e-06,
+      "loss": 1.1559,
+      "step": 2124
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901827612618813e-06,
+      "loss": 1.1102,
+      "step": 2125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90169032830571e-06,
+      "loss": 1.0849,
+      "step": 2126
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901552949023468e-06,
+      "loss": 0.9425,
+      "step": 2127
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901415474774751e-06,
+      "loss": 1.0577,
+      "step": 2128
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901277905562219e-06,
+      "loss": 0.8314,
+      "step": 2129
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90114024138854e-06,
+      "loss": 0.9345,
+      "step": 2130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901002482256382e-06,
+      "loss": 1.0119,
+      "step": 2131
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900864628168413e-06,
+      "loss": 0.9254,
+      "step": 2132
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900726679127301e-06,
+      "loss": 1.0264,
+      "step": 2133
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900588635135723e-06,
+      "loss": 0.823,
+      "step": 2134
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90045049619635e-06,
+      "loss": 1.1144,
+      "step": 2135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900312262311862e-06,
+      "loss": 1.0105,
+      "step": 2136
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900173933484936e-06,
+      "loss": 0.9182,
+      "step": 2137
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900035509718252e-06,
+      "loss": 1.1084,
+      "step": 2138
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899896991014488e-06,
+      "loss": 0.8774,
+      "step": 2139
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899758377376334e-06,
+      "loss": 1.0384,
+      "step": 2140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899619668806473e-06,
+      "loss": 1.1329,
+      "step": 2141
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899480865307591e-06,
+      "loss": 1.1202,
+      "step": 2142
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899341966882379e-06,
+      "loss": 1.0175,
+      "step": 2143
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899202973533529e-06,
+      "loss": 0.9894,
+      "step": 2144
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89906388526373e-06,
+      "loss": 0.9669,
+      "step": 2145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898924702075683e-06,
+      "loss": 1.2104,
+      "step": 2146
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898785423972079e-06,
+      "loss": 1.0437,
+      "step": 2147
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898646050955618e-06,
+      "loss": 0.9604,
+      "step": 2148
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898506583029003e-06,
+      "loss": 1.0491,
+      "step": 2149
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898367020194932e-06,
+      "loss": 1.0984,
+      "step": 2150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898227362456113e-06,
+      "loss": 0.8719,
+      "step": 2151
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89808760981525e-06,
+      "loss": 0.9985,
+      "step": 2152
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897947762275049e-06,
+      "loss": 0.923,
+      "step": 2153
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897807819838222e-06,
+      "loss": 0.908,
+      "step": 2154
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897667782507482e-06,
+      "loss": 1.1044,
+      "step": 2155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897527650285534e-06,
+      "loss": 0.9254,
+      "step": 2156
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897387423175103e-06,
+      "loss": 1.0114,
+      "step": 2157
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897247101178901e-06,
+      "loss": 0.9903,
+      "step": 2158
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897106684299646e-06,
+      "loss": 1.0211,
+      "step": 2159
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896966172540061e-06,
+      "loss": 0.9783,
+      "step": 2160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896825565902866e-06,
+      "loss": 1.268,
+      "step": 2161
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89668486439079e-06,
+      "loss": 1.0838,
+      "step": 2162
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89654406800655e-06,
+      "loss": 1.0945,
+      "step": 2163
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896403176752883e-06,
+      "loss": 0.8402,
+      "step": 2164
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896262190632515e-06,
+      "loss": 0.8627,
+      "step": 2165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896121109648177e-06,
+      "loss": 1.0719,
+      "step": 2166
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895979933802603e-06,
+      "loss": 0.9629,
+      "step": 2167
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89583866309853e-06,
+      "loss": 0.889,
+      "step": 2168
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89569729753869e-06,
+      "loss": 0.9531,
+      "step": 2169
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895555837125827e-06,
+      "loss": 0.9591,
+      "step": 2170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895414281862682e-06,
+      "loss": 0.9419,
+      "step": 2171
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895272631751994e-06,
+      "loss": 1.0504,
+      "step": 2172
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895130886796509e-06,
+      "loss": 0.959,
+      "step": 2173
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894989046998976e-06,
+      "loss": 0.9853,
+      "step": 2174
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894847112362137e-06,
+      "loss": 0.9407,
+      "step": 2175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894705082888748e-06,
+      "loss": 1.139,
+      "step": 2176
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894562958581557e-06,
+      "loss": 1.0266,
+      "step": 2177
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89442073944332e-06,
+      "loss": 0.8981,
+      "step": 2178
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89427842547679e-06,
+      "loss": 0.8458,
+      "step": 2179
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894136016684727e-06,
+      "loss": 0.911,
+      "step": 2180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893993513069889e-06,
+      "loss": 0.8557,
+      "step": 2181
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893850914635035e-06,
+      "loss": 1.0884,
+      "step": 2182
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893708221382932e-06,
+      "loss": 1.035,
+      "step": 2183
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89356543331634e-06,
+      "loss": 1.1362,
+      "step": 2184
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89342255043803e-06,
+      "loss": 1.1536,
+      "step": 2185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893279572750767e-06,
+      "loss": 1.1306,
+      "step": 2186
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89313650025732e-06,
+      "loss": 1.0583,
+      "step": 2187
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892993332960468e-06,
+      "loss": 1.1567,
+      "step": 2188
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892850070862976e-06,
+      "loss": 0.9353,
+      "step": 2189
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892706713967626e-06,
+      "loss": 1.1629,
+      "step": 2190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892563262277192e-06,
+      "loss": 0.8733,
+      "step": 2191
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892419715794456e-06,
+      "loss": 1.023,
+      "step": 2192
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892276074522197e-06,
+      "loss": 0.8365,
+      "step": 2193
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892132338463199e-06,
+      "loss": 1.1561,
+      "step": 2194
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891988507620247e-06,
+      "loss": 0.9317,
+      "step": 2195
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891844581996126e-06,
+      "loss": 0.9355,
+      "step": 2196
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891700561593627e-06,
+      "loss": 0.9578,
+      "step": 2197
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891556446415538e-06,
+      "loss": 0.9141,
+      "step": 2198
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891412236464654e-06,
+      "loss": 1.073,
+      "step": 2199
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891267931743768e-06,
+      "loss": 1.0639,
+      "step": 2200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891123532255674e-06,
+      "loss": 0.9011,
+      "step": 2201
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890979038003172e-06,
+      "loss": 0.9542,
+      "step": 2202
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89083444898906e-06,
+      "loss": 1.0745,
+      "step": 2203
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890689765216139e-06,
+      "loss": 0.9976,
+      "step": 2204
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890544986687215e-06,
+      "loss": 0.9341,
+      "step": 2205
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89040011340509e-06,
+      "loss": 1.0815,
+      "step": 2206
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890255145372571e-06,
+      "loss": 0.9199,
+      "step": 2207
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89011008259247e-06,
+      "loss": 0.9921,
+      "step": 2208
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.889964925067594e-06,
+      "loss": 0.9755,
+      "step": 2209
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.889819672800759e-06,
+      "loss": 1.0589,
+      "step": 2210
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889674325794776e-06,
+      "loss": 1.1118,
+      "step": 2211
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88952888405246e-06,
+      "loss": 1.1507,
+      "step": 2212
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889383347576633e-06,
+      "loss": 1.047,
+      "step": 2213
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.889237716370112e-06,
+      "loss": 1.0349,
+      "step": 2214
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88909199043572e-06,
+      "loss": 0.9429,
+      "step": 2215
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88894616977628e-06,
+      "loss": 0.9302,
+      "step": 2216
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888800254394617e-06,
+      "loss": 0.9388,
+      "step": 2217
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888654244293557e-06,
+      "loss": 0.9831,
+      "step": 2218
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88850813947593e-06,
+      "loss": 0.8525,
+      "step": 2219
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888361939944567e-06,
+      "loss": 1.0816,
+      "step": 2220
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8882156457023e-06,
+      "loss": 0.9074,
+      "step": 2221
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888069256751964e-06,
+      "loss": 1.0899,
+      "step": 2222
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887922773096396e-06,
+      "loss": 1.0823,
+      "step": 2223
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887776194738433e-06,
+      "loss": 0.9967,
+      "step": 2224
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887629521680912e-06,
+      "loss": 1.055,
+      "step": 2225
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88748275392668e-06,
+      "loss": 0.8592,
+      "step": 2226
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887335891478577e-06,
+      "loss": 1.0412,
+      "step": 2227
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887188934339452e-06,
+      "loss": 0.948,
+      "step": 2228
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88704188251215e-06,
+      "loss": 1.0673,
+      "step": 2229
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886894735999519e-06,
+      "loss": 0.8399,
+      "step": 2230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88674749480441e-06,
+      "loss": 1.0843,
+      "step": 2231
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886600158929677e-06,
+      "loss": 1.0078,
+      "step": 2232
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886452728378175e-06,
+      "loss": 1.1508,
+      "step": 2233
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88630520315276e-06,
+      "loss": 0.7341,
+      "step": 2234
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88615758325629e-06,
+      "loss": 0.9835,
+      "step": 2235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886009868691627e-06,
+      "loss": 0.931,
+      "step": 2236
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885862059461629e-06,
+      "loss": 0.9623,
+      "step": 2237
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885714155569162e-06,
+      "loss": 1.0402,
+      "step": 2238
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88556615701709e-06,
+      "loss": 1.1421,
+      "step": 2239
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885418063808284e-06,
+      "loss": 1.0191,
+      "step": 2240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88526987594561e-06,
+      "loss": 1.0566,
+      "step": 2241
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885121593431942e-06,
+      "loss": 1.0259,
+      "step": 2242
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88497321627015e-06,
+      "loss": 0.8944,
+      "step": 2243
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88482474446311e-06,
+      "loss": 1.0278,
+      "step": 2244
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884676178013701e-06,
+      "loss": 0.9411,
+      "step": 2245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884527516924796e-06,
+      "loss": 1.0819,
+      "step": 2246
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88437876119928e-06,
+      "loss": 0.9947,
+      "step": 2247
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884229910840033e-06,
+      "loss": 0.9618,
+      "step": 2248
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88408096584994e-06,
+      "loss": 0.9549,
+      "step": 2249
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883931926231885e-06,
+      "loss": 1.0526,
+      "step": 2250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883782791988758e-06,
+      "loss": 0.8928,
+      "step": 2251
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883633563123446e-06,
+      "loss": 0.9993,
+      "step": 2252
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883484239638842e-06,
+      "loss": 0.9999,
+      "step": 2253
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883334821537839e-06,
+      "loss": 1.0212,
+      "step": 2254
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883185308823331e-06,
+      "loss": 0.9447,
+      "step": 2255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883035701498215e-06,
+      "loss": 1.1187,
+      "step": 2256
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88288599956539e-06,
+      "loss": 1.0858,
+      "step": 2257
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882736203027758e-06,
+      "loss": 0.9328,
+      "step": 2258
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882586311888218e-06,
+      "loss": 1.0635,
+      "step": 2259
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882436326149678e-06,
+      "loss": 1.1055,
+      "step": 2260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88228624581504e-06,
+      "loss": 1.0063,
+      "step": 2261
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882136070887213e-06,
+      "loss": 0.9695,
+      "step": 2262
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881985801369107e-06,
+      "loss": 1.0586,
+      "step": 2263
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881835437263634e-06,
+      "loss": 0.9262,
+      "step": 2264
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881684978573707e-06,
+      "loss": 1.0221,
+      "step": 2265
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88153442530224e-06,
+      "loss": 1.0481,
+      "step": 2266
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881383777452152e-06,
+      "loss": 0.8605,
+      "step": 2267
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88123303502636e-06,
+      "loss": 0.9681,
+      "step": 2268
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881082198027787e-06,
+      "loss": 1.0531,
+      "step": 2269
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880931266459352e-06,
+      "loss": 0.9887,
+      "step": 2270
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88078024032398e-06,
+      "loss": 0.9678,
+      "step": 2271
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880629119624599e-06,
+      "loss": 0.9335,
+      "step": 2272
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880477904364137e-06,
+      "loss": 1.1037,
+      "step": 2273
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880326594545521e-06,
+      "loss": 1.1123,
+      "step": 2274
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880175190171684e-06,
+      "loss": 0.9661,
+      "step": 2275
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880023691245562e-06,
+      "loss": 0.9695,
+      "step": 2276
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879872097770086e-06,
+      "loss": 1.0925,
+      "step": 2277
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879720409748195e-06,
+      "loss": 1.0267,
+      "step": 2278
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879568627182829e-06,
+      "loss": 0.7982,
+      "step": 2279
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879416750076927e-06,
+      "loss": 1.0364,
+      "step": 2280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879264778433433e-06,
+      "loss": 1.1058,
+      "step": 2281
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87911271225529e-06,
+      "loss": 1.0011,
+      "step": 2282
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878960551545445e-06,
+      "loss": 1.1223,
+      "step": 2283
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878808296306846e-06,
+      "loss": 1.0621,
+      "step": 2284
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878655946542443e-06,
+      "loss": 1.014,
+      "step": 2285
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878503502255188e-06,
+      "loss": 1.0156,
+      "step": 2286
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878350963448034e-06,
+      "loss": 1.0028,
+      "step": 2287
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878198330123936e-06,
+      "loss": 0.837,
+      "step": 2288
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878045602285854e-06,
+      "loss": 1.0169,
+      "step": 2289
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877892779936744e-06,
+      "loss": 0.9939,
+      "step": 2290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877739863079567e-06,
+      "loss": 1.0994,
+      "step": 2291
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877586851717287e-06,
+      "loss": 0.8951,
+      "step": 2292
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87743374585287e-06,
+      "loss": 1.0425,
+      "step": 2293
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877280545489279e-06,
+      "loss": 0.9389,
+      "step": 2294
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877127250629483e-06,
+      "loss": 0.8695,
+      "step": 2295
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876973861276453e-06,
+      "loss": 0.9854,
+      "step": 2296
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876820377433161e-06,
+      "loss": 0.9777,
+      "step": 2297
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876666799102582e-06,
+      "loss": 0.8194,
+      "step": 2298
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876513126287688e-06,
+      "loss": 0.9229,
+      "step": 2299
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876359358991459e-06,
+      "loss": 1.0366,
+      "step": 2300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876205497216874e-06,
+      "loss": 1.0377,
+      "step": 2301
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876051540966914e-06,
+      "loss": 0.8644,
+      "step": 2302
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875897490244558e-06,
+      "loss": 1.0429,
+      "step": 2303
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875743345052798e-06,
+      "loss": 1.0786,
+      "step": 2304
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875589105394616e-06,
+      "loss": 0.983,
+      "step": 2305
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875434771273e-06,
+      "loss": 0.7868,
+      "step": 2306
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875280342690942e-06,
+      "loss": 1.0055,
+      "step": 2307
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875125819651433e-06,
+      "loss": 0.9067,
+      "step": 2308
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87497120215747e-06,
+      "loss": 1.1402,
+      "step": 2309
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87481649021204e-06,
+      "loss": 1.1688,
+      "step": 2310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87466168381815e-06,
+      "loss": 0.9285,
+      "step": 2311
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874506782978794e-06,
+      "loss": 0.9674,
+      "step": 2312
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874351787696976e-06,
+      "loss": 1.0219,
+      "step": 2313
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874196697975698e-06,
+      "loss": 1.1262,
+      "step": 2314
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874041513817964e-06,
+      "loss": 0.9123,
+      "step": 2315
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873886235226781e-06,
+      "loss": 1.0651,
+      "step": 2316
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87373086220516e-06,
+      "loss": 1.0908,
+      "step": 2317
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873575394756107e-06,
+      "loss": 1.0169,
+      "step": 2318
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873419832882636e-06,
+      "loss": 0.888,
+      "step": 2319
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873264176587763e-06,
+      "loss": 1.0429,
+      "step": 2320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8731084258745e-06,
+      "loss": 1.0813,
+      "step": 2321
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872952580745867e-06,
+      "loss": 0.9122,
+      "step": 2322
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872796641204883e-06,
+      "loss": 1.0091,
+      "step": 2323
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872640607254571e-06,
+      "loss": 0.9647,
+      "step": 2324
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87248447889795e-06,
+      "loss": 1.1485,
+      "step": 2325
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87232825613805e-06,
+      "loss": 1.0194,
+      "step": 2326
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872171938977895e-06,
+      "loss": 1.028,
+      "step": 2327
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872015527420512e-06,
+      "loss": 1.0272,
+      "step": 2328
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871859021468935e-06,
+      "loss": 1.0701,
+      "step": 2329
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871702421126194e-06,
+      "loss": 1.0129,
+      "step": 2330
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871545726395323e-06,
+      "loss": 0.9297,
+      "step": 2331
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87138893727936e-06,
+      "loss": 1.0673,
+      "step": 2332
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871232053781341e-06,
+      "loss": 1.0246,
+      "step": 2333
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871075075904306e-06,
+      "loss": 1.0358,
+      "step": 2334
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870918003651295e-06,
+      "loss": 1.0048,
+      "step": 2335
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870760837025354e-06,
+      "loss": 1.1324,
+      "step": 2336
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870603576029526e-06,
+      "loss": 1.126,
+      "step": 2337
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870446220666859e-06,
+      "loss": 0.949,
+      "step": 2338
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870288770940401e-06,
+      "loss": 1.0455,
+      "step": 2339
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870131226853203e-06,
+      "loss": 1.1193,
+      "step": 2340
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869973588408316e-06,
+      "loss": 0.8292,
+      "step": 2341
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869815855608798e-06,
+      "loss": 0.9426,
+      "step": 2342
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8696580284577e-06,
+      "loss": 1.1387,
+      "step": 2343
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869500106958083e-06,
+      "loss": 0.8933,
+      "step": 2344
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869342091113007e-06,
+      "loss": 0.9518,
+      "step": 2345
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869183980925531e-06,
+      "loss": 1.0737,
+      "step": 2346
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86902577639872e-06,
+      "loss": 0.9776,
+      "step": 2347
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86886747753564e-06,
+      "loss": 1.0277,
+      "step": 2348
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868709084339357e-06,
+      "loss": 1.0519,
+      "step": 2349
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86855059681294e-06,
+      "loss": 1.123,
+      "step": 2350
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86839201495946e-06,
+      "loss": 0.9182,
+      "step": 2351
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868233338781988e-06,
+      "loss": 1.022,
+      "step": 2352
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868074568283598e-06,
+      "loss": 1.0417,
+      "step": 2353
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86791570346737e-06,
+      "loss": 1.0028,
+      "step": 2354
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86775674433638e-06,
+      "loss": 0.8906,
+      "step": 2355
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867597690893706e-06,
+      "loss": 1.0531,
+      "step": 2356
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86743854314243e-06,
+      "loss": 1.1253,
+      "step": 2357
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867279301085637e-06,
+      "loss": 1.117,
+      "step": 2358
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867119964726413e-06,
+      "loss": 0.9277,
+      "step": 2359
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866960534067842e-06,
+      "loss": 0.8395,
+      "step": 2360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866801009113015e-06,
+      "loss": 0.959,
+      "step": 2361
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866641389865022e-06,
+      "loss": 0.9744,
+      "step": 2362
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866481676326958e-06,
+      "loss": 0.9542,
+      "step": 2363
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866321868501914e-06,
+      "loss": 1.0008,
+      "step": 2364
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866161966392987e-06,
+      "loss": 1.0931,
+      "step": 2365
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866001970003276e-06,
+      "loss": 0.9089,
+      "step": 2366
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865841879335882e-06,
+      "loss": 1.0686,
+      "step": 2367
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865681694393904e-06,
+      "loss": 1.0609,
+      "step": 2368
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865521415180448e-06,
+      "loss": 0.9623,
+      "step": 2369
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865361041698618e-06,
+      "loss": 0.9867,
+      "step": 2370
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86520057395152e-06,
+      "loss": 1.1053,
+      "step": 2371
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865040011942266e-06,
+      "loss": 0.9333,
+      "step": 2372
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864879355673965e-06,
+      "loss": 1.0284,
+      "step": 2373
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86471860514973e-06,
+      "loss": 1.0119,
+      "step": 2374
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864557760372675e-06,
+      "loss": 1.1761,
+      "step": 2375
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864396821345917e-06,
+      "loss": 0.9539,
+      "step": 2376
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864235788072573e-06,
+      "loss": 1.0354,
+      "step": 2377
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864074660555767e-06,
+      "loss": 1.0567,
+      "step": 2378
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863913438798616e-06,
+      "loss": 1.0233,
+      "step": 2379
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863752122804245e-06,
+      "loss": 0.9951,
+      "step": 2380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863590712575782e-06,
+      "loss": 0.9351,
+      "step": 2381
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86342920811635e-06,
+      "loss": 1.0038,
+      "step": 2382
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86326760942908e-06,
+      "loss": 0.9891,
+      "step": 2383
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863105916517104e-06,
+      "loss": 1.0899,
+      "step": 2384
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862944129383555e-06,
+      "loss": 1.0169,
+      "step": 2385
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862782248031566e-06,
+      "loss": 0.9335,
+      "step": 2386
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862620272464274e-06,
+      "loss": 1.083,
+      "step": 2387
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862458202684817e-06,
+      "loss": 0.8237,
+      "step": 2388
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862296038696335e-06,
+      "loss": 0.8787,
+      "step": 2389
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86213378050197e-06,
+      "loss": 1.0026,
+      "step": 2390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861971428104867e-06,
+      "loss": 1.0947,
+      "step": 2391
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86180898150817e-06,
+      "loss": 1.004,
+      "step": 2392
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861646440715027e-06,
+      "loss": 0.9716,
+      "step": 2393
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861483805728587e-06,
+      "loss": 0.9687,
+      "step": 2394
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861321076552e-06,
+      "loss": 1.0236,
+      "step": 2395
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86115825318842e-06,
+      "loss": 0.9519,
+      "step": 2396
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860995335641004e-06,
+      "loss": 0.8527,
+      "step": 2397
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860832323912903e-06,
+      "loss": 1.1031,
+      "step": 2398
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860669218007279e-06,
+      "loss": 1.0429,
+      "step": 2399
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860506017927294e-06,
+      "loss": 0.9568,
+      "step": 2400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860342723676105e-06,
+      "loss": 1.1127,
+      "step": 2401
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860179335256879e-06,
+      "loss": 0.9029,
+      "step": 2402
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86001585267278e-06,
+      "loss": 0.9858,
+      "step": 2403
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859852275926978e-06,
+      "loss": 1.1952,
+      "step": 2404
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85968860502264e-06,
+      "loss": 1.0464,
+      "step": 2405
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859524839962934e-06,
+      "loss": 1.1583,
+      "step": 2406
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85936098075104e-06,
+      "loss": 0.8265,
+      "step": 2407
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85919702739013e-06,
+      "loss": 0.8864,
+      "step": 2408
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859032979883378e-06,
+      "loss": 0.7196,
+      "step": 2409
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858868838233966e-06,
+      "loss": 0.8508,
+      "step": 2410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858704602445072e-06,
+      "loss": 0.9084,
+      "step": 2411
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858540272519878e-06,
+      "loss": 1.0501,
+      "step": 2412
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858375848461567e-06,
+      "loss": 1.0012,
+      "step": 2413
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858211330273328e-06,
+      "loss": 1.0351,
+      "step": 2414
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858046717958345e-06,
+      "loss": 0.9547,
+      "step": 2415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85788201151981e-06,
+      "loss": 1.1402,
+      "step": 2416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857717210960913e-06,
+      "loss": 1.0166,
+      "step": 2417
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857552316284846e-06,
+      "loss": 1.2591,
+      "step": 2418
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857387327494805e-06,
+      "loss": 0.8552,
+      "step": 2419
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857222244593987e-06,
+      "loss": 0.9867,
+      "step": 2420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857057067585589e-06,
+      "loss": 0.9116,
+      "step": 2421
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856891796472815e-06,
+      "loss": 1.063,
+      "step": 2422
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85672643125886e-06,
+      "loss": 1.0062,
+      "step": 2423
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856560971946934e-06,
+      "loss": 0.8649,
+      "step": 2424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85639541854024e-06,
+      "loss": 1.008,
+      "step": 2425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856229771041988e-06,
+      "loss": 1.0221,
+      "step": 2426
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856064029455383e-06,
+      "loss": 1.0951,
+      "step": 2427
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855898193783641e-06,
+      "loss": 0.9796,
+      "step": 2428
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855732264029974e-06,
+      "loss": 1.0766,
+      "step": 2429
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855566240197593e-06,
+      "loss": 0.8844,
+      "step": 2430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85540012228972e-06,
+      "loss": 1.0379,
+      "step": 2431
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85523391030957e-06,
+      "loss": 0.9392,
+      "step": 2432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855067604260365e-06,
+      "loss": 0.9964,
+      "step": 2433
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854901204145325e-06,
+      "loss": 1.0076,
+      "step": 2434
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854734709967677e-06,
+      "loss": 1.0034,
+      "step": 2435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854568121730645e-06,
+      "loss": 0.8645,
+      "step": 2436
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854401439437455e-06,
+      "loss": 1.0872,
+      "step": 2437
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85423466309134e-06,
+      "loss": 0.923,
+      "step": 2438
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85406779269553e-06,
+      "loss": 1.075,
+      "step": 2439
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.853900828253257e-06,
+      "loss": 0.9775,
+      "step": 2440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.853733769767756e-06,
+      "loss": 1.021,
+      "step": 2441
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.853566617242265e-06,
+      "loss": 1.1366,
+      "step": 2442
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85339937068002e-06,
+      "loss": 1.13,
+      "step": 2443
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.853232030084266e-06,
+      "loss": 0.9892,
+      "step": 2444
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85306459545824e-06,
+      "loss": 1.0584,
+      "step": 2445
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852897066805189e-06,
+      "loss": 0.9548,
+      "step": 2446
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852729444128359e-06,
+      "loss": 0.9421,
+      "step": 2447
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852561727430997e-06,
+      "loss": 0.9665,
+      "step": 2448
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85239391671635e-06,
+      "loss": 1.1438,
+      "step": 2449
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852226011987674e-06,
+      "loss": 1.0215,
+      "step": 2450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852058013248217e-06,
+      "loss": 1.1307,
+      "step": 2451
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851889920501239e-06,
+      "loss": 1.0202,
+      "step": 2452
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851721733749993e-06,
+      "loss": 1.006,
+      "step": 2453
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85155345299774e-06,
+      "loss": 1.0501,
+      "step": 2454
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851385078247739e-06,
+      "loss": 1.0072,
+      "step": 2455
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851216609503252e-06,
+      "loss": 1.0947,
+      "step": 2456
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851048046767544e-06,
+      "loss": 0.9572,
+      "step": 2457
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850879390043881e-06,
+      "loss": 1.1178,
+      "step": 2458
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85071063933553e-06,
+      "loss": 0.9146,
+      "step": 2459
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85054179464576e-06,
+      "loss": 0.8791,
+      "step": 2460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850372855977844e-06,
+      "loss": 1.0774,
+      "step": 2461
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850203823335053e-06,
+      "loss": 1.0257,
+      "step": 2462
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850034696720664e-06,
+      "loss": 1.0108,
+      "step": 2463
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849865476137951e-06,
+      "loss": 1.0567,
+      "step": 2464
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849696161590199e-06,
+      "loss": 0.9913,
+      "step": 2465
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84952675308068e-06,
+      "loss": 1.0017,
+      "step": 2466
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849357250612682e-06,
+      "loss": 0.9829,
+      "step": 2467
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849187654189486e-06,
+      "loss": 1.0307,
+      "step": 2468
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84901796381438e-06,
+      "loss": 1.1042,
+      "step": 2469
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848848179490651e-06,
+      "loss": 0.9341,
+      "step": 2470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848678301221588e-06,
+      "loss": 0.8452,
+      "step": 2471
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848508329010483e-06,
+      "loss": 0.9413,
+      "step": 2472
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84833826286063e-06,
+      "loss": 1.084,
+      "step": 2473
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848168102775321e-06,
+      "loss": 0.8902,
+      "step": 2474
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847997848757855e-06,
+      "loss": 1.1265,
+      "step": 2475
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847827500811531e-06,
+      "loss": 1.0333,
+      "step": 2476
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84765705893965e-06,
+      "loss": 0.9677,
+      "step": 2477
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847486523145512e-06,
+      "loss": 1.0092,
+      "step": 2478
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847315893432422e-06,
+      "loss": 0.8347,
+      "step": 2479
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847145169803686e-06,
+      "loss": 1.0273,
+      "step": 2480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846974352262613e-06,
+      "loss": 0.9619,
+      "step": 2481
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84680344081251e-06,
+      "loss": 0.9159,
+      "step": 2482
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846632435456691e-06,
+      "loss": 1.2027,
+      "step": 2483
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846461336198467e-06,
+      "loss": 1.1754,
+      "step": 2484
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846290143041154e-06,
+      "loss": 1.1608,
+      "step": 2485
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846118855988069e-06,
+      "loss": 1.0773,
+      "step": 2486
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845947475042532e-06,
+      "loss": 0.9359,
+      "step": 2487
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84577600020786e-06,
+      "loss": 0.9958,
+      "step": 2488
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845604431487376e-06,
+      "loss": 1.1376,
+      "step": 2489
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845432768884408e-06,
+      "loss": 1.0853,
+      "step": 2490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845261012402279e-06,
+      "loss": 1.062,
+      "step": 2491
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845089162044314e-06,
+      "loss": 0.857,
+      "step": 2492
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84491721781385e-06,
+      "loss": 0.9233,
+      "step": 2493
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84474517971421e-06,
+      "loss": 0.9582,
+      "step": 2494
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844573047748733e-06,
+      "loss": 1.0171,
+      "step": 2495
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844400821920752e-06,
+      "loss": 0.9501,
+      "step": 2496
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844228502233603e-06,
+      "loss": 1.0698,
+      "step": 2497
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844056088690629e-06,
+      "loss": 1.004,
+      "step": 2498
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843883581295163e-06,
+      "loss": 0.9322,
+      "step": 2499
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843710980050553e-06,
+      "loss": 1.1444,
+      "step": 2500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843538284960141e-06,
+      "loss": 1.0529,
+      "step": 2501
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843365496027275e-06,
+      "loss": 1.1057,
+      "step": 2502
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843192613255299e-06,
+      "loss": 0.9423,
+      "step": 2503
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843019636647565e-06,
+      "loss": 0.9524,
+      "step": 2504
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842846566207426e-06,
+      "loss": 0.9302,
+      "step": 2505
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842673401938232e-06,
+      "loss": 0.8672,
+      "step": 2506
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84250014384334e-06,
+      "loss": 0.9929,
+      "step": 2507
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842326791926105e-06,
+      "loss": 0.9568,
+      "step": 2508
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842153346189889e-06,
+      "loss": 0.862,
+      "step": 2509
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841979806638049e-06,
+      "loss": 0.8844,
+      "step": 2510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841806173273948e-06,
+      "loss": 0.9228,
+      "step": 2511
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84163244610095e-06,
+      "loss": 0.8125,
+      "step": 2512
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841458625122425e-06,
+      "loss": 1.1006,
+      "step": 2513
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841284710341735e-06,
+      "loss": 1.1083,
+      "step": 2514
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84111070176225e-06,
+      "loss": 0.8208,
+      "step": 2515
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840936599387347e-06,
+      "loss": 0.9556,
+      "step": 2516
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840762403220394e-06,
+      "loss": 1.0405,
+      "step": 2517
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840588113264767e-06,
+      "loss": 1.1498,
+      "step": 2518
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840413729523843e-06,
+      "loss": 1.0167,
+      "step": 2519
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840239252001002e-06,
+      "loss": 0.9126,
+      "step": 2520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840064680699624e-06,
+      "loss": 0.9792,
+      "step": 2521
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839890015623089e-06,
+      "loss": 0.9378,
+      "step": 2522
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839715256774783e-06,
+      "loss": 1.0931,
+      "step": 2523
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839540404158092e-06,
+      "loss": 1.053,
+      "step": 2524
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839365457776404e-06,
+      "loss": 1.0921,
+      "step": 2525
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839190417633108e-06,
+      "loss": 0.9497,
+      "step": 2526
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839015283731597e-06,
+      "loss": 0.8795,
+      "step": 2527
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838840056075259e-06,
+      "loss": 1.0796,
+      "step": 2528
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838664734667496e-06,
+      "loss": 1.1463,
+      "step": 2529
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8384893195117e-06,
+      "loss": 1.0829,
+      "step": 2530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838313810611268e-06,
+      "loss": 1.1206,
+      "step": 2531
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838138207969608e-06,
+      "loss": 0.9538,
+      "step": 2532
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837962511590115e-06,
+      "loss": 0.8922,
+      "step": 2533
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837786721476197e-06,
+      "loss": 0.7943,
+      "step": 2534
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837610837631257e-06,
+      "loss": 0.9833,
+      "step": 2535
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837434860058706e-06,
+      "loss": 1.0667,
+      "step": 2536
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837258788761953e-06,
+      "loss": 0.984,
+      "step": 2537
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837082623744407e-06,
+      "loss": 1.0632,
+      "step": 2538
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83690636500948e-06,
+      "loss": 0.9506,
+      "step": 2539
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836730012560593e-06,
+      "loss": 0.9473,
+      "step": 2540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836553566401157e-06,
+      "loss": 0.9518,
+      "step": 2541
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836377026534595e-06,
+      "loss": 0.9605,
+      "step": 2542
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836200392964323e-06,
+      "loss": 1.0283,
+      "step": 2543
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836023665693766e-06,
+      "loss": 1.0214,
+      "step": 2544
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835846844726347e-06,
+      "loss": 1.0046,
+      "step": 2545
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835669930065493e-06,
+      "loss": 1.0239,
+      "step": 2546
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835492921714632e-06,
+      "loss": 1.0241,
+      "step": 2547
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83531581967719e-06,
+      "loss": 0.9525,
+      "step": 2548
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835138623956603e-06,
+      "loss": 1.2701,
+      "step": 2549
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834961334556301e-06,
+      "loss": 1.2189,
+      "step": 2550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83478395147972e-06,
+      "loss": 0.9895,
+      "step": 2551
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834606474730297e-06,
+      "loss": 0.9066,
+      "step": 2552
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83442890431147e-06,
+      "loss": 0.8243,
+      "step": 2553
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83425124022668e-06,
+      "loss": 1.0442,
+      "step": 2554
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834073482479368e-06,
+      "loss": 0.9833,
+      "step": 2555
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833895631072979e-06,
+      "loss": 1.066,
+      "step": 2556
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833717686010958e-06,
+      "loss": 1.1349,
+      "step": 2557
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833539647296754e-06,
+      "loss": 0.7936,
+      "step": 2558
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833361514933817e-06,
+      "loss": 0.9009,
+      "step": 2559
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833183288925596e-06,
+      "loss": 1.1026,
+      "step": 2560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833004969275544e-06,
+      "loss": 0.9258,
+      "step": 2561
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832826555987119e-06,
+      "loss": 1.1106,
+      "step": 2562
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832648049063776e-06,
+      "loss": 1.0426,
+      "step": 2563
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832469448508972e-06,
+      "loss": 0.9791,
+      "step": 2564
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832290754326169e-06,
+      "loss": 0.9308,
+      "step": 2565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832111966518827e-06,
+      "loss": 1.0149,
+      "step": 2566
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831933085090414e-06,
+      "loss": 0.7425,
+      "step": 2567
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831754110044393e-06,
+      "loss": 0.8367,
+      "step": 2568
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831575041384232e-06,
+      "loss": 0.8348,
+      "step": 2569
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8313958791134e-06,
+      "loss": 0.8452,
+      "step": 2570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831216623235369e-06,
+      "loss": 1.0034,
+      "step": 2571
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831037273753612e-06,
+      "loss": 0.9796,
+      "step": 2572
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830857830671605e-06,
+      "loss": 1.2685,
+      "step": 2573
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83067829399282e-06,
+      "loss": 1.0048,
+      "step": 2574
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830498663720743e-06,
+      "loss": 1.0197,
+      "step": 2575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830318939858846e-06,
+      "loss": 1.0678,
+      "step": 2576
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830139122410619e-06,
+      "loss": 0.9935,
+      "step": 2577
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82995921137954e-06,
+      "loss": 0.9237,
+      "step": 2578
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829779206769099e-06,
+      "loss": 1.1212,
+      "step": 2579
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829599108582779e-06,
+      "loss": 0.9437,
+      "step": 2580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829418916824073e-06,
+      "loss": 1.1019,
+      "step": 2581
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82923863149647e-06,
+      "loss": 0.956,
+      "step": 2582
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829058252603465e-06,
+      "loss": 1.0942,
+      "step": 2583
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828877780148551e-06,
+      "loss": 1.0656,
+      "step": 2584
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828697214135226e-06,
+      "loss": 1.1379,
+      "step": 2585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828516554566988e-06,
+      "loss": 1.0842,
+      "step": 2586
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828335801447338e-06,
+      "loss": 0.9282,
+      "step": 2587
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828154954779774e-06,
+      "loss": 0.8907,
+      "step": 2588
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827974014567805e-06,
+      "loss": 0.9798,
+      "step": 2589
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827792980814934e-06,
+      "loss": 1.0113,
+      "step": 2590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82761185352467e-06,
+      "loss": 1.0468,
+      "step": 2591
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82743063270052e-06,
+      "loss": 1.0255,
+      "step": 2592
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827249318345999e-06,
+      "loss": 1.2465,
+      "step": 2593
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827067910464616e-06,
+      "loss": 0.9458,
+      "step": 2594
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826886409059885e-06,
+      "loss": 1.0273,
+      "step": 2595
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826704814135329e-06,
+      "loss": 0.9551,
+      "step": 2596
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82652312569446e-06,
+      "loss": 0.9983,
+      "step": 2597
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8263413437408e-06,
+      "loss": 0.8976,
+      "step": 2598
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82615946827787e-06,
+      "loss": 1.0789,
+      "step": 2599
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825977499309198e-06,
+      "loss": 0.8466,
+      "step": 2600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825795436838304e-06,
+      "loss": 1.0585,
+      "step": 2601
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82561328086872e-06,
+      "loss": 1.0027,
+      "step": 2602
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825431031403972e-06,
+      "loss": 0.9843,
+      "step": 2603
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825248688447593e-06,
+      "loss": 1.0154,
+      "step": 2604
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825066252003116e-06,
+      "loss": 0.9904,
+      "step": 2605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824883722074074e-06,
+      "loss": 1.0434,
+      "step": 2606
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824701098664004e-06,
+      "loss": 0.8894,
+      "step": 2607
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824518381776445e-06,
+      "loss": 1.0153,
+      "step": 2608
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824335571414937e-06,
+      "loss": 1.1768,
+      "step": 2609
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824152667583022e-06,
+      "loss": 1.0825,
+      "step": 2610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823969670284242e-06,
+      "loss": 1.0285,
+      "step": 2611
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823786579522145e-06,
+      "loss": 0.835,
+      "step": 2612
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823603395300277e-06,
+      "loss": 1.0315,
+      "step": 2613
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823420117622186e-06,
+      "loss": 1.1074,
+      "step": 2614
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823236746491427e-06,
+      "loss": 1.0932,
+      "step": 2615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82305328191155e-06,
+      "loss": 1.1886,
+      "step": 2616
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822869723886107e-06,
+      "loss": 1.0313,
+      "step": 2617
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822686072418659e-06,
+      "loss": 0.9382,
+      "step": 2618
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822502327512762e-06,
+      "loss": 0.9878,
+      "step": 2619
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822318489171976e-06,
+      "loss": 0.841,
+      "step": 2620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822134557399863e-06,
+      "loss": 1.0803,
+      "step": 2621
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82195053219999e-06,
+      "loss": 0.9409,
+      "step": 2622
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821766413575915e-06,
+      "loss": 0.8514,
+      "step": 2623
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82158220153121e-06,
+      "loss": 0.9896,
+      "step": 2624
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821397896069444e-06,
+      "loss": 0.8591,
+      "step": 2625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821213497194188e-06,
+      "loss": 1.022,
+      "step": 2626
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821029004909016e-06,
+      "loss": 1.0475,
+      "step": 2627
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820844419217498e-06,
+      "loss": 0.8781,
+      "step": 2628
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820659740123214e-06,
+      "loss": 0.8995,
+      "step": 2629
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82047496762974e-06,
+      "loss": 0.9708,
+      "step": 2630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820290101740657e-06,
+      "loss": 0.8883,
+      "step": 2631
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820105142459549e-06,
+      "loss": 1.0147,
+      "step": 2632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819920089789996e-06,
+      "loss": 0.7254,
+      "step": 2633
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819734943735583e-06,
+      "loss": 0.9489,
+      "step": 2634
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8195497042999e-06,
+      "loss": 1.0979,
+      "step": 2635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819364371486536e-06,
+      "loss": 0.8512,
+      "step": 2636
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819178945299078e-06,
+      "loss": 1.0283,
+      "step": 2637
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818993425741122e-06,
+      "loss": 1.1168,
+      "step": 2638
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818807812816263e-06,
+      "loss": 1.129,
+      "step": 2639
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818622106528094e-06,
+      "loss": 0.8409,
+      "step": 2640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818436306880214e-06,
+      "loss": 1.0398,
+      "step": 2641
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818250413876226e-06,
+      "loss": 0.9288,
+      "step": 2642
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818064427519727e-06,
+      "loss": 1.0952,
+      "step": 2643
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817878347814325e-06,
+      "loss": 1.0192,
+      "step": 2644
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817692174763621e-06,
+      "loss": 1.0966,
+      "step": 2645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817505908371225e-06,
+      "loss": 1.1038,
+      "step": 2646
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817319548640745e-06,
+      "loss": 1.2072,
+      "step": 2647
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817133095575791e-06,
+      "loss": 1.056,
+      "step": 2648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816946549179976e-06,
+      "loss": 1.1456,
+      "step": 2649
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816759909456912e-06,
+      "loss": 1.042,
+      "step": 2650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81657317641022e-06,
+      "loss": 1.0293,
+      "step": 2651
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816386350043516e-06,
+      "loss": 0.8481,
+      "step": 2652
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816199430360417e-06,
+      "loss": 1.0516,
+      "step": 2653
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816012417364546e-06,
+      "loss": 1.082,
+      "step": 2654
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81582531105953e-06,
+      "loss": 0.936,
+      "step": 2655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815638111448989e-06,
+      "loss": 1.0841,
+      "step": 2656
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815450818536551e-06,
+      "loss": 0.8886,
+      "step": 2657
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815263432325848e-06,
+      "loss": 1.0617,
+      "step": 2658
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815075952820507e-06,
+      "loss": 1.0257,
+      "step": 2659
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814888380024161e-06,
+      "loss": 0.9841,
+      "step": 2660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814700713940445e-06,
+      "loss": 0.9578,
+      "step": 2661
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814512954572997e-06,
+      "loss": 1.1284,
+      "step": 2662
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81432510192545e-06,
+      "loss": 0.899,
+      "step": 2663
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814137156001447e-06,
+      "loss": 1.1227,
+      "step": 2664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81394911680463e-06,
+      "loss": 1.0412,
+      "step": 2665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81376098433864e-06,
+      "loss": 0.9245,
+      "step": 2666
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813572758607121e-06,
+      "loss": 0.9454,
+      "step": 2667
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813384439613724e-06,
+      "loss": 1.1357,
+      "step": 2668
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813196027362095e-06,
+      "loss": 1.0367,
+      "step": 2669
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813007521855884e-06,
+      "loss": 0.8568,
+      "step": 2670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812818923098745e-06,
+      "loss": 1.0193,
+      "step": 2671
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81263023109433e-06,
+      "loss": 1.0525,
+      "step": 2672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812441445846297e-06,
+      "loss": 1.2425,
+      "step": 2673
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812252567358302e-06,
+      "loss": 1.0007,
+      "step": 2674
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812063595634007e-06,
+      "loss": 0.9505,
+      "step": 2675
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811874530677069e-06,
+      "loss": 0.9935,
+      "step": 2676
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811685372491153e-06,
+      "loss": 0.9819,
+      "step": 2677
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811496121079926e-06,
+      "loss": 0.8977,
+      "step": 2678
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811306776447054e-06,
+      "loss": 1.0289,
+      "step": 2679
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811117338596203e-06,
+      "loss": 0.8399,
+      "step": 2680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810927807531044e-06,
+      "loss": 1.1077,
+      "step": 2681
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810738183255251e-06,
+      "loss": 1.1024,
+      "step": 2682
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810548465772498e-06,
+      "loss": 1.0107,
+      "step": 2683
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810358655086459e-06,
+      "loss": 1.0888,
+      "step": 2684
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.81016875120081e-06,
+      "loss": 1.0347,
+      "step": 2685
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809978754119236e-06,
+      "loss": 0.9278,
+      "step": 2686
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809788663845414e-06,
+      "loss": 0.8688,
+      "step": 2687
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809598480383026e-06,
+      "loss": 0.9899,
+      "step": 2688
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809408203735759e-06,
+      "loss": 0.9154,
+      "step": 2689
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.8092178339073e-06,
+      "loss": 0.8951,
+      "step": 2690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809027370901336e-06,
+      "loss": 1.0834,
+      "step": 2691
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808836814721558e-06,
+      "loss": 0.8499,
+      "step": 2692
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808646165371657e-06,
+      "loss": 0.9216,
+      "step": 2693
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808455422855327e-06,
+      "loss": 1.0086,
+      "step": 2694
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808264587176263e-06,
+      "loss": 0.9863,
+      "step": 2695
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808073658338167e-06,
+      "loss": 0.9655,
+      "step": 2696
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80788263634473e-06,
+      "loss": 1.1629,
+      "step": 2697
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80769152119966e-06,
+      "loss": 0.8364,
+      "step": 2698
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807500312906656e-06,
+      "loss": 0.8614,
+      "step": 2699
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807309011469426e-06,
+      "loss": 0.9284,
+      "step": 2700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807117616891674e-06,
+      "loss": 1.0179,
+      "step": 2701
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806926129177107e-06,
+      "loss": 0.9673,
+      "step": 2702
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806734548329437e-06,
+      "loss": 0.9488,
+      "step": 2703
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806542874352377e-06,
+      "loss": 1.088,
+      "step": 2704
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806351107249639e-06,
+      "loss": 0.9902,
+      "step": 2705
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806159247024937e-06,
+      "loss": 0.9751,
+      "step": 2706
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80596729368199e-06,
+      "loss": 1.1199,
+      "step": 2707
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805775247224518e-06,
+      "loss": 1.1359,
+      "step": 2708
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805583107656241e-06,
+      "loss": 0.9431,
+      "step": 2709
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80539087498088e-06,
+      "loss": 1.039,
+      "step": 2710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80519854920216e-06,
+      "loss": 0.9147,
+      "step": 2711
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80500613032381e-06,
+      "loss": 1.1184,
+      "step": 2712
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804813618349556e-06,
+      "loss": 1.0339,
+      "step": 2713
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804621013283127e-06,
+      "loss": 0.8504,
+      "step": 2714
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804428315128256e-06,
+      "loss": 0.9783,
+      "step": 2715
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804235523888676e-06,
+      "loss": 1.1169,
+      "step": 2716
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804042639568123e-06,
+      "loss": 1.0304,
+      "step": 2717
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803849662170336e-06,
+      "loss": 0.9658,
+      "step": 2718
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803656591699048e-06,
+      "loss": 0.9679,
+      "step": 2719
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803463428158005e-06,
+      "loss": 1.0716,
+      "step": 2720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803270171550948e-06,
+      "loss": 1.0131,
+      "step": 2721
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803076821881619e-06,
+      "loss": 0.8996,
+      "step": 2722
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802883379153768e-06,
+      "loss": 0.8773,
+      "step": 2723
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802689843371138e-06,
+      "loss": 1.0039,
+      "step": 2724
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802496214537485e-06,
+      "loss": 1.1316,
+      "step": 2725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802302492656556e-06,
+      "loss": 1.0438,
+      "step": 2726
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802108677732106e-06,
+      "loss": 0.8656,
+      "step": 2727
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801914769767888e-06,
+      "loss": 0.8662,
+      "step": 2728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801720768767663e-06,
+      "loss": 0.891,
+      "step": 2729
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801526674735187e-06,
+      "loss": 1.0668,
+      "step": 2730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801332487674218e-06,
+      "loss": 1.1309,
+      "step": 2731
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801138207588525e-06,
+      "loss": 1.1141,
+      "step": 2732
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800943834481866e-06,
+      "loss": 1.1323,
+      "step": 2733
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80074936835801e-06,
+      "loss": 1.0242,
+      "step": 2734
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800554809220724e-06,
+      "loss": 1.0343,
+      "step": 2735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800360157073775e-06,
+      "loss": 0.9831,
+      "step": 2736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80016541192094e-06,
+      "loss": 0.9801,
+      "step": 2737
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799970573765989e-06,
+      "loss": 1.0135,
+      "step": 2738
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799775642612696e-06,
+      "loss": 1.0322,
+      "step": 2739
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799580618464839e-06,
+      "loss": 1.1699,
+      "step": 2740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799385501326196e-06,
+      "loss": 1.0137,
+      "step": 2741
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799190291200548e-06,
+      "loss": 1.0056,
+      "step": 2742
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798994988091677e-06,
+      "loss": 0.9684,
+      "step": 2743
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798799592003366e-06,
+      "loss": 0.998,
+      "step": 2744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798604102939401e-06,
+      "loss": 0.9914,
+      "step": 2745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798408520903571e-06,
+      "loss": 0.9391,
+      "step": 2746
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798212845899665e-06,
+      "loss": 0.874,
+      "step": 2747
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79801707793147e-06,
+      "loss": 0.835,
+      "step": 2748
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797821217002786e-06,
+      "loss": 1.0557,
+      "step": 2749
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797625263117403e-06,
+      "loss": 0.9515,
+      "step": 2750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797429216279119e-06,
+      "loss": 0.9343,
+      "step": 2751
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797233076491732e-06,
+      "loss": 0.9651,
+      "step": 2752
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797036843759043e-06,
+      "loss": 0.8364,
+      "step": 2753
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796840518084853e-06,
+      "loss": 1.0921,
+      "step": 2754
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796644099472965e-06,
+      "loss": 1.0969,
+      "step": 2755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796447587927187e-06,
+      "loss": 1.0231,
+      "step": 2756
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796250983451325e-06,
+      "loss": 0.8944,
+      "step": 2757
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796054286049188e-06,
+      "loss": 0.9152,
+      "step": 2758
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795857495724585e-06,
+      "loss": 0.8455,
+      "step": 2759
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795660612481334e-06,
+      "loss": 1.0789,
+      "step": 2760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795463636323244e-06,
+      "loss": 1.0111,
+      "step": 2761
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795266567254136e-06,
+      "loss": 0.9726,
+      "step": 2762
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795069405277825e-06,
+      "loss": 1.0024,
+      "step": 2763
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794872150398132e-06,
+      "loss": 0.9436,
+      "step": 2764
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794674802618878e-06,
+      "loss": 1.2432,
+      "step": 2765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794477361943888e-06,
+      "loss": 1.181,
+      "step": 2766
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794279828376987e-06,
+      "loss": 0.968,
+      "step": 2767
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794082201922001e-06,
+      "loss": 0.798,
+      "step": 2768
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79388448258276e-06,
+      "loss": 1.0142,
+      "step": 2769
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793686670363097e-06,
+      "loss": 0.9615,
+      "step": 2770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793488765266838e-06,
+      "loss": 1.1387,
+      "step": 2771
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793290767297825e-06,
+      "loss": 0.7847,
+      "step": 2772
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79309267645989e-06,
+      "loss": 0.81,
+      "step": 2773
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79289449275687e-06,
+      "loss": 0.8051,
+      "step": 2774
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792696216192608e-06,
+      "loss": 1.0723,
+      "step": 2775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792497846770943e-06,
+      "loss": 1.0707,
+      "step": 2776
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79229938449572e-06,
+      "loss": 1.0695,
+      "step": 2777
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792100829370783e-06,
+      "loss": 1.0333,
+      "step": 2778
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791902181399979e-06,
+      "loss": 1.0689,
+      "step": 2779
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791703440587159e-06,
+      "loss": 1.0891,
+      "step": 2780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791504606936171e-06,
+      "loss": 1.0062,
+      "step": 2781
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791305680450868e-06,
+      "loss": 1.1041,
+      "step": 2782
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791106661135104e-06,
+      "loss": 0.8275,
+      "step": 2783
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790907548992735e-06,
+      "loss": 0.9981,
+      "step": 2784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79070834402762e-06,
+      "loss": 0.9505,
+      "step": 2785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790509046243617e-06,
+      "loss": 0.995,
+      "step": 2786
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790309655644587e-06,
+      "loss": 0.8745,
+      "step": 2787
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790110172234396e-06,
+      "loss": 0.8687,
+      "step": 2788
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789910596016905e-06,
+      "loss": 0.7725,
+      "step": 2789
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789710926995983e-06,
+      "loss": 0.9004,
+      "step": 2790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789511165175499e-06,
+      "loss": 1.1418,
+      "step": 2791
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789311310559322e-06,
+      "loss": 0.8064,
+      "step": 2792
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789111363151326e-06,
+      "loss": 1.0367,
+      "step": 2793
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788911322955383e-06,
+      "loss": 0.994,
+      "step": 2794
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78871118997537e-06,
+      "loss": 0.9297,
+      "step": 2795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788510964215163e-06,
+      "loss": 1.0777,
+      "step": 2796
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788310645678643e-06,
+      "loss": 0.9686,
+      "step": 2797
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78811023436969e-06,
+      "loss": 0.8875,
+      "step": 2798
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787909730292186e-06,
+      "loss": 0.9741,
+      "step": 2799
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787709133450018e-06,
+      "loss": 0.9954,
+      "step": 2800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787508443847073e-06,
+      "loss": 1.007,
+      "step": 2801
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787307661487237e-06,
+      "loss": 0.8782,
+      "step": 2802
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.7871067863744e-06,
+      "loss": 0.9694,
+      "step": 2803
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786905818512457e-06,
+      "loss": 0.9148,
+      "step": 2804
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786704757905298e-06,
+      "loss": 1.0401,
+      "step": 2805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78650360455682e-06,
+      "loss": 0.9862,
+      "step": 2806
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78630235847092e-06,
+      "loss": 0.9041,
+      "step": 2807
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786101019651499e-06,
+      "loss": 1.1949,
+      "step": 2808
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785899588102456e-06,
+      "loss": 1.1883,
+      "step": 2809
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785698063827695e-06,
+      "loss": 1.1204,
+      "step": 2810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785496446831117e-06,
+      "loss": 1.0954,
+      "step": 2811
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785294737116632e-06,
+      "loss": 0.9323,
+      "step": 2812
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785092934688149e-06,
+      "loss": 1.0699,
+      "step": 2813
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784891039549573e-06,
+      "loss": 0.8906,
+      "step": 2814
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78468905170482e-06,
+      "loss": 0.8273,
+      "step": 2815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784486971157803e-06,
+      "loss": 1.0161,
+      "step": 2816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784284797912434e-06,
+      "loss": 0.919,
+      "step": 2817
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784082531972635e-06,
+      "loss": 1.0249,
+      "step": 2818
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78388017334232e-06,
+      "loss": 1.0053,
+      "step": 2819
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78367772202541e-06,
+      "loss": 0.9201,
+      "step": 2820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783475178025831e-06,
+      "loss": 0.957,
+      "step": 2821
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783272541347508e-06,
+      "loss": 1.1176,
+      "step": 2822
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783069811994362e-06,
+      "loss": 0.865,
+      "step": 2823
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782866989970324e-06,
+      "loss": 1.0543,
+      "step": 2824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782664075279322e-06,
+      "loss": 0.942,
+      "step": 2825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782461067925291e-06,
+      "loss": 0.9686,
+      "step": 2826
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782257967912159e-06,
+      "loss": 1.0296,
+      "step": 2827
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782054775243866e-06,
+      "loss": 0.9952,
+      "step": 2828
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781851489924342e-06,
+      "loss": 1.1069,
+      "step": 2829
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781648111957536e-06,
+      "loss": 1.0317,
+      "step": 2830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781444641347379e-06,
+      "loss": 0.8901,
+      "step": 2831
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781241078097816e-06,
+      "loss": 1.1658,
+      "step": 2832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781037422212793e-06,
+      "loss": 1.0448,
+      "step": 2833
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780833673696255e-06,
+      "loss": 1.1415,
+      "step": 2834
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780629832552148e-06,
+      "loss": 0.9158,
+      "step": 2835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780425898784425e-06,
+      "loss": 0.8939,
+      "step": 2836
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780221872397031e-06,
+      "loss": 0.9143,
+      "step": 2837
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780017753393925e-06,
+      "loss": 0.9719,
+      "step": 2838
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779813541779058e-06,
+      "loss": 0.9054,
+      "step": 2839
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779609237556389e-06,
+      "loss": 1.0897,
+      "step": 2840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779404840729874e-06,
+      "loss": 0.8566,
+      "step": 2841
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779200351303475e-06,
+      "loss": 0.9011,
+      "step": 2842
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778995769281154e-06,
+      "loss": 0.8911,
+      "step": 2843
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778791094666873e-06,
+      "loss": 0.9219,
+      "step": 2844
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.7785863274646e-06,
+      "loss": 1.0799,
+      "step": 2845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778381467678297e-06,
+      "loss": 0.8747,
+      "step": 2846
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77817651531194e-06,
+      "loss": 1.0556,
+      "step": 2847
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777971470369496e-06,
+      "loss": 0.9248,
+      "step": 2848
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77776633285494e-06,
+      "loss": 0.9731,
+      "step": 2849
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777561102772243e-06,
+      "loss": 0.9268,
+      "step": 2850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777355780125385e-06,
+      "loss": 0.9356,
+      "step": 2851
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777150364918341e-06,
+      "loss": 0.8313,
+      "step": 2852
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776944857155096e-06,
+      "loss": 0.9627,
+      "step": 2853
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776739256839625e-06,
+      "loss": 0.9902,
+      "step": 2854
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776533563975914e-06,
+      "loss": 1.0809,
+      "step": 2855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776327778567949e-06,
+      "loss": 1.0379,
+      "step": 2856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776121900619718e-06,
+      "loss": 1.1284,
+      "step": 2857
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775915930135207e-06,
+      "loss": 1.075,
+      "step": 2858
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775709867118409e-06,
+      "loss": 1.1296,
+      "step": 2859
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775503711573315e-06,
+      "loss": 1.1595,
+      "step": 2860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775297463503921e-06,
+      "loss": 0.9753,
+      "step": 2861
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775091122914223e-06,
+      "loss": 0.9466,
+      "step": 2862
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774884689808214e-06,
+      "loss": 0.8777,
+      "step": 2863
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.7746781641899e-06,
+      "loss": 1.0668,
+      "step": 2864
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77447154606328e-06,
+      "loss": 0.8245,
+      "step": 2865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774264835432354e-06,
+      "loss": 1.0203,
+      "step": 2866
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774058032301132e-06,
+      "loss": 1.0574,
+      "step": 2867
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773851136673619e-06,
+      "loss": 0.9641,
+      "step": 2868
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773644148553822e-06,
+      "loss": 0.8597,
+      "step": 2869
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773437067945752e-06,
+      "loss": 0.9334,
+      "step": 2870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773229894853421e-06,
+      "loss": 0.9131,
+      "step": 2871
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773022629280846e-06,
+      "loss": 0.8625,
+      "step": 2872
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772815271232039e-06,
+      "loss": 0.9774,
+      "step": 2873
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772607820711018e-06,
+      "loss": 1.1137,
+      "step": 2874
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772400277721805e-06,
+      "loss": 1.0164,
+      "step": 2875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772192642268418e-06,
+      "loss": 0.9557,
+      "step": 2876
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771984914354882e-06,
+      "loss": 0.9949,
+      "step": 2877
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77177709398522e-06,
+      "loss": 0.8739,
+      "step": 2878
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771569181163458e-06,
+      "loss": 0.9047,
+      "step": 2879
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771361175893626e-06,
+      "loss": 1.1165,
+      "step": 2880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771153078179756e-06,
+      "loss": 0.9201,
+      "step": 2881
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770944888025874e-06,
+      "loss": 0.9619,
+      "step": 2882
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77073660543602e-06,
+      "loss": 0.9706,
+      "step": 2883
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770528230414225e-06,
+      "loss": 1.1002,
+      "step": 2884
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770319762964526e-06,
+      "loss": 0.8144,
+      "step": 2885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770111203090965e-06,
+      "loss": 1.0341,
+      "step": 2886
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769902550797584e-06,
+      "loss": 0.875,
+      "step": 2887
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769693806088421e-06,
+      "loss": 0.9062,
+      "step": 2888
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769484968967522e-06,
+      "loss": 0.808,
+      "step": 2889
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769276039438932e-06,
+      "loss": 1.0016,
+      "step": 2890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769067017506703e-06,
+      "loss": 0.9555,
+      "step": 2891
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768857903174884e-06,
+      "loss": 1.142,
+      "step": 2892
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768648696447523e-06,
+      "loss": 0.9487,
+      "step": 2893
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768439397328674e-06,
+      "loss": 0.8667,
+      "step": 2894
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768230005822394e-06,
+      "loss": 1.03,
+      "step": 2895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768020521932739e-06,
+      "loss": 0.8868,
+      "step": 2896
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.76781094566377e-06,
+      "loss": 0.9433,
+      "step": 2897
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767601277019544e-06,
+      "loss": 0.9165,
+      "step": 2898
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767391516004124e-06,
+      "loss": 0.9793,
+      "step": 2899
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767181662621576e-06,
+      "loss": 1.0926,
+      "step": 2900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766971716875964e-06,
+      "loss": 1.1068,
+      "step": 2901
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766761678771358e-06,
+      "loss": 1.086,
+      "step": 2902
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766551548311823e-06,
+      "loss": 0.785,
+      "step": 2903
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766341325501433e-06,
+      "loss": 1.0696,
+      "step": 2904
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766131010344264e-06,
+      "loss": 1.1058,
+      "step": 2905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765920602844386e-06,
+      "loss": 0.9969,
+      "step": 2906
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765710103005877e-06,
+      "loss": 1.0312,
+      "step": 2907
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765499510832816e-06,
+      "loss": 1.1366,
+      "step": 2908
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.765288826329285e-06,
+      "loss": 0.961,
+      "step": 2909
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.765078049499361e-06,
+      "loss": 0.9219,
+      "step": 2910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764867180347133e-06,
+      "loss": 0.9443,
+      "step": 2911
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764656218876683e-06,
+      "loss": 0.9248,
+      "step": 2912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.7644451650921e-06,
+      "loss": 1.0073,
+      "step": 2913
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764234018997472e-06,
+      "loss": 1.0814,
+      "step": 2914
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76402278059689e-06,
+      "loss": 0.8927,
+      "step": 2915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76381144989445e-06,
+      "loss": 0.9368,
+      "step": 2916
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76360002689424e-06,
+      "loss": 1.1807,
+      "step": 2917
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763388511600364e-06,
+      "loss": 0.8466,
+      "step": 2918
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763176904016914e-06,
+      "loss": 1.0061,
+      "step": 2919
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762965204147992e-06,
+      "loss": 1.0353,
+      "step": 2920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.7627534119977e-06,
+      "loss": 1.0345,
+      "step": 2921
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76254152757014e-06,
+      "loss": 1.1022,
+      "step": 2922
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76232955086942e-06,
+      "loss": 1.04,
+      "step": 2923
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762117481899643e-06,
+      "loss": 0.899,
+      "step": 2924
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761905320664921e-06,
+      "loss": 0.9914,
+      "step": 2925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761693067169365e-06,
+      "loss": 1.0142,
+      "step": 2926
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761480721417085e-06,
+      "loss": 0.8685,
+      "step": 2927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761268283412194e-06,
+      "loss": 1.0599,
+      "step": 2928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761055753158814e-06,
+      "loss": 0.9546,
+      "step": 2929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760843130661057e-06,
+      "loss": 0.924,
+      "step": 2930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760630415923043e-06,
+      "loss": 1.0519,
+      "step": 2931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760417608948896e-06,
+      "loss": 0.8488,
+      "step": 2932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760204709742738e-06,
+      "loss": 0.984,
+      "step": 2933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759991718308693e-06,
+      "loss": 1.0806,
+      "step": 2934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759778634650887e-06,
+      "loss": 1.2042,
+      "step": 2935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75956545877345e-06,
+      "loss": 0.9601,
+      "step": 2936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759352190680514e-06,
+      "loss": 1.0147,
+      "step": 2937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759138830376207e-06,
+      "loss": 1.0697,
+      "step": 2938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758925377864665e-06,
+      "loss": 1.2477,
+      "step": 2939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758711833150023e-06,
+      "loss": 1.025,
+      "step": 2940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758498196236418e-06,
+      "loss": 1.1126,
+      "step": 2941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758284467127992e-06,
+      "loss": 0.9627,
+      "step": 2942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758070645828884e-06,
+      "loss": 1.0821,
+      "step": 2943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757856732343234e-06,
+      "loss": 1.1479,
+      "step": 2944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757642726675191e-06,
+      "loss": 1.0384,
+      "step": 2945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757428628828899e-06,
+      "loss": 0.9153,
+      "step": 2946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757214438808507e-06,
+      "loss": 0.9147,
+      "step": 2947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757000156618162e-06,
+      "loss": 1.0801,
+      "step": 2948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75678578226202e-06,
+      "loss": 1.0989,
+      "step": 2949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756571315744233e-06,
+      "loss": 0.9466,
+      "step": 2950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756356757068954e-06,
+      "loss": 0.9618,
+      "step": 2951
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756142106240342e-06,
+      "loss": 0.8984,
+      "step": 2952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755927363262558e-06,
+      "loss": 1.14,
+      "step": 2953
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755712528139757e-06,
+      "loss": 0.8882,
+      "step": 2954
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755497600876106e-06,
+      "loss": 0.9353,
+      "step": 2955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755282581475769e-06,
+      "loss": 0.9531,
+      "step": 2956
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755067469942909e-06,
+      "loss": 1.0182,
+      "step": 2957
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754852266281696e-06,
+      "loss": 0.9354,
+      "step": 2958
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754636970496298e-06,
+      "loss": 1.07,
+      "step": 2959
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754421582590888e-06,
+      "loss": 0.959,
+      "step": 2960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754206102569639e-06,
+      "loss": 1.0459,
+      "step": 2961
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753990530436724e-06,
+      "loss": 1.0079,
+      "step": 2962
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75377486619632e-06,
+      "loss": 1.004,
+      "step": 2963
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753559109852608e-06,
+      "loss": 0.8301,
+      "step": 2964
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753343261409766e-06,
+      "loss": 1.2224,
+      "step": 2965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753127320871978e-06,
+      "loss": 1.093,
+      "step": 2966
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752911288243426e-06,
+      "loss": 0.9984,
+      "step": 2967
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752695163528295e-06,
+      "loss": 1.0503,
+      "step": 2968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752478946730772e-06,
+      "loss": 0.8993,
+      "step": 2969
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752262637855048e-06,
+      "loss": 1.1101,
+      "step": 2970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752046236905315e-06,
+      "loss": 0.961,
+      "step": 2971
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751829743885762e-06,
+      "loss": 0.9997,
+      "step": 2972
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751613158800587e-06,
+      "loss": 0.8642,
+      "step": 2973
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751396481653985e-06,
+      "loss": 0.816,
+      "step": 2974
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751179712450154e-06,
+      "loss": 1.0347,
+      "step": 2975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750962851193293e-06,
+      "loss": 1.1537,
+      "step": 2976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750745897887605e-06,
+      "loss": 1.1045,
+      "step": 2977
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750528852537293e-06,
+      "loss": 0.8765,
+      "step": 2978
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75031171514656e-06,
+      "loss": 1.1705,
+      "step": 2979
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750094485719619e-06,
+      "loss": 0.9176,
+      "step": 2980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749877164260673e-06,
+      "loss": 0.9529,
+      "step": 2981
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749659750773933e-06,
+      "loss": 1.0191,
+      "step": 2982
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749442245263614e-06,
+      "loss": 1.0838,
+      "step": 2983
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749224647733929e-06,
+      "loss": 1.0594,
+      "step": 2984
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749006958189094e-06,
+      "loss": 0.9625,
+      "step": 2985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748789176633326e-06,
+      "loss": 1.035,
+      "step": 2986
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748571303070844e-06,
+      "loss": 0.9873,
+      "step": 2987
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748353337505871e-06,
+      "loss": 0.9494,
+      "step": 2988
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74813527994263e-06,
+      "loss": 0.8391,
+      "step": 2989
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.747917130385344e-06,
+      "loss": 1.0439,
+      "step": 2990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74769888883824e-06,
+      "loss": 0.9232,
+      "step": 2991
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.747480555305549e-06,
+      "loss": 1.0202,
+      "step": 2992
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.747262129791497e-06,
+      "loss": 1.1077,
+      "step": 2993
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74704361230032e-06,
+      "loss": 1.0887,
+      "step": 2994
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.746825002836249e-06,
+      "loss": 1.0159,
+      "step": 2995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.746606301403519e-06,
+      "loss": 1.1232,
+      "step": 2996
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74638750800637e-06,
+      "loss": 1.0099,
+      "step": 2997
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74616862264904e-06,
+      "loss": 0.84,
+      "step": 2998
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745949645335768e-06,
+      "loss": 0.9325,
+      "step": 2999
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.7457305760708e-06,
+      "loss": 0.9051,
+      "step": 3000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745511414858377e-06,
+      "loss": 0.8013,
+      "step": 3001
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745292161702748e-06,
+      "loss": 0.9986,
+      "step": 3002
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74507281660816e-06,
+      "loss": 0.9538,
+      "step": 3003
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744853379578862e-06,
+      "loss": 0.9665,
+      "step": 3004
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744633850619107e-06,
+      "loss": 0.9432,
+      "step": 3005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744414229733146e-06,
+      "loss": 0.9616,
+      "step": 3006
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744194516925237e-06,
+      "loss": 0.9939,
+      "step": 3007
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743974712199633e-06,
+      "loss": 0.8283,
+      "step": 3008
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743754815560595e-06,
+      "loss": 1.0202,
+      "step": 3009
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743534827012387e-06,
+      "loss": 0.9067,
+      "step": 3010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743314746559265e-06,
+      "loss": 0.796,
+      "step": 3011
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743094574205497e-06,
+      "loss": 0.8825,
+      "step": 3012
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742874309955346e-06,
+      "loss": 0.902,
+      "step": 3013
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742653953813082e-06,
+      "loss": 1.0063,
+      "step": 3014
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742433505782974e-06,
+      "loss": 0.8158,
+      "step": 3015
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742212965869292e-06,
+      "loss": 1.0331,
+      "step": 3016
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741992334076309e-06,
+      "loss": 1.0158,
+      "step": 3017
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741771610408301e-06,
+      "loss": 1.201,
+      "step": 3018
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741550794869543e-06,
+      "loss": 1.114,
+      "step": 3019
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741329887464315e-06,
+      "loss": 0.997,
+      "step": 3020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741108888196894e-06,
+      "loss": 1.0256,
+      "step": 3021
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740887797071566e-06,
+      "loss": 0.9016,
+      "step": 3022
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740666614092612e-06,
+      "loss": 0.9935,
+      "step": 3023
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740445339264318e-06,
+      "loss": 1.1353,
+      "step": 3024
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74022397259097e-06,
+      "loss": 0.9109,
+      "step": 3025
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740002514076858e-06,
+      "loss": 1.0871,
+      "step": 3026
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73978096372627e-06,
+      "loss": 1.0548,
+      "step": 3027
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739559321543505e-06,
+      "loss": 0.9599,
+      "step": 3028
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739337587532852e-06,
+      "loss": 0.905,
+      "step": 3029
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739115761698607e-06,
+      "loss": 0.999,
+      "step": 3030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738893844045069e-06,
+      "loss": 1.1569,
+      "step": 3031
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73867183457654e-06,
+      "loss": 0.8565,
+      "step": 3032
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738449733297318e-06,
+      "loss": 1.0304,
+      "step": 3033
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738227540211706e-06,
+      "loss": 0.9015,
+      "step": 3034
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738005255324012e-06,
+      "loss": 1.0459,
+      "step": 3035
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73778287863854e-06,
+      "loss": 1.0851,
+      "step": 3036
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.7375604101596e-06,
+      "loss": 0.8451,
+      "step": 3037
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737337849891503e-06,
+      "loss": 0.9947,
+      "step": 3038
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737115197838556e-06,
+      "loss": 0.7861,
+      "step": 3039
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73689245400508e-06,
+      "loss": 0.904,
+      "step": 3040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736669618395387e-06,
+      "loss": 1.014,
+      "step": 3041
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736446691013793e-06,
+      "loss": 1.0968,
+      "step": 3042
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736223671864619e-06,
+      "loss": 1.0229,
+      "step": 3043
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736000560952185e-06,
+      "loss": 0.9306,
+      "step": 3044
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735777358280816e-06,
+      "loss": 1.0299,
+      "step": 3045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735554063854835e-06,
+      "loss": 0.9474,
+      "step": 3046
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735330677678567e-06,
+      "loss": 0.9147,
+      "step": 3047
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735107199756343e-06,
+      "loss": 1.113,
+      "step": 3048
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734883630092489e-06,
+      "loss": 1.0731,
+      "step": 3049
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73465996869134e-06,
+      "loss": 1.1633,
+      "step": 3050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734436215557228e-06,
+      "loss": 1.0449,
+      "step": 3051
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734212370694488e-06,
+      "loss": 1.0285,
+      "step": 3052
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733988434107458e-06,
+      "loss": 1.2047,
+      "step": 3053
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733764405800474e-06,
+      "loss": 1.1251,
+      "step": 3054
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73354028577788e-06,
+      "loss": 0.89,
+      "step": 3055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733316074044016e-06,
+      "loss": 0.9947,
+      "step": 3056
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733091770603229e-06,
+      "loss": 0.8804,
+      "step": 3057
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73286737545986e-06,
+      "loss": 1.0813,
+      "step": 3058
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73264288861826e-06,
+      "loss": 1.0569,
+      "step": 3059
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.732418310082776e-06,
+      "loss": 1.0879,
+      "step": 3060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.732193639857762e-06,
+      "loss": 0.8429,
+      "step": 3061
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731968877947571e-06,
+      "loss": 0.9802,
+      "step": 3062
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731744024356555e-06,
+      "loss": 0.8427,
+      "step": 3063
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731519079089074e-06,
+      "loss": 1.0685,
+      "step": 3064
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731294042149482e-06,
+      "loss": 0.8633,
+      "step": 3065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731068913542142e-06,
+      "loss": 1.0665,
+      "step": 3066
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730843693271413e-06,
+      "loss": 1.0239,
+      "step": 3067
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730618381341662e-06,
+      "loss": 0.9739,
+      "step": 3068
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730392977757252e-06,
+      "loss": 0.9751,
+      "step": 3069
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730167482522554e-06,
+      "loss": 0.9571,
+      "step": 3070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729941895641932e-06,
+      "loss": 0.904,
+      "step": 3071
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729716217119759e-06,
+      "loss": 0.962,
+      "step": 3072
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729490446960406e-06,
+      "loss": 0.8726,
+      "step": 3073
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72926458516825e-06,
+      "loss": 0.8693,
+      "step": 3074
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729038631747665e-06,
+      "loss": 0.9489,
+      "step": 3075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72881258670303e-06,
+      "loss": 0.9567,
+      "step": 3076
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.728586450038722e-06,
+      "loss": 1.0197,
+      "step": 3077
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.728360221759125e-06,
+      "loss": 0.9585,
+      "step": 3078
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.728133901868621e-06,
+      "loss": 0.9936,
+      "step": 3079
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727907490371595e-06,
+      "loss": 0.8901,
+      "step": 3080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727680987272434e-06,
+      "loss": 1.0256,
+      "step": 3081
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727454392575528e-06,
+      "loss": 1.0016,
+      "step": 3082
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727227706285263e-06,
+      "loss": 1.0281,
+      "step": 3083
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727000928406036e-06,
+      "loss": 0.9871,
+      "step": 3084
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726774058942236e-06,
+      "loss": 0.8849,
+      "step": 3085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726547097898262e-06,
+      "loss": 0.9127,
+      "step": 3086
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72632004527851e-06,
+      "loss": 1.0802,
+      "step": 3087
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726092901087378e-06,
+      "loss": 0.965,
+      "step": 3088
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725865665329271e-06,
+      "loss": 1.0056,
+      "step": 3089
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725638338008585e-06,
+      "loss": 0.8757,
+      "step": 3090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725410919129731e-06,
+      "loss": 1.0251,
+      "step": 3091
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725183408697112e-06,
+      "loss": 0.9287,
+      "step": 3092
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724955806715136e-06,
+      "loss": 0.9887,
+      "step": 3093
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724728113188212e-06,
+      "loss": 0.9448,
+      "step": 3094
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724500328120753e-06,
+      "loss": 0.999,
+      "step": 3095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724272451517172e-06,
+      "loss": 0.8489,
+      "step": 3096
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724044483381885e-06,
+      "loss": 0.9402,
+      "step": 3097
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723816423719306e-06,
+      "loss": 1.0049,
+      "step": 3098
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723588272533855e-06,
+      "loss": 0.8522,
+      "step": 3099
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723360029829956e-06,
+      "loss": 1.0361,
+      "step": 3100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723131695612025e-06,
+      "loss": 1.0433,
+      "step": 3101
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72290326988449e-06,
+      "loss": 0.9239,
+      "step": 3102
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722674752651774e-06,
+      "loss": 0.9922,
+      "step": 3103
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722446143918307e-06,
+      "loss": 0.8959,
+      "step": 3104
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722217443688517e-06,
+      "loss": 1.0044,
+      "step": 3105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721988651966836e-06,
+      "loss": 0.9229,
+      "step": 3106
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721759768757694e-06,
+      "loss": 0.891,
+      "step": 3107
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72153079406553e-06,
+      "loss": 0.9767,
+      "step": 3108
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721301727894776e-06,
+      "loss": 1.0558,
+      "step": 3109
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721072570249872e-06,
+      "loss": 0.7936,
+      "step": 3110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72084332113526e-06,
+      "loss": 0.9607,
+      "step": 3111
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720613980555378e-06,
+      "loss": 0.9026,
+      "step": 3112
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720384548514671e-06,
+      "loss": 1.0288,
+      "step": 3113
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720155025017583e-06,
+      "loss": 0.9592,
+      "step": 3114
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719925410068565e-06,
+      "loss": 1.1304,
+      "step": 3115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719695703672061e-06,
+      "loss": 1.0329,
+      "step": 3116
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719465905832524e-06,
+      "loss": 1.0699,
+      "step": 3117
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719236016554404e-06,
+      "loss": 1.043,
+      "step": 3118
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719006035842158e-06,
+      "loss": 0.9371,
+      "step": 3119
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71877596370024e-06,
+      "loss": 1.0766,
+      "step": 3120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718545800133107e-06,
+      "loss": 0.8392,
+      "step": 3121
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718315545145224e-06,
+      "loss": 0.8765,
+      "step": 3122
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718085198741043e-06,
+      "loss": 1.0807,
+      "step": 3123
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717854760925033e-06,
+      "loss": 1.0679,
+      "step": 3124
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717624231701657e-06,
+      "loss": 1.0878,
+      "step": 3125
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717393611075383e-06,
+      "loss": 1.0213,
+      "step": 3126
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717162899050676e-06,
+      "loss": 0.9663,
+      "step": 3127
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716932095632008e-06,
+      "loss": 0.8965,
+      "step": 3128
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716701200823852e-06,
+      "loss": 0.8126,
+      "step": 3129
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716470214630679e-06,
+      "loss": 1.111,
+      "step": 3130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716239137056965e-06,
+      "loss": 0.9853,
+      "step": 3131
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716007968107189e-06,
+      "loss": 1.1008,
+      "step": 3132
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715776707785829e-06,
+      "loss": 0.996,
+      "step": 3133
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715545356097362e-06,
+      "loss": 1.2532,
+      "step": 3134
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715313913046276e-06,
+      "loss": 0.8223,
+      "step": 3135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715082378637052e-06,
+      "loss": 0.8223,
+      "step": 3136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714850752874175e-06,
+      "loss": 0.9933,
+      "step": 3137
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714619035762136e-06,
+      "loss": 0.9629,
+      "step": 3138
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714387227305422e-06,
+      "loss": 0.9605,
+      "step": 3139
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714155327508524e-06,
+      "loss": 0.9807,
+      "step": 3140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.713923336375936e-06,
+      "loss": 0.9916,
+      "step": 3141
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.713691253912155e-06,
+      "loss": 1.0708,
+      "step": 3142
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.713459080121674e-06,
+      "loss": 1.1226,
+      "step": 3143
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.713226815008991e-06,
+      "loss": 0.9564,
+      "step": 3144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712994458578608e-06,
+      "loss": 0.8588,
+      "step": 3145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.71276201083503e-06,
+      "loss": 1.0403,
+      "step": 3146
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712529471782752e-06,
+      "loss": 0.8827,
+      "step": 3147
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712296841426288e-06,
+      "loss": 0.9981,
+      "step": 3148
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.71206411977014e-06,
+      "loss": 1.047,
+      "step": 3149
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711831306818817e-06,
+      "loss": 1.0443,
+      "step": 3150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711598402576835e-06,
+      "loss": 0.9409,
+      "step": 3151
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.7113654070487e-06,
+      "loss": 0.9857,
+      "step": 3152
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711132320238929e-06,
+      "loss": 0.8389,
+      "step": 3153
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.71089914215204e-06,
+      "loss": 0.8939,
+      "step": 3154
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710665872792547e-06,
+      "loss": 1.0299,
+      "step": 3155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710432512164972e-06,
+      "loss": 1.0903,
+      "step": 3156
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710199060273836e-06,
+      "loss": 1.0618,
+      "step": 3157
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709965517123661e-06,
+      "loss": 1.2748,
+      "step": 3158
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709731882718972e-06,
+      "loss": 1.0511,
+      "step": 3159
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709498157064297e-06,
+      "loss": 1.0331,
+      "step": 3160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709264340164161e-06,
+      "loss": 0.9388,
+      "step": 3161
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.7090304320231e-06,
+      "loss": 0.9606,
+      "step": 3162
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708796432645642e-06,
+      "loss": 0.8624,
+      "step": 3163
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708562342036319e-06,
+      "loss": 1.0881,
+      "step": 3164
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708328160199671e-06,
+      "loss": 1.079,
+      "step": 3165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708093887140234e-06,
+      "loss": 0.9512,
+      "step": 3166
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707859522862542e-06,
+      "loss": 1.0412,
+      "step": 3167
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707625067371142e-06,
+      "loss": 0.9806,
+      "step": 3168
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707390520670574e-06,
+      "loss": 0.9247,
+      "step": 3169
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707155882765382e-06,
+      "loss": 0.8916,
+      "step": 3170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706921153660114e-06,
+      "loss": 0.9215,
+      "step": 3171
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706686333359315e-06,
+      "loss": 1.0211,
+      "step": 3172
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706451421867536e-06,
+      "loss": 1.0989,
+      "step": 3173
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706216419189329e-06,
+      "loss": 0.8791,
+      "step": 3174
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705981325329246e-06,
+      "loss": 0.8405,
+      "step": 3175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705746140291841e-06,
+      "loss": 1.0282,
+      "step": 3176
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705510864081674e-06,
+      "loss": 1.1551,
+      "step": 3177
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705275496703302e-06,
+      "loss": 1.0875,
+      "step": 3178
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705040038161283e-06,
+      "loss": 0.912,
+      "step": 3179
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70480448846018e-06,
+      "loss": 0.9963,
+      "step": 3180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704568847604558e-06,
+      "loss": 0.8916,
+      "step": 3181
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704333115598982e-06,
+      "loss": 0.8398,
+      "step": 3182
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704097292448018e-06,
+      "loss": 0.6877,
+      "step": 3183
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.703861378156237e-06,
+      "loss": 1.0,
+      "step": 3184
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.703625372728208e-06,
+      "loss": 1.0404,
+      "step": 3185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.703389276168507e-06,
+      "loss": 0.9043,
+      "step": 3186
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.703153088481702e-06,
+      "loss": 0.8786,
+      "step": 3187
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702916809672376e-06,
+      "loss": 0.8787,
+      "step": 3188
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702680439745101e-06,
+      "loss": 0.9115,
+      "step": 3189
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702443978704461e-06,
+      "loss": 0.9622,
+      "step": 3190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702207426555034e-06,
+      "loss": 1.0896,
+      "step": 3191
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.701970783301407e-06,
+      "loss": 1.0191,
+      "step": 3192
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70173404894816e-06,
+      "loss": 0.9741,
+      "step": 3193
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.701497223499886e-06,
+      "loss": 1.0937,
+      "step": 3194
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.701260306961166e-06,
+      "loss": 0.9972,
+      "step": 3195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.701023299336596e-06,
+      "loss": 1.0215,
+      "step": 3196
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700786200630765e-06,
+      "loss": 0.9275,
+      "step": 3197
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700549010848268e-06,
+      "loss": 0.9266,
+      "step": 3198
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.7003117299937e-06,
+      "loss": 1.0179,
+      "step": 3199
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700074358071658e-06,
+      "loss": 0.9006,
+      "step": 3200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699836895086743e-06,
+      "loss": 1.0028,
+      "step": 3201
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699599341043552e-06,
+      "loss": 1.1059,
+      "step": 3202
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699361695946692e-06,
+      "loss": 1.0906,
+      "step": 3203
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699123959800764e-06,
+      "loss": 0.831,
+      "step": 3204
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698886132610377e-06,
+      "loss": 1.0241,
+      "step": 3205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698648214380136e-06,
+      "loss": 1.0221,
+      "step": 3206
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69841020511465e-06,
+      "loss": 1.0204,
+      "step": 3207
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698172104818533e-06,
+      "loss": 1.1643,
+      "step": 3208
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697933913496398e-06,
+      "loss": 0.9549,
+      "step": 3209
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697695631152858e-06,
+      "loss": 0.8881,
+      "step": 3210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697457257792532e-06,
+      "loss": 0.9733,
+      "step": 3211
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697218793420036e-06,
+      "loss": 1.0564,
+      "step": 3212
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696980238039993e-06,
+      "loss": 1.0657,
+      "step": 3213
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696741591657022e-06,
+      "loss": 1.0193,
+      "step": 3214
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69650285427575e-06,
+      "loss": 0.8427,
+      "step": 3215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696264025900796e-06,
+      "loss": 0.9498,
+      "step": 3216
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696025106536796e-06,
+      "loss": 1.0378,
+      "step": 3217
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695786096188373e-06,
+      "loss": 1.0428,
+      "step": 3218
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695546994860161e-06,
+      "loss": 0.9237,
+      "step": 3219
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69530780255679e-06,
+      "loss": 1.1242,
+      "step": 3220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695068519282895e-06,
+      "loss": 0.943,
+      "step": 3221
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694829145043113e-06,
+      "loss": 0.9599,
+      "step": 3222
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694589679842081e-06,
+      "loss": 0.9565,
+      "step": 3223
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69435012368444e-06,
+      "loss": 1.0334,
+      "step": 3224
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694110476574829e-06,
+      "loss": 0.9256,
+      "step": 3225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693870738517893e-06,
+      "loss": 1.0049,
+      "step": 3226
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693630909518275e-06,
+      "loss": 1.0197,
+      "step": 3227
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693390989580624e-06,
+      "loss": 0.8856,
+      "step": 3228
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693150978709586e-06,
+      "loss": 0.9932,
+      "step": 3229
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692910876909814e-06,
+      "loss": 0.9269,
+      "step": 3230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692670684185958e-06,
+      "loss": 0.9303,
+      "step": 3231
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692430400542671e-06,
+      "loss": 0.9727,
+      "step": 3232
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692190025984612e-06,
+      "loss": 0.9344,
+      "step": 3233
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691949560516433e-06,
+      "loss": 0.9604,
+      "step": 3234
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691709004142794e-06,
+      "loss": 1.0369,
+      "step": 3235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691468356868362e-06,
+      "loss": 1.1697,
+      "step": 3236
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691227618697791e-06,
+      "loss": 0.8706,
+      "step": 3237
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69098678963575e-06,
+      "loss": 1.018,
+      "step": 3238
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690745869686904e-06,
+      "loss": 0.952,
+      "step": 3239
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690504858855922e-06,
+      "loss": 1.1188,
+      "step": 3240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690263757147471e-06,
+      "loss": 1.0422,
+      "step": 3241
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690022564566225e-06,
+      "loss": 0.8378,
+      "step": 3242
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689781281116855e-06,
+      "loss": 1.0438,
+      "step": 3243
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689539906804035e-06,
+      "loss": 0.9331,
+      "step": 3244
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689298441632445e-06,
+      "loss": 1.0578,
+      "step": 3245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689056885606761e-06,
+      "loss": 1.0561,
+      "step": 3246
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688815238731664e-06,
+      "loss": 1.114,
+      "step": 3247
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688573501011835e-06,
+      "loss": 0.9551,
+      "step": 3248
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688331672451955e-06,
+      "loss": 0.84,
+      "step": 3249
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688089753056718e-06,
+      "loss": 1.0393,
+      "step": 3250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.6878477428308e-06,
+      "loss": 0.9606,
+      "step": 3251
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687605641778899e-06,
+      "loss": 0.9507,
+      "step": 3252
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.6873634499057e-06,
+      "loss": 0.9688,
+      "step": 3253
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687121167215898e-06,
+      "loss": 0.8926,
+      "step": 3254
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686878793714187e-06,
+      "loss": 0.8574,
+      "step": 3255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686636329405263e-06,
+      "loss": 1.1156,
+      "step": 3256
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686393774293824e-06,
+      "loss": 1.0762,
+      "step": 3257
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686151128384567e-06,
+      "loss": 1.1009,
+      "step": 3258
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685908391682194e-06,
+      "loss": 0.8643,
+      "step": 3259
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68566556419141e-06,
+      "loss": 0.9417,
+      "step": 3260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68542264591692e-06,
+      "loss": 0.9317,
+      "step": 3261
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685179636863427e-06,
+      "loss": 0.9766,
+      "step": 3262
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68493653703564e-06,
+      "loss": 0.9542,
+      "step": 3263
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684693346438274e-06,
+      "loss": 0.891,
+      "step": 3264
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684450065076034e-06,
+      "loss": 1.0949,
+      "step": 3265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68420669295364e-06,
+      "loss": 1.0624,
+      "step": 3266
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683963230075802e-06,
+      "loss": 1.0602,
+      "step": 3267
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683719676447238e-06,
+      "loss": 1.0255,
+      "step": 3268
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683476032072668e-06,
+      "loss": 1.1905,
+      "step": 3269
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683232296956812e-06,
+      "loss": 0.8738,
+      "step": 3270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682988471104393e-06,
+      "loss": 1.0846,
+      "step": 3271
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682744554520136e-06,
+      "loss": 1.0816,
+      "step": 3272
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682500547208764e-06,
+      "loss": 1.0375,
+      "step": 3273
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682256449175008e-06,
+      "loss": 1.0902,
+      "step": 3274
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682012260423593e-06,
+      "loss": 0.949,
+      "step": 3275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.681767980959253e-06,
+      "loss": 0.978,
+      "step": 3276
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68152361078672e-06,
+      "loss": 0.9591,
+      "step": 3277
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.681279149910731e-06,
+      "loss": 0.977,
+      "step": 3278
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68103459833602e-06,
+      "loss": 1.0574,
+      "step": 3279
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680789956067323e-06,
+      "loss": 0.9323,
+      "step": 3280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680545223109386e-06,
+      "loss": 0.8976,
+      "step": 3281
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680300399466944e-06,
+      "loss": 1.0447,
+      "step": 3282
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680055485144744e-06,
+      "loss": 0.8658,
+      "step": 3283
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679810480147532e-06,
+      "loss": 1.0745,
+      "step": 3284
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679565384480051e-06,
+      "loss": 0.9391,
+      "step": 3285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679320198147055e-06,
+      "loss": 1.0068,
+      "step": 3286
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679074921153291e-06,
+      "loss": 0.9356,
+      "step": 3287
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.67882955350351e-06,
+      "loss": 1.1393,
+      "step": 3288
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678584095202468e-06,
+      "loss": 0.9501,
+      "step": 3289
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678338546254921e-06,
+      "loss": 0.8215,
+      "step": 3290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678092906665628e-06,
+      "loss": 0.7257,
+      "step": 3291
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677847176439344e-06,
+      "loss": 0.9977,
+      "step": 3292
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677601355580833e-06,
+      "loss": 0.9769,
+      "step": 3293
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677355444094856e-06,
+      "loss": 1.1099,
+      "step": 3294
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677109441986178e-06,
+      "loss": 0.9793,
+      "step": 3295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676863349259566e-06,
+      "loss": 1.0065,
+      "step": 3296
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676617165919788e-06,
+      "loss": 1.0009,
+      "step": 3297
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676370891971612e-06,
+      "loss": 0.8515,
+      "step": 3298
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676124527419813e-06,
+      "loss": 1.1656,
+      "step": 3299
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.67587807226916e-06,
+      "loss": 1.1661,
+      "step": 3300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675631526524429e-06,
+      "loss": 0.9485,
+      "step": 3301
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675384890190401e-06,
+      "loss": 1.0085,
+      "step": 3302
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675138163271849e-06,
+      "loss": 1.0,
+      "step": 3303
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.674891345773557e-06,
+      "loss": 1.1812,
+      "step": 3304
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.674644437700304e-06,
+      "loss": 0.8605,
+      "step": 3305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.674397439056876e-06,
+      "loss": 0.8616,
+      "step": 3306
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.674150349848058e-06,
+      "loss": 0.8824,
+      "step": 3307
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673903170078636e-06,
+      "loss": 0.9958,
+      "step": 3308
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673655899753403e-06,
+      "loss": 1.0997,
+      "step": 3309
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673408538877145e-06,
+      "loss": 0.9991,
+      "step": 3310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673161087454658e-06,
+      "loss": 0.7942,
+      "step": 3311
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672913545490735e-06,
+      "loss": 1.0537,
+      "step": 3312
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672665912990172e-06,
+      "loss": 0.9274,
+      "step": 3313
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672418189957765e-06,
+      "loss": 0.8114,
+      "step": 3314
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672170376398318e-06,
+      "loss": 1.0581,
+      "step": 3315
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671922472316628e-06,
+      "loss": 1.0863,
+      "step": 3316
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.6716744777175e-06,
+      "loss": 0.9097,
+      "step": 3317
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671426392605739e-06,
+      "loss": 1.0545,
+      "step": 3318
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671178216986153e-06,
+      "loss": 0.9674,
+      "step": 3319
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670929950863545e-06,
+      "loss": 1.0645,
+      "step": 3320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670681594242732e-06,
+      "loss": 0.951,
+      "step": 3321
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670433147128522e-06,
+      "loss": 1.0513,
+      "step": 3322
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670184609525728e-06,
+      "loss": 1.1845,
+      "step": 3323
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669935981439167e-06,
+      "loss": 1.1107,
+      "step": 3324
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669687262873656e-06,
+      "loss": 1.0915,
+      "step": 3325
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669438453834014e-06,
+      "loss": 1.0605,
+      "step": 3326
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66918955432506e-06,
+      "loss": 1.0799,
+      "step": 3327
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66894056435162e-06,
+      "loss": 1.0359,
+      "step": 3328
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668691483918512e-06,
+      "loss": 1.0831,
+      "step": 3329
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668442313030567e-06,
+      "loss": 0.937,
+      "step": 3330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66819305169261e-06,
+      "loss": 0.8209,
+      "step": 3331
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667943699909474e-06,
+      "loss": 0.8105,
+      "step": 3332
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667694257685987e-06,
+      "loss": 1.0337,
+      "step": 3333
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667444725026983e-06,
+      "loss": 1.0929,
+      "step": 3334
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667195101937294e-06,
+      "loss": 0.9507,
+      "step": 3335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666945388421761e-06,
+      "loss": 0.9427,
+      "step": 3336
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666695584485219e-06,
+      "loss": 0.9329,
+      "step": 3337
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666445690132509e-06,
+      "loss": 1.0077,
+      "step": 3338
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666195705368471e-06,
+      "loss": 0.8954,
+      "step": 3339
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66594563019795e-06,
+      "loss": 0.9632,
+      "step": 3340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665695464625792e-06,
+      "loss": 0.8806,
+      "step": 3341
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665445208656843e-06,
+      "loss": 0.9504,
+      "step": 3342
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66519486229595e-06,
+      "loss": 0.9813,
+      "step": 3343
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664944425547964e-06,
+      "loss": 1.0402,
+      "step": 3344
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66469389841774e-06,
+      "loss": 1.0026,
+      "step": 3345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664443280910128e-06,
+      "loss": 1.1826,
+      "step": 3346
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664192573029987e-06,
+      "loss": 0.9676,
+      "step": 3347
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663941774782173e-06,
+      "loss": 1.0299,
+      "step": 3348
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663690886171544e-06,
+      "loss": 0.8072,
+      "step": 3349
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663439907202962e-06,
+      "loss": 1.0644,
+      "step": 3350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66318883788129e-06,
+      "loss": 1.1214,
+      "step": 3351
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662937678211393e-06,
+      "loss": 0.8977,
+      "step": 3352
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662686428198135e-06,
+      "loss": 0.9813,
+      "step": 3353
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662435087846385e-06,
+      "loss": 0.9937,
+      "step": 3354
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662183657161012e-06,
+      "loss": 1.0982,
+      "step": 3355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661932136146891e-06,
+      "loss": 0.9874,
+      "step": 3356
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661680524808891e-06,
+      "loss": 1.2998,
+      "step": 3357
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661428823151889e-06,
+      "loss": 0.8786,
+      "step": 3358
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66117703118076e-06,
+      "loss": 0.8986,
+      "step": 3359
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660925148900382e-06,
+      "loss": 0.9756,
+      "step": 3360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660673176315637e-06,
+      "loss": 0.7007,
+      "step": 3361
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660421113431409e-06,
+      "loss": 0.9758,
+      "step": 3362
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660168960252575e-06,
+      "loss": 0.9452,
+      "step": 3363
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.65991671678403e-06,
+      "loss": 0.8905,
+      "step": 3364
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.659664383030652e-06,
+      "loss": 0.9349,
+      "step": 3365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.659411958997333e-06,
+      "loss": 1.1138,
+      "step": 3366
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.659159444688965e-06,
+      "loss": 0.9978,
+      "step": 3367
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.65890684011044e-06,
+      "loss": 0.9893,
+      "step": 3368
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658654145266652e-06,
+      "loss": 0.9283,
+      "step": 3369
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658401360162497e-06,
+      "loss": 1.065,
+      "step": 3370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658148484802871e-06,
+      "loss": 0.8378,
+      "step": 3371
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.657895519192678e-06,
+      "loss": 1.2497,
+      "step": 3372
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.657642463336811e-06,
+      "loss": 1.0457,
+      "step": 3373
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.657389317240181e-06,
+      "loss": 1.025,
+      "step": 3374
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65713608090769e-06,
+      "loss": 1.0927,
+      "step": 3375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65688275434424e-06,
+      "loss": 1.0057,
+      "step": 3376
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.656629337554746e-06,
+      "loss": 0.8704,
+      "step": 3377
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.656375830544116e-06,
+      "loss": 1.121,
+      "step": 3378
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.656122233317259e-06,
+      "loss": 0.9293,
+      "step": 3379
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65586854587909e-06,
+      "loss": 0.9978,
+      "step": 3380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.655614768234524e-06,
+      "loss": 0.912,
+      "step": 3381
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.655360900388477e-06,
+      "loss": 1.126,
+      "step": 3382
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65510694234587e-06,
+      "loss": 1.0137,
+      "step": 3383
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654852894111621e-06,
+      "loss": 1.106,
+      "step": 3384
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654598755690654e-06,
+      "loss": 0.9546,
+      "step": 3385
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654344527087892e-06,
+      "loss": 0.9862,
+      "step": 3386
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65409020830826e-06,
+      "loss": 1.0114,
+      "step": 3387
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653835799356686e-06,
+      "loss": 1.1741,
+      "step": 3388
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653581300238097e-06,
+      "loss": 0.9805,
+      "step": 3389
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653326710957429e-06,
+      "loss": 1.1527,
+      "step": 3390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65307203151961e-06,
+      "loss": 1.1146,
+      "step": 3391
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652817261929577e-06,
+      "loss": 0.9007,
+      "step": 3392
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652562402192264e-06,
+      "loss": 1.0376,
+      "step": 3393
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65230745231261e-06,
+      "loss": 0.8973,
+      "step": 3394
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652052412295554e-06,
+      "loss": 0.9161,
+      "step": 3395
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65179728214604e-06,
+      "loss": 1.2511,
+      "step": 3396
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.651542061869007e-06,
+      "loss": 0.7441,
+      "step": 3397
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.651286751469401e-06,
+      "loss": 1.0639,
+      "step": 3398
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65103135095217e-06,
+      "loss": 1.1692,
+      "step": 3399
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650775860322263e-06,
+      "loss": 1.1258,
+      "step": 3400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650520279584627e-06,
+      "loss": 0.8108,
+      "step": 3401
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650264608744215e-06,
+      "loss": 0.9837,
+      "step": 3402
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650008847805983e-06,
+      "loss": 1.0267,
+      "step": 3403
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649752996774883e-06,
+      "loss": 1.1218,
+      "step": 3404
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649497055655873e-06,
+      "loss": 1.11,
+      "step": 3405
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649241024453914e-06,
+      "loss": 1.1078,
+      "step": 3406
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.648984903173964e-06,
+      "loss": 0.9577,
+      "step": 3407
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.648728691820986e-06,
+      "loss": 1.0571,
+      "step": 3408
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.648472390399943e-06,
+      "loss": 0.9489,
+      "step": 3409
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.648215998915804e-06,
+      "loss": 1.1861,
+      "step": 3410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.647959517373532e-06,
+      "loss": 0.8346,
+      "step": 3411
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.647702945778101e-06,
+      "loss": 1.0867,
+      "step": 3412
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.647446284134479e-06,
+      "loss": 0.9712,
+      "step": 3413
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.647189532447638e-06,
+      "loss": 1.1856,
+      "step": 3414
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646932690722556e-06,
+      "loss": 1.0788,
+      "step": 3415
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646675758964206e-06,
+      "loss": 1.1498,
+      "step": 3416
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646418737177567e-06,
+      "loss": 0.9019,
+      "step": 3417
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646161625367621e-06,
+      "loss": 0.9977,
+      "step": 3418
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645904423539345e-06,
+      "loss": 0.8538,
+      "step": 3419
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645647131697726e-06,
+      "loss": 1.0446,
+      "step": 3420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645389749847747e-06,
+      "loss": 0.949,
+      "step": 3421
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645132277994395e-06,
+      "loss": 1.0121,
+      "step": 3422
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64487471614266e-06,
+      "loss": 0.9897,
+      "step": 3423
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64461706429753e-06,
+      "loss": 0.8671,
+      "step": 3424
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.644359322463998e-06,
+      "loss": 1.0506,
+      "step": 3425
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.644101490647058e-06,
+      "loss": 0.989,
+      "step": 3426
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643843568851705e-06,
+      "loss": 1.1288,
+      "step": 3427
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643585557082937e-06,
+      "loss": 0.9611,
+      "step": 3428
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643327455345751e-06,
+      "loss": 0.9848,
+      "step": 3429
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64306926364515e-06,
+      "loss": 1.0196,
+      "step": 3430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.642810981986134e-06,
+      "loss": 1.0619,
+      "step": 3431
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.642552610373709e-06,
+      "loss": 0.9949,
+      "step": 3432
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64229414881288e-06,
+      "loss": 0.9042,
+      "step": 3433
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.642035597308655e-06,
+      "loss": 0.9457,
+      "step": 3434
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641776955866045e-06,
+      "loss": 0.9634,
+      "step": 3435
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641518224490057e-06,
+      "loss": 1.0077,
+      "step": 3436
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641259403185706e-06,
+      "loss": 1.118,
+      "step": 3437
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641000491958008e-06,
+      "loss": 0.9732,
+      "step": 3438
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640741490811979e-06,
+      "loss": 0.9718,
+      "step": 3439
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640482399752635e-06,
+      "loss": 0.95,
+      "step": 3440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640223218785e-06,
+      "loss": 1.0176,
+      "step": 3441
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.639963947914088e-06,
+      "loss": 0.9993,
+      "step": 3442
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63970458714493e-06,
+      "loss": 0.9858,
+      "step": 3443
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.639445136482549e-06,
+      "loss": 0.9459,
+      "step": 3444
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.639185595931968e-06,
+      "loss": 1.1614,
+      "step": 3445
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63892596549822e-06,
+      "loss": 1.1638,
+      "step": 3446
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638666245186333e-06,
+      "loss": 0.9361,
+      "step": 3447
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638406435001341e-06,
+      "loss": 1.0019,
+      "step": 3448
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638146534948275e-06,
+      "loss": 1.0219,
+      "step": 3449
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637886545032174e-06,
+      "loss": 1.0782,
+      "step": 3450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637626465258071e-06,
+      "loss": 0.8215,
+      "step": 3451
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637366295631008e-06,
+      "loss": 1.0404,
+      "step": 3452
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637106036156026e-06,
+      "loss": 1.1195,
+      "step": 3453
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636845686838165e-06,
+      "loss": 0.8558,
+      "step": 3454
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63658524768247e-06,
+      "loss": 0.9749,
+      "step": 3455
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63632471869399e-06,
+      "loss": 0.8528,
+      "step": 3456
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636064099877769e-06,
+      "loss": 0.839,
+      "step": 3457
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635803391238857e-06,
+      "loss": 1.0125,
+      "step": 3458
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635542592782308e-06,
+      "loss": 1.0141,
+      "step": 3459
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635281704513171e-06,
+      "loss": 0.9268,
+      "step": 3460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6350207264365e-06,
+      "loss": 1.0446,
+      "step": 3461
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.634759658557358e-06,
+      "loss": 1.0977,
+      "step": 3462
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.634498500880797e-06,
+      "loss": 1.0773,
+      "step": 3463
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.634237253411881e-06,
+      "loss": 1.1079,
+      "step": 3464
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.633975916155668e-06,
+      "loss": 0.9412,
+      "step": 3465
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.633714489117223e-06,
+      "loss": 1.1245,
+      "step": 3466
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63345297230161e-06,
+      "loss": 1.0631,
+      "step": 3467
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.633191365713897e-06,
+      "loss": 0.8519,
+      "step": 3468
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.632929669359152e-06,
+      "loss": 0.9874,
+      "step": 3469
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.632667883242445e-06,
+      "loss": 1.0462,
+      "step": 3470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63240600736885e-06,
+      "loss": 0.9502,
+      "step": 3471
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.632144041743441e-06,
+      "loss": 0.9916,
+      "step": 3472
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63188198637129e-06,
+      "loss": 1.0068,
+      "step": 3473
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631619841257477e-06,
+      "loss": 1.0705,
+      "step": 3474
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631357606407079e-06,
+      "loss": 1.1249,
+      "step": 3475
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631095281825178e-06,
+      "loss": 0.9815,
+      "step": 3476
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.630832867516855e-06,
+      "loss": 0.9072,
+      "step": 3477
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6305703634872e-06,
+      "loss": 1.0784,
+      "step": 3478
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.630307769741292e-06,
+      "loss": 1.0652,
+      "step": 3479
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.630045086284221e-06,
+      "loss": 0.868,
+      "step": 3480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.629782313121077e-06,
+      "loss": 0.8949,
+      "step": 3481
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62951945025695e-06,
+      "loss": 0.7712,
+      "step": 3482
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.629256497696935e-06,
+      "loss": 1.0139,
+      "step": 3483
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628993455446126e-06,
+      "loss": 0.9353,
+      "step": 3484
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628730323509618e-06,
+      "loss": 1.023,
+      "step": 3485
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62846710189251e-06,
+      "loss": 1.0011,
+      "step": 3486
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628203790599901e-06,
+      "loss": 0.9716,
+      "step": 3487
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627940389636895e-06,
+      "loss": 0.8595,
+      "step": 3488
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627676899008592e-06,
+      "loss": 0.9941,
+      "step": 3489
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6274133187201e-06,
+      "loss": 0.9384,
+      "step": 3490
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627149648776524e-06,
+      "loss": 1.1717,
+      "step": 3491
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.626885889182974e-06,
+      "loss": 1.0975,
+      "step": 3492
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62662203994456e-06,
+      "loss": 1.0647,
+      "step": 3493
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62635810106639e-06,
+      "loss": 1.0641,
+      "step": 3494
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.626094072553584e-06,
+      "loss": 0.9272,
+      "step": 3495
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.625829954411256e-06,
+      "loss": 0.9848,
+      "step": 3496
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62556574664452e-06,
+      "loss": 1.0616,
+      "step": 3497
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.625301449258498e-06,
+      "loss": 1.0125,
+      "step": 3498
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.625037062258309e-06,
+      "loss": 1.0608,
+      "step": 3499
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624772585649076e-06,
+      "loss": 0.9556,
+      "step": 3500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624508019435922e-06,
+      "loss": 0.8838,
+      "step": 3501
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624243363623974e-06,
+      "loss": 1.0093,
+      "step": 3502
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.623978618218363e-06,
+      "loss": 1.1596,
+      "step": 3503
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.623713783224213e-06,
+      "loss": 1.0218,
+      "step": 3504
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.623448858646658e-06,
+      "loss": 0.8876,
+      "step": 3505
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.623183844490828e-06,
+      "loss": 0.8443,
+      "step": 3506
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622918740761863e-06,
+      "loss": 1.0162,
+      "step": 3507
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622653547464894e-06,
+      "loss": 1.1919,
+      "step": 3508
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622388264605061e-06,
+      "loss": 1.0591,
+      "step": 3509
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622122892187505e-06,
+      "loss": 1.0999,
+      "step": 3510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621857430217366e-06,
+      "loss": 1.0468,
+      "step": 3511
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621591878699787e-06,
+      "loss": 1.0298,
+      "step": 3512
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621326237639915e-06,
+      "loss": 1.026,
+      "step": 3513
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621060507042895e-06,
+      "loss": 0.9254,
+      "step": 3514
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.620794686913876e-06,
+      "loss": 1.114,
+      "step": 3515
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.620528777258007e-06,
+      "loss": 0.9687,
+      "step": 3516
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.620262778080443e-06,
+      "loss": 0.8377,
+      "step": 3517
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.619996689386335e-06,
+      "loss": 0.9827,
+      "step": 3518
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.619730511180839e-06,
+      "loss": 1.1615,
+      "step": 3519
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61946424346911e-06,
+      "loss": 0.9651,
+      "step": 3520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.619197886256312e-06,
+      "loss": 0.9275,
+      "step": 3521
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6189314395476e-06,
+      "loss": 1.2023,
+      "step": 3522
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618664903348143e-06,
+      "loss": 1.0712,
+      "step": 3523
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6183982776631e-06,
+      "loss": 0.8711,
+      "step": 3524
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618131562497636e-06,
+      "loss": 0.9889,
+      "step": 3525
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617864757856922e-06,
+      "loss": 1.1135,
+      "step": 3526
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617597863746125e-06,
+      "loss": 0.9118,
+      "step": 3527
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617330880170417e-06,
+      "loss": 0.9361,
+      "step": 3528
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617063807134972e-06,
+      "loss": 1.1261,
+      "step": 3529
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616796644644961e-06,
+      "loss": 0.7961,
+      "step": 3530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616529392705563e-06,
+      "loss": 0.9459,
+      "step": 3531
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616262051321956e-06,
+      "loss": 0.9427,
+      "step": 3532
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.615994620499318e-06,
+      "loss": 1.0342,
+      "step": 3533
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.615727100242831e-06,
+      "loss": 1.0481,
+      "step": 3534
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61545949055768e-06,
+      "loss": 1.0424,
+      "step": 3535
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.615191791449047e-06,
+      "loss": 1.0142,
+      "step": 3536
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614924002922122e-06,
+      "loss": 1.0045,
+      "step": 3537
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614656124982089e-06,
+      "loss": 0.9609,
+      "step": 3538
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614388157634142e-06,
+      "loss": 0.9982,
+      "step": 3539
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61412010088347e-06,
+      "loss": 0.6884,
+      "step": 3540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613851954735267e-06,
+      "loss": 1.0416,
+      "step": 3541
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61358371919473e-06,
+      "loss": 0.9621,
+      "step": 3542
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613315394267055e-06,
+      "loss": 0.9284,
+      "step": 3543
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613046979957442e-06,
+      "loss": 0.9201,
+      "step": 3544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.612778476271087e-06,
+      "loss": 1.0911,
+      "step": 3545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.612509883213198e-06,
+      "loss": 1.0457,
+      "step": 3546
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.612241200788975e-06,
+      "loss": 0.8445,
+      "step": 3547
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611972429003626e-06,
+      "loss": 0.9246,
+      "step": 3548
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611703567862356e-06,
+      "loss": 0.9786,
+      "step": 3549
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611434617370376e-06,
+      "loss": 0.9554,
+      "step": 3550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611165577532898e-06,
+      "loss": 1.0503,
+      "step": 3551
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61089644835513e-06,
+      "loss": 0.9588,
+      "step": 3552
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610627229842292e-06,
+      "loss": 0.9171,
+      "step": 3553
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610357921999596e-06,
+      "loss": 0.9795,
+      "step": 3554
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610088524832262e-06,
+      "loss": 1.041,
+      "step": 3555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.609819038345508e-06,
+      "loss": 1.0727,
+      "step": 3556
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.609549462544554e-06,
+      "loss": 1.0318,
+      "step": 3557
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.609279797434627e-06,
+      "loss": 1.0228,
+      "step": 3558
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.609010043020949e-06,
+      "loss": 1.0594,
+      "step": 3559
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608740199308746e-06,
+      "loss": 1.0155,
+      "step": 3560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608470266303248e-06,
+      "loss": 1.0588,
+      "step": 3561
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608200244009684e-06,
+      "loss": 1.0926,
+      "step": 3562
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.607930132433285e-06,
+      "loss": 1.167,
+      "step": 3563
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.607659931579286e-06,
+      "loss": 0.8398,
+      "step": 3564
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60738964145292e-06,
+      "loss": 0.9843,
+      "step": 3565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.607119262059426e-06,
+      "loss": 1.0493,
+      "step": 3566
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60684879340404e-06,
+      "loss": 0.9536,
+      "step": 3567
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.606578235492004e-06,
+      "loss": 0.9739,
+      "step": 3568
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.606307588328559e-06,
+      "loss": 1.0459,
+      "step": 3569
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60603685191895e-06,
+      "loss": 1.0679,
+      "step": 3570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.605766026268422e-06,
+      "loss": 1.1718,
+      "step": 3571
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.605495111382221e-06,
+      "loss": 0.8813,
+      "step": 3572
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6052241072656e-06,
+      "loss": 0.9089,
+      "step": 3573
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604953013923804e-06,
+      "loss": 1.1398,
+      "step": 3574
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604681831362088e-06,
+      "loss": 1.0062,
+      "step": 3575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604410559585706e-06,
+      "loss": 0.9577,
+      "step": 3576
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604139198599914e-06,
+      "loss": 1.0256,
+      "step": 3577
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60386774840997e-06,
+      "loss": 1.1842,
+      "step": 3578
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.603596209021133e-06,
+      "loss": 1.0085,
+      "step": 3579
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.603324580438662e-06,
+      "loss": 0.9971,
+      "step": 3580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.603052862667824e-06,
+      "loss": 0.9635,
+      "step": 3581
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.602781055713879e-06,
+      "loss": 1.0547,
+      "step": 3582
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.602509159582094e-06,
+      "loss": 1.0952,
+      "step": 3583
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.602237174277741e-06,
+      "loss": 1.0025,
+      "step": 3584
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.601965099806085e-06,
+      "loss": 1.1078,
+      "step": 3585
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6016929361724e-06,
+      "loss": 1.0183,
+      "step": 3586
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.601420683381957e-06,
+      "loss": 0.9182,
+      "step": 3587
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.601148341440035e-06,
+      "loss": 0.9894,
+      "step": 3588
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600875910351904e-06,
+      "loss": 0.9833,
+      "step": 3589
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600603390122848e-06,
+      "loss": 0.912,
+      "step": 3590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600330780758142e-06,
+      "loss": 0.9218,
+      "step": 3591
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600058082263072e-06,
+      "loss": 1.0507,
+      "step": 3592
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.599785294642923e-06,
+      "loss": 1.0811,
+      "step": 3593
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.599512417902974e-06,
+      "loss": 0.9121,
+      "step": 3594
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.599239452048517e-06,
+      "loss": 0.6643,
+      "step": 3595
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598966397084839e-06,
+      "loss": 0.9575,
+      "step": 3596
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.59869325301723e-06,
+      "loss": 1.0737,
+      "step": 3597
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598420019850982e-06,
+      "loss": 1.0476,
+      "step": 3598
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598146697591388e-06,
+      "loss": 1.1439,
+      "step": 3599
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597873286243746e-06,
+      "loss": 0.958,
+      "step": 3600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597599785813351e-06,
+      "loss": 0.998,
+      "step": 3601
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597326196305505e-06,
+      "loss": 0.8405,
+      "step": 3602
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597052517725506e-06,
+      "loss": 1.0412,
+      "step": 3603
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.596778750078657e-06,
+      "loss": 0.9916,
+      "step": 3604
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.596504893370264e-06,
+      "loss": 1.0186,
+      "step": 3605
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.59623094760563e-06,
+      "loss": 0.7117,
+      "step": 3606
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.595956912790064e-06,
+      "loss": 1.016,
+      "step": 3607
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.595682788928875e-06,
+      "loss": 1.0197,
+      "step": 3608
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.595408576027375e-06,
+      "loss": 0.9488,
+      "step": 3609
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.595134274090877e-06,
+      "loss": 1.0623,
+      "step": 3610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594859883124692e-06,
+      "loss": 1.1421,
+      "step": 3611
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59458540313414e-06,
+      "loss": 0.9629,
+      "step": 3612
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594310834124539e-06,
+      "loss": 0.9552,
+      "step": 3613
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59403617610121e-06,
+      "loss": 0.9064,
+      "step": 3614
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.593761429069469e-06,
+      "loss": 0.8624,
+      "step": 3615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.593486593034642e-06,
+      "loss": 1.0035,
+      "step": 3616
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.593211668002056e-06,
+      "loss": 0.9768,
+      "step": 3617
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.592936653977035e-06,
+      "loss": 0.8527,
+      "step": 3618
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.592661550964909e-06,
+      "loss": 1.0382,
+      "step": 3619
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.592386358971006e-06,
+      "loss": 0.9816,
+      "step": 3620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59211107800066e-06,
+      "loss": 1.0917,
+      "step": 3621
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.591835708059202e-06,
+      "loss": 1.0359,
+      "step": 3622
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59156024915197e-06,
+      "loss": 1.0639,
+      "step": 3623
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5912847012843e-06,
+      "loss": 1.0444,
+      "step": 3624
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.591009064461531e-06,
+      "loss": 0.9602,
+      "step": 3625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.590733338689002e-06,
+      "loss": 1.064,
+      "step": 3626
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.590457523972055e-06,
+      "loss": 1.0222,
+      "step": 3627
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.590181620316036e-06,
+      "loss": 0.9668,
+      "step": 3628
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.589905627726288e-06,
+      "loss": 0.9022,
+      "step": 3629
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58962954620816e-06,
+      "loss": 1.0159,
+      "step": 3630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.589353375767002e-06,
+      "loss": 0.989,
+      "step": 3631
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.589077116408163e-06,
+      "loss": 1.0789,
+      "step": 3632
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588800768136996e-06,
+      "loss": 0.9015,
+      "step": 3633
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588524330958855e-06,
+      "loss": 0.8538,
+      "step": 3634
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588247804879093e-06,
+      "loss": 1.0333,
+      "step": 3635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587971189903073e-06,
+      "loss": 0.8851,
+      "step": 3636
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587694486036152e-06,
+      "loss": 0.961,
+      "step": 3637
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587417693283691e-06,
+      "loss": 1.1473,
+      "step": 3638
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587140811651054e-06,
+      "loss": 0.9605,
+      "step": 3639
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.586863841143602e-06,
+      "loss": 1.1056,
+      "step": 3640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.586586781766705e-06,
+      "loss": 0.8589,
+      "step": 3641
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58630963352573e-06,
+      "loss": 1.0584,
+      "step": 3642
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.586032396426044e-06,
+      "loss": 0.9065,
+      "step": 3643
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.585755070473022e-06,
+      "loss": 1.0766,
+      "step": 3644
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.585477655672035e-06,
+      "loss": 1.0768,
+      "step": 3645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58520015202846e-06,
+      "loss": 1.1806,
+      "step": 3646
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584922559547672e-06,
+      "loss": 1.068,
+      "step": 3647
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584644878235047e-06,
+      "loss": 0.9561,
+      "step": 3648
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584367108095969e-06,
+      "loss": 0.8105,
+      "step": 3649
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584089249135819e-06,
+      "loss": 1.0377,
+      "step": 3650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.583811301359979e-06,
+      "loss": 0.9338,
+      "step": 3651
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.583533264773835e-06,
+      "loss": 1.1863,
+      "step": 3652
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.583255139382774e-06,
+      "loss": 0.9604,
+      "step": 3653
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.582976925192183e-06,
+      "loss": 1.0481,
+      "step": 3654
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.582698622207455e-06,
+      "loss": 0.9855,
+      "step": 3655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58242023043398e-06,
+      "loss": 0.9724,
+      "step": 3656
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.582141749877153e-06,
+      "loss": 1.0722,
+      "step": 3657
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581863180542368e-06,
+      "loss": 0.9305,
+      "step": 3658
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581584522435025e-06,
+      "loss": 0.9658,
+      "step": 3659
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58130577556052e-06,
+      "loss": 1.0238,
+      "step": 3660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581026939924253e-06,
+      "loss": 1.1121,
+      "step": 3661
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58074801553163e-06,
+      "loss": 0.9613,
+      "step": 3662
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.580469002388055e-06,
+      "loss": 1.0175,
+      "step": 3663
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58018990049893e-06,
+      "loss": 0.983,
+      "step": 3664
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.579910709869667e-06,
+      "loss": 1.0974,
+      "step": 3665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57963143050567e-06,
+      "loss": 1.0438,
+      "step": 3666
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.579352062412354e-06,
+      "loss": 1.1376,
+      "step": 3667
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57907260559513e-06,
+      "loss": 1.156,
+      "step": 3668
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.578793060059413e-06,
+      "loss": 0.9465,
+      "step": 3669
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57851342581062e-06,
+      "loss": 1.0664,
+      "step": 3670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.578233702854168e-06,
+      "loss": 0.9974,
+      "step": 3671
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577953891195477e-06,
+      "loss": 0.8924,
+      "step": 3672
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577673990839966e-06,
+      "loss": 0.9209,
+      "step": 3673
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577394001793061e-06,
+      "loss": 0.929,
+      "step": 3674
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577113924060185e-06,
+      "loss": 1.0038,
+      "step": 3675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.576833757646765e-06,
+      "loss": 0.8545,
+      "step": 3676
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57655350255823e-06,
+      "loss": 1.0675,
+      "step": 3677
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.576273158800007e-06,
+      "loss": 1.0953,
+      "step": 3678
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575992726377531e-06,
+      "loss": 1.0634,
+      "step": 3679
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575712205296234e-06,
+      "loss": 0.9226,
+      "step": 3680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575431595561552e-06,
+      "loss": 0.9921,
+      "step": 3681
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57515089717892e-06,
+      "loss": 0.9206,
+      "step": 3682
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.574870110153775e-06,
+      "loss": 0.9095,
+      "step": 3683
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.574589234491562e-06,
+      "loss": 0.8429,
+      "step": 3684
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57430827019772e-06,
+      "loss": 1.0412,
+      "step": 3685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.574027217277692e-06,
+      "loss": 0.9846,
+      "step": 3686
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.573746075736924e-06,
+      "loss": 1.0188,
+      "step": 3687
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.573464845580864e-06,
+      "loss": 0.8828,
+      "step": 3688
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.573183526814959e-06,
+      "loss": 0.8926,
+      "step": 3689
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572902119444663e-06,
+      "loss": 0.817,
+      "step": 3690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572620623475423e-06,
+      "loss": 0.9394,
+      "step": 3691
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572339038912697e-06,
+      "loss": 0.9355,
+      "step": 3692
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57205736576194e-06,
+      "loss": 0.9838,
+      "step": 3693
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571775604028607e-06,
+      "loss": 1.078,
+      "step": 3694
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571493753718159e-06,
+      "loss": 1.0876,
+      "step": 3695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571211814836059e-06,
+      "loss": 0.861,
+      "step": 3696
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570929787387763e-06,
+      "loss": 1.0582,
+      "step": 3697
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570647671378742e-06,
+      "loss": 0.9973,
+      "step": 3698
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570365466814458e-06,
+      "loss": 1.1891,
+      "step": 3699
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570083173700381e-06,
+      "loss": 1.1446,
+      "step": 3700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.569800792041979e-06,
+      "loss": 0.9411,
+      "step": 3701
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.569518321844722e-06,
+      "loss": 0.9475,
+      "step": 3702
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.569235763114084e-06,
+      "loss": 1.0275,
+      "step": 3703
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.568953115855541e-06,
+      "loss": 1.0529,
+      "step": 3704
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.568670380074568e-06,
+      "loss": 1.0229,
+      "step": 3705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56838755577664e-06,
+      "loss": 1.0541,
+      "step": 3706
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.568104642967242e-06,
+      "loss": 0.9491,
+      "step": 3707
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.567821641651851e-06,
+      "loss": 0.9859,
+      "step": 3708
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.567538551835955e-06,
+      "loss": 0.9528,
+      "step": 3709
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.567255373525032e-06,
+      "loss": 0.8124,
+      "step": 3710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566972106724574e-06,
+      "loss": 1.0466,
+      "step": 3711
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566688751440068e-06,
+      "loss": 1.0274,
+      "step": 3712
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566405307677002e-06,
+      "loss": 1.0147,
+      "step": 3713
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56612177544087e-06,
+      "loss": 0.8505,
+      "step": 3714
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.565838154737163e-06,
+      "loss": 0.8905,
+      "step": 3715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.565554445571378e-06,
+      "loss": 0.9226,
+      "step": 3716
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.565270647949013e-06,
+      "loss": 0.8,
+      "step": 3717
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.564986761875562e-06,
+      "loss": 1.091,
+      "step": 3718
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.564702787356529e-06,
+      "loss": 0.9217,
+      "step": 3719
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.564418724397415e-06,
+      "loss": 1.071,
+      "step": 3720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.564134573003723e-06,
+      "loss": 1.1179,
+      "step": 3721
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56385033318096e-06,
+      "loss": 0.9076,
+      "step": 3722
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.563566004934632e-06,
+      "loss": 0.9166,
+      "step": 3723
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.563281588270249e-06,
+      "loss": 0.982,
+      "step": 3724
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.562997083193318e-06,
+      "loss": 1.0585,
+      "step": 3725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.562712489709354e-06,
+      "loss": 0.92,
+      "step": 3726
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.562427807823872e-06,
+      "loss": 1.1295,
+      "step": 3727
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.562143037542384e-06,
+      "loss": 0.9764,
+      "step": 3728
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561858178870411e-06,
+      "loss": 1.1412,
+      "step": 3729
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561573231813471e-06,
+      "loss": 0.9937,
+      "step": 3730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561288196377082e-06,
+      "loss": 0.8899,
+      "step": 3731
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561003072566773e-06,
+      "loss": 1.0554,
+      "step": 3732
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.560717860388061e-06,
+      "loss": 0.8942,
+      "step": 3733
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.560432559846476e-06,
+      "loss": 0.8818,
+      "step": 3734
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.560147170947544e-06,
+      "loss": 1.0407,
+      "step": 3735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559861693696795e-06,
+      "loss": 0.9743,
+      "step": 3736
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55957612809976e-06,
+      "loss": 0.7852,
+      "step": 3737
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559290474161972e-06,
+      "loss": 1.0007,
+      "step": 3738
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559004731888966e-06,
+      "loss": 0.8769,
+      "step": 3739
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.558718901286276e-06,
+      "loss": 0.8158,
+      "step": 3740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.558432982359443e-06,
+      "loss": 0.8726,
+      "step": 3741
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.558146975114004e-06,
+      "loss": 1.0546,
+      "step": 3742
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.557860879555502e-06,
+      "loss": 0.9749,
+      "step": 3743
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.557574695689478e-06,
+      "loss": 0.8846,
+      "step": 3744
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.557288423521481e-06,
+      "loss": 0.9973,
+      "step": 3745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55700206305705e-06,
+      "loss": 0.9114,
+      "step": 3746
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.556715614301742e-06,
+      "loss": 1.0648,
+      "step": 3747
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5564290772611e-06,
+      "loss": 0.9722,
+      "step": 3748
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55614245194068e-06,
+      "loss": 1.1794,
+      "step": 3749
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.555855738346031e-06,
+      "loss": 1.1727,
+      "step": 3750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.555568936482713e-06,
+      "loss": 0.8733,
+      "step": 3751
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55528204635628e-06,
+      "loss": 1.1688,
+      "step": 3752
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554995067972289e-06,
+      "loss": 0.9421,
+      "step": 3753
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554708001336303e-06,
+      "loss": 1.1168,
+      "step": 3754
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554420846453883e-06,
+      "loss": 1.0405,
+      "step": 3755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554133603330592e-06,
+      "loss": 1.0573,
+      "step": 3756
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.553846271971994e-06,
+      "loss": 0.966,
+      "step": 3757
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.553558852383659e-06,
+      "loss": 1.0673,
+      "step": 3758
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.553271344571153e-06,
+      "loss": 1.0318,
+      "step": 3759
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55298374854005e-06,
+      "loss": 1.0216,
+      "step": 3760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.552696064295916e-06,
+      "loss": 0.8755,
+      "step": 3761
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.552408291844332e-06,
+      "loss": 0.9826,
+      "step": 3762
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.552120431190868e-06,
+      "loss": 1.0663,
+      "step": 3763
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551832482341104e-06,
+      "loss": 1.0737,
+      "step": 3764
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551544445300618e-06,
+      "loss": 1.0233,
+      "step": 3765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551256320074992e-06,
+      "loss": 1.0177,
+      "step": 3766
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.550968106669805e-06,
+      "loss": 0.9155,
+      "step": 3767
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.550679805090643e-06,
+      "loss": 0.9885,
+      "step": 3768
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.550391415343095e-06,
+      "loss": 0.8834,
+      "step": 3769
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.550102937432743e-06,
+      "loss": 0.9484,
+      "step": 3770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549814371365178e-06,
+      "loss": 0.9445,
+      "step": 3771
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549525717145994e-06,
+      "loss": 0.9883,
+      "step": 3772
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54923697478078e-06,
+      "loss": 0.9319,
+      "step": 3773
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.548948144275133e-06,
+      "loss": 1.1312,
+      "step": 3774
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.548659225634646e-06,
+      "loss": 0.9725,
+      "step": 3775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.548370218864919e-06,
+      "loss": 1.0193,
+      "step": 3776
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54808112397155e-06,
+      "loss": 0.9146,
+      "step": 3777
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547791940960142e-06,
+      "loss": 0.998,
+      "step": 3778
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547502669836294e-06,
+      "loss": 0.922,
+      "step": 3779
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547213310605618e-06,
+      "loss": 1.1067,
+      "step": 3780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.546923863273711e-06,
+      "loss": 1.0337,
+      "step": 3781
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54663432784619e-06,
+      "loss": 1.0865,
+      "step": 3782
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.546344704328655e-06,
+      "loss": 1.0092,
+      "step": 3783
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.546054992726728e-06,
+      "loss": 1.1553,
+      "step": 3784
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.545765193046014e-06,
+      "loss": 0.8747,
+      "step": 3785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.545475305292132e-06,
+      "loss": 1.0409,
+      "step": 3786
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.545185329470694e-06,
+      "loss": 0.8482,
+      "step": 3787
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544895265587323e-06,
+      "loss": 1.0999,
+      "step": 3788
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544605113647637e-06,
+      "loss": 1.0127,
+      "step": 3789
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544314873657259e-06,
+      "loss": 1.155,
+      "step": 3790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54402454562181e-06,
+      "loss": 0.9824,
+      "step": 3791
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.543734129546916e-06,
+      "loss": 0.7173,
+      "step": 3792
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.543443625438205e-06,
+      "loss": 1.0365,
+      "step": 3793
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.543153033301304e-06,
+      "loss": 0.7895,
+      "step": 3794
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542862353141843e-06,
+      "loss": 1.0054,
+      "step": 3795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542571584965454e-06,
+      "loss": 1.1342,
+      "step": 3796
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542280728777773e-06,
+      "loss": 1.0322,
+      "step": 3797
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541989784584431e-06,
+      "loss": 1.0314,
+      "step": 3798
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54169875239107e-06,
+      "loss": 0.9258,
+      "step": 3799
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541407632203324e-06,
+      "loss": 0.665,
+      "step": 3800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541116424026834e-06,
+      "loss": 0.8896,
+      "step": 3801
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.540825127867245e-06,
+      "loss": 0.9232,
+      "step": 3802
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.540533743730198e-06,
+      "loss": 0.9211,
+      "step": 3803
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54024227162134e-06,
+      "loss": 1.0602,
+      "step": 3804
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539950711546319e-06,
+      "loss": 0.9019,
+      "step": 3805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539659063510782e-06,
+      "loss": 0.9974,
+      "step": 3806
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539367327520382e-06,
+      "loss": 1.2168,
+      "step": 3807
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539075503580769e-06,
+      "loss": 0.8817,
+      "step": 3808
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.538783591697596e-06,
+      "loss": 1.082,
+      "step": 3809
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.538491591876522e-06,
+      "loss": 0.993,
+      "step": 3810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.538199504123202e-06,
+      "loss": 0.8281,
+      "step": 3811
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537907328443296e-06,
+      "loss": 0.9111,
+      "step": 3812
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537615064842465e-06,
+      "loss": 1.0435,
+      "step": 3813
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537322713326374e-06,
+      "loss": 0.8724,
+      "step": 3814
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537030273900681e-06,
+      "loss": 1.0232,
+      "step": 3815
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.536737746571058e-06,
+      "loss": 0.9954,
+      "step": 3816
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.536445131343169e-06,
+      "loss": 0.9782,
+      "step": 3817
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.536152428222684e-06,
+      "loss": 0.9777,
+      "step": 3818
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.535859637215277e-06,
+      "loss": 0.9446,
+      "step": 3819
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.535566758326618e-06,
+      "loss": 0.9581,
+      "step": 3820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53527379156238e-06,
+      "loss": 0.9565,
+      "step": 3821
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.534980736928242e-06,
+      "loss": 0.8752,
+      "step": 3822
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53468759442988e-06,
+      "loss": 1.0542,
+      "step": 3823
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.534394364072977e-06,
+      "loss": 0.9009,
+      "step": 3824
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53410104586321e-06,
+      "loss": 0.9887,
+      "step": 3825
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.533807639806265e-06,
+      "loss": 1.1055,
+      "step": 3826
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.533514145907823e-06,
+      "loss": 0.8367,
+      "step": 3827
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.533220564173576e-06,
+      "loss": 0.8674,
+      "step": 3828
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532926894609205e-06,
+      "loss": 0.9736,
+      "step": 3829
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532633137220406e-06,
+      "loss": 0.835,
+      "step": 3830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532339292012869e-06,
+      "loss": 0.8588,
+      "step": 3831
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532045358992286e-06,
+      "loss": 1.0025,
+      "step": 3832
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53175133816435e-06,
+      "loss": 1.0261,
+      "step": 3833
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.531457229534762e-06,
+      "loss": 0.9739,
+      "step": 3834
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.531163033109217e-06,
+      "loss": 1.0375,
+      "step": 3835
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.530868748893416e-06,
+      "loss": 0.9798,
+      "step": 3836
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.530574376893062e-06,
+      "loss": 1.1943,
+      "step": 3837
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.530279917113856e-06,
+      "loss": 1.0093,
+      "step": 3838
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.529985369561506e-06,
+      "loss": 0.8828,
+      "step": 3839
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.529690734241714e-06,
+      "loss": 1.0425,
+      "step": 3840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.529396011160194e-06,
+      "loss": 0.8994,
+      "step": 3841
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.529101200322653e-06,
+      "loss": 0.9465,
+      "step": 3842
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528806301734804e-06,
+      "loss": 0.9745,
+      "step": 3843
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528511315402358e-06,
+      "loss": 0.9962,
+      "step": 3844
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528216241331034e-06,
+      "loss": 1.0882,
+      "step": 3845
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527921079526548e-06,
+      "loss": 0.9037,
+      "step": 3846
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527625829994617e-06,
+      "loss": 0.7629,
+      "step": 3847
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527330492740964e-06,
+      "loss": 1.0217,
+      "step": 3848
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527035067771309e-06,
+      "loss": 0.8884,
+      "step": 3849
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526739555091378e-06,
+      "loss": 0.9705,
+      "step": 3850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526443954706891e-06,
+      "loss": 1.1082,
+      "step": 3851
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526148266623583e-06,
+      "loss": 1.0152,
+      "step": 3852
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.525852490847179e-06,
+      "loss": 0.9261,
+      "step": 3853
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.525556627383408e-06,
+      "loss": 0.9855,
+      "step": 3854
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.525260676238004e-06,
+      "loss": 0.8825,
+      "step": 3855
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.5249646374167e-06,
+      "loss": 0.9894,
+      "step": 3856
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524668510925235e-06,
+      "loss": 1.0749,
+      "step": 3857
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524372296769342e-06,
+      "loss": 1.2053,
+      "step": 3858
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524075994954762e-06,
+      "loss": 1.0281,
+      "step": 3859
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.523779605487238e-06,
+      "loss": 0.9512,
+      "step": 3860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.523483128372508e-06,
+      "loss": 0.9732,
+      "step": 3861
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.52318656361632e-06,
+      "loss": 0.8322,
+      "step": 3862
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522889911224417e-06,
+      "loss": 0.9464,
+      "step": 3863
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.52259317120255e-06,
+      "loss": 0.9572,
+      "step": 3864
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522296343556464e-06,
+      "loss": 0.886,
+      "step": 3865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.521999428291913e-06,
+      "loss": 1.1353,
+      "step": 3866
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.52170242541465e-06,
+      "loss": 1.0967,
+      "step": 3867
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.521405334930428e-06,
+      "loss": 0.9946,
+      "step": 3868
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.521108156845e-06,
+      "loss": 0.9085,
+      "step": 3869
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.520810891164129e-06,
+      "loss": 1.0243,
+      "step": 3870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.520513537893574e-06,
+      "loss": 1.0139,
+      "step": 3871
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.520216097039094e-06,
+      "loss": 0.9029,
+      "step": 3872
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.519918568606452e-06,
+      "loss": 1.1122,
+      "step": 3873
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.519620952601412e-06,
+      "loss": 0.9825,
+      "step": 3874
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.519323249029743e-06,
+      "loss": 0.8314,
+      "step": 3875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51902545789721e-06,
+      "loss": 1.0629,
+      "step": 3876
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.518727579209584e-06,
+      "loss": 1.0637,
+      "step": 3877
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.518429612972637e-06,
+      "loss": 0.7662,
+      "step": 3878
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51813155919214e-06,
+      "loss": 1.1284,
+      "step": 3879
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51783341787387e-06,
+      "loss": 0.838,
+      "step": 3880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.517535189023602e-06,
+      "loss": 1.1566,
+      "step": 3881
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.517236872647115e-06,
+      "loss": 0.9824,
+      "step": 3882
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516938468750187e-06,
+      "loss": 0.8613,
+      "step": 3883
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516639977338602e-06,
+      "loss": 0.8358,
+      "step": 3884
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516341398418142e-06,
+      "loss": 0.8829,
+      "step": 3885
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51604273199459e-06,
+      "loss": 0.7862,
+      "step": 3886
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.515743978073737e-06,
+      "loss": 0.9647,
+      "step": 3887
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.515445136661367e-06,
+      "loss": 0.8794,
+      "step": 3888
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.515146207763273e-06,
+      "loss": 0.9664,
+      "step": 3889
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514847191385244e-06,
+      "loss": 0.9197,
+      "step": 3890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514548087533076e-06,
+      "loss": 0.9252,
+      "step": 3891
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514248896212562e-06,
+      "loss": 0.9386,
+      "step": 3892
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.5139496174295e-06,
+      "loss": 1.0213,
+      "step": 3893
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.513650251189687e-06,
+      "loss": 0.9854,
+      "step": 3894
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.513350797498926e-06,
+      "loss": 0.8601,
+      "step": 3895
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.513051256363017e-06,
+      "loss": 0.8217,
+      "step": 3896
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.512751627787763e-06,
+      "loss": 0.9509,
+      "step": 3897
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51245191177897e-06,
+      "loss": 1.0673,
+      "step": 3898
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.512152108342445e-06,
+      "loss": 1.0292,
+      "step": 3899
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.511852217483996e-06,
+      "loss": 1.0579,
+      "step": 3900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.511552239209434e-06,
+      "loss": 0.9357,
+      "step": 3901
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51125217352457e-06,
+      "loss": 1.0394,
+      "step": 3902
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51095202043522e-06,
+      "loss": 0.9486,
+      "step": 3903
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.510651779947197e-06,
+      "loss": 1.0603,
+      "step": 3904
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51035145206632e-06,
+      "loss": 1.0913,
+      "step": 3905
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.510051036798404e-06,
+      "loss": 0.8974,
+      "step": 3906
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509750534149275e-06,
+      "loss": 0.9388,
+      "step": 3907
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509449944124751e-06,
+      "loss": 0.9574,
+      "step": 3908
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509149266730657e-06,
+      "loss": 0.8826,
+      "step": 3909
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50884850197282e-06,
+      "loss": 1.0024,
+      "step": 3910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.508547649857064e-06,
+      "loss": 0.8777,
+      "step": 3911
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.508246710389224e-06,
+      "loss": 0.955,
+      "step": 3912
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507945683575123e-06,
+      "loss": 0.8823,
+      "step": 3913
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507644569420598e-06,
+      "loss": 0.9563,
+      "step": 3914
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507343367931482e-06,
+      "loss": 1.0315,
+      "step": 3915
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507042079113611e-06,
+      "loss": 1.0208,
+      "step": 3916
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50674070297282e-06,
+      "loss": 1.0391,
+      "step": 3917
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.506439239514954e-06,
+      "loss": 1.0397,
+      "step": 3918
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.506137688745847e-06,
+      "loss": 0.822,
+      "step": 3919
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.505836050671347e-06,
+      "loss": 0.9753,
+      "step": 3920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.505534325297291e-06,
+      "loss": 0.9701,
+      "step": 3921
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.505232512629535e-06,
+      "loss": 0.9072,
+      "step": 3922
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504930612673918e-06,
+      "loss": 1.0431,
+      "step": 3923
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504628625436291e-06,
+      "loss": 1.1009,
+      "step": 3924
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504326550922508e-06,
+      "loss": 0.7595,
+      "step": 3925
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504024389138422e-06,
+      "loss": 1.0778,
+      "step": 3926
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503722140089882e-06,
+      "loss": 0.9744,
+      "step": 3927
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503419803782749e-06,
+      "loss": 0.8866,
+      "step": 3928
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503117380222877e-06,
+      "loss": 0.9127,
+      "step": 3929
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.502814869416128e-06,
+      "loss": 1.0327,
+      "step": 3930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.502512271368362e-06,
+      "loss": 0.9619,
+      "step": 3931
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.502209586085444e-06,
+      "loss": 0.915,
+      "step": 3932
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.501906813573234e-06,
+      "loss": 0.9356,
+      "step": 3933
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.501603953837602e-06,
+      "loss": 0.7801,
+      "step": 3934
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.501301006884414e-06,
+      "loss": 0.7953,
+      "step": 3935
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.500997972719541e-06,
+      "loss": 0.7971,
+      "step": 3936
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.500694851348853e-06,
+      "loss": 1.0042,
+      "step": 3937
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.500391642778221e-06,
+      "loss": 0.9773,
+      "step": 3938
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.500088347013525e-06,
+      "loss": 1.0237,
+      "step": 3939
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.499784964060637e-06,
+      "loss": 0.8257,
+      "step": 3940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.499481493925436e-06,
+      "loss": 0.837,
+      "step": 3941
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.4991779366138e-06,
+      "loss": 0.9645,
+      "step": 3942
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.498874292131615e-06,
+      "loss": 0.8891,
+      "step": 3943
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49857056048476e-06,
+      "loss": 0.9482,
+      "step": 3944
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.498266741679119e-06,
+      "loss": 0.918,
+      "step": 3945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497962835720582e-06,
+      "loss": 1.0745,
+      "step": 3946
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497658842615036e-06,
+      "loss": 1.0082,
+      "step": 3947
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497354762368368e-06,
+      "loss": 1.1411,
+      "step": 3948
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497050594986472e-06,
+      "loss": 1.167,
+      "step": 3949
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.496746340475243e-06,
+      "loss": 0.94,
+      "step": 3950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.496441998840572e-06,
+      "loss": 0.9355,
+      "step": 3951
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.496137570088358e-06,
+      "loss": 1.0097,
+      "step": 3952
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495833054224496e-06,
+      "loss": 1.1562,
+      "step": 3953
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495528451254892e-06,
+      "loss": 1.0443,
+      "step": 3954
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495223761185443e-06,
+      "loss": 1.005,
+      "step": 3955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.494918984022053e-06,
+      "loss": 1.07,
+      "step": 3956
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.494614119770627e-06,
+      "loss": 0.9577,
+      "step": 3957
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.494309168437072e-06,
+      "loss": 0.9378,
+      "step": 3958
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.494004130027296e-06,
+      "loss": 0.8606,
+      "step": 3959
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49369900454721e-06,
+      "loss": 0.8364,
+      "step": 3960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.493393792002725e-06,
+      "loss": 0.9146,
+      "step": 3961
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.493088492399754e-06,
+      "loss": 1.065,
+      "step": 3962
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.492783105744215e-06,
+      "loss": 1.0977,
+      "step": 3963
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.492477632042019e-06,
+      "loss": 1.1681,
+      "step": 3964
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49217207129909e-06,
+      "loss": 0.842,
+      "step": 3965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.491866423521346e-06,
+      "loss": 0.9665,
+      "step": 3966
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.491560688714708e-06,
+      "loss": 1.0642,
+      "step": 3967
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.491254866885102e-06,
+      "loss": 0.9795,
+      "step": 3968
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49094895803845e-06,
+      "loss": 1.0487,
+      "step": 3969
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490642962180681e-06,
+      "loss": 1.0338,
+      "step": 3970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490336879317725e-06,
+      "loss": 0.7785,
+      "step": 3971
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490030709455509e-06,
+      "loss": 0.7931,
+      "step": 3972
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.489724452599967e-06,
+      "loss": 1.1374,
+      "step": 3973
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.489418108757032e-06,
+      "loss": 0.8703,
+      "step": 3974
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.489111677932639e-06,
+      "loss": 0.942,
+      "step": 3975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.488805160132726e-06,
+      "loss": 1.115,
+      "step": 3976
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.488498555363232e-06,
+      "loss": 0.9182,
+      "step": 3977
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.488191863630096e-06,
+      "loss": 0.9577,
+      "step": 3978
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.487885084939261e-06,
+      "loss": 1.1222,
+      "step": 3979
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48757821929667e-06,
+      "loss": 0.9394,
+      "step": 3980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.487271266708268e-06,
+      "loss": 1.0079,
+      "step": 3981
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486964227180005e-06,
+      "loss": 0.9928,
+      "step": 3982
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486657100717827e-06,
+      "loss": 1.0213,
+      "step": 3983
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486349887327685e-06,
+      "loss": 1.0124,
+      "step": 3984
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486042587015534e-06,
+      "loss": 0.9011,
+      "step": 3985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.485735199787322e-06,
+      "loss": 0.949,
+      "step": 3986
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48542772564901e-06,
+      "loss": 0.8097,
+      "step": 3987
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.485120164606553e-06,
+      "loss": 0.9225,
+      "step": 3988
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48481251666591e-06,
+      "loss": 1.1104,
+      "step": 3989
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.484504781833043e-06,
+      "loss": 1.1451,
+      "step": 3990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.484196960113911e-06,
+      "loss": 0.7911,
+      "step": 3991
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.483889051514483e-06,
+      "loss": 1.0406,
+      "step": 3992
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48358105604072e-06,
+      "loss": 0.9028,
+      "step": 3993
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.483272973698591e-06,
+      "loss": 0.8913,
+      "step": 3994
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.482964804494065e-06,
+      "loss": 0.7737,
+      "step": 3995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.482656548433113e-06,
+      "loss": 1.1483,
+      "step": 3996
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.482348205521708e-06,
+      "loss": 1.1143,
+      "step": 3997
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.482039775765821e-06,
+      "loss": 1.1038,
+      "step": 3998
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481731259171432e-06,
+      "loss": 0.8709,
+      "step": 3999
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481422655744516e-06,
+      "loss": 0.8607,
+      "step": 4000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481113965491054e-06,
+      "loss": 0.97,
+      "step": 4001
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.480805188417022e-06,
+      "loss": 0.8368,
+      "step": 4002
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.480496324528408e-06,
+      "loss": 1.1491,
+      "step": 4003
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.480187373831195e-06,
+      "loss": 1.024,
+      "step": 4004
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.479878336331366e-06,
+      "loss": 1.0745,
+      "step": 4005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.479569212034912e-06,
+      "loss": 0.7684,
+      "step": 4006
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47926000094782e-06,
+      "loss": 0.9157,
+      "step": 4007
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.478950703076081e-06,
+      "loss": 1.0993,
+      "step": 4008
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.478641318425689e-06,
+      "loss": 0.9802,
+      "step": 4009
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.478331847002637e-06,
+      "loss": 1.0382,
+      "step": 4010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47802228881292e-06,
+      "loss": 1.0014,
+      "step": 4011
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47771264386254e-06,
+      "loss": 0.9843,
+      "step": 4012
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.477402912157491e-06,
+      "loss": 1.0469,
+      "step": 4013
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.477093093703776e-06,
+      "loss": 1.0588,
+      "step": 4014
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.4767831885074e-06,
+      "loss": 0.9285,
+      "step": 4015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.476473196574362e-06,
+      "loss": 0.9355,
+      "step": 4016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.476163117910674e-06,
+      "loss": 0.9966,
+      "step": 4017
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47585295252234e-06,
+      "loss": 0.9243,
+      "step": 4018
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47554270041537e-06,
+      "loss": 0.9627,
+      "step": 4019
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.475232361595776e-06,
+      "loss": 0.822,
+      "step": 4020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47492193606957e-06,
+      "loss": 0.742,
+      "step": 4021
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.474611423842767e-06,
+      "loss": 0.9433,
+      "step": 4022
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.474300824921383e-06,
+      "loss": 0.8525,
+      "step": 4023
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.473990139311433e-06,
+      "loss": 1.0279,
+      "step": 4024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.473679367018942e-06,
+      "loss": 0.8979,
+      "step": 4025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.473368508049925e-06,
+      "loss": 0.8949,
+      "step": 4026
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47305756241041e-06,
+      "loss": 1.1066,
+      "step": 4027
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47274653010642e-06,
+      "loss": 1.0068,
+      "step": 4028
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.472435411143979e-06,
+      "loss": 1.0278,
+      "step": 4029
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.472124205529117e-06,
+      "loss": 1.3112,
+      "step": 4030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471812913267863e-06,
+      "loss": 1.0318,
+      "step": 4031
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471501534366247e-06,
+      "loss": 0.8096,
+      "step": 4032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471190068830305e-06,
+      "loss": 0.8911,
+      "step": 4033
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.470878516666068e-06,
+      "loss": 0.9571,
+      "step": 4034
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.470566877879574e-06,
+      "loss": 0.8989,
+      "step": 4035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.470255152476862e-06,
+      "loss": 0.8893,
+      "step": 4036
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.469943340463971e-06,
+      "loss": 1.0542,
+      "step": 4037
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46963144184694e-06,
+      "loss": 1.0006,
+      "step": 4038
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.469319456631814e-06,
+      "loss": 0.9556,
+      "step": 4039
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.469007384824638e-06,
+      "loss": 1.0228,
+      "step": 4040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.468695226431457e-06,
+      "loss": 1.0335,
+      "step": 4041
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.468382981458321e-06,
+      "loss": 0.9377,
+      "step": 4042
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.468070649911277e-06,
+      "loss": 1.0305,
+      "step": 4043
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.467758231796378e-06,
+      "loss": 1.0711,
+      "step": 4044
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.467445727119677e-06,
+      "loss": 0.9847,
+      "step": 4045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.467133135887227e-06,
+      "loss": 0.9097,
+      "step": 4046
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.466820458105088e-06,
+      "loss": 1.0446,
+      "step": 4047
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.466507693779314e-06,
+      "loss": 1.0061,
+      "step": 4048
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.466194842915969e-06,
+      "loss": 0.9568,
+      "step": 4049
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.465881905521111e-06,
+      "loss": 1.1664,
+      "step": 4050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.465568881600801e-06,
+      "loss": 1.0875,
+      "step": 4051
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46525577116111e-06,
+      "loss": 1.0333,
+      "step": 4052
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464942574208102e-06,
+      "loss": 1.0776,
+      "step": 4053
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464629290747844e-06,
+      "loss": 0.8863,
+      "step": 4054
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464315920786404e-06,
+      "loss": 1.1398,
+      "step": 4055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464002464329858e-06,
+      "loss": 0.9822,
+      "step": 4056
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463688921384276e-06,
+      "loss": 1.1289,
+      "step": 4057
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463375291955735e-06,
+      "loss": 1.0321,
+      "step": 4058
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463061576050308e-06,
+      "loss": 0.9018,
+      "step": 4059
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462747773674075e-06,
+      "loss": 0.9971,
+      "step": 4060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462433884833119e-06,
+      "loss": 0.9972,
+      "step": 4061
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462119909533517e-06,
+      "loss": 0.9277,
+      "step": 4062
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461805847781353e-06,
+      "loss": 0.9918,
+      "step": 4063
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461491699582715e-06,
+      "loss": 1.038,
+      "step": 4064
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461177464943687e-06,
+      "loss": 0.9471,
+      "step": 4065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460863143870355e-06,
+      "loss": 1.1162,
+      "step": 4066
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460548736368814e-06,
+      "loss": 0.9127,
+      "step": 4067
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460234242445152e-06,
+      "loss": 0.9851,
+      "step": 4068
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459919662105464e-06,
+      "loss": 0.9556,
+      "step": 4069
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459604995355844e-06,
+      "loss": 1.0395,
+      "step": 4070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459290242202389e-06,
+      "loss": 0.9083,
+      "step": 4071
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.458975402651196e-06,
+      "loss": 0.9867,
+      "step": 4072
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.458660476708368e-06,
+      "loss": 0.8609,
+      "step": 4073
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.458345464380003e-06,
+      "loss": 1.0778,
+      "step": 4074
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.458030365672208e-06,
+      "loss": 1.1471,
+      "step": 4075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.457715180591084e-06,
+      "loss": 1.031,
+      "step": 4076
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45739990914274e-06,
+      "loss": 0.9776,
+      "step": 4077
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.457084551333284e-06,
+      "loss": 1.0764,
+      "step": 4078
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.456769107168827e-06,
+      "loss": 0.9007,
+      "step": 4079
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.456453576655477e-06,
+      "loss": 0.8761,
+      "step": 4080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45613795979935e-06,
+      "loss": 1.0969,
+      "step": 4081
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.455822256606563e-06,
+      "loss": 0.9555,
+      "step": 4082
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.455506467083228e-06,
+      "loss": 0.9002,
+      "step": 4083
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.455190591235468e-06,
+      "loss": 0.9532,
+      "step": 4084
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4548746290694e-06,
+      "loss": 0.7666,
+      "step": 4085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.454558580591147e-06,
+      "loss": 1.0425,
+      "step": 4086
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.454242445806832e-06,
+      "loss": 0.9769,
+      "step": 4087
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45392622472258e-06,
+      "loss": 1.0683,
+      "step": 4088
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.453609917344517e-06,
+      "loss": 0.9596,
+      "step": 4089
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.453293523678772e-06,
+      "loss": 1.047,
+      "step": 4090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452977043731477e-06,
+      "loss": 0.8157,
+      "step": 4091
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45266047750876e-06,
+      "loss": 0.9089,
+      "step": 4092
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452343825016757e-06,
+      "loss": 0.9436,
+      "step": 4093
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452027086261602e-06,
+      "loss": 1.1044,
+      "step": 4094
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.451710261249432e-06,
+      "loss": 1.0186,
+      "step": 4095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.451393349986387e-06,
+      "loss": 1.1055,
+      "step": 4096
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.451076352478605e-06,
+      "loss": 0.8687,
+      "step": 4097
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.450759268732226e-06,
+      "loss": 1.0004,
+      "step": 4098
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.450442098753398e-06,
+      "loss": 1.1026,
+      "step": 4099
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.450124842548264e-06,
+      "loss": 0.7289,
+      "step": 4100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.44980750012297e-06,
+      "loss": 1.053,
+      "step": 4101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.449490071483664e-06,
+      "loss": 0.9645,
+      "step": 4102
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4491725566365e-06,
+      "loss": 0.9571,
+      "step": 4103
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.448854955587625e-06,
+      "loss": 0.9329,
+      "step": 4104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.448537268343194e-06,
+      "loss": 0.8676,
+      "step": 4105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.448219494909363e-06,
+      "loss": 0.8795,
+      "step": 4106
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.447901635292289e-06,
+      "loss": 1.0211,
+      "step": 4107
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.44758368949813e-06,
+      "loss": 0.8426,
+      "step": 4108
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.447265657533045e-06,
+      "loss": 1.2483,
+      "step": 4109
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.446947539403198e-06,
+      "loss": 0.8907,
+      "step": 4110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.446629335114748e-06,
+      "loss": 0.9631,
+      "step": 4111
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.446311044673866e-06,
+      "loss": 0.9815,
+      "step": 4112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.445992668086716e-06,
+      "loss": 0.902,
+      "step": 4113
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.445674205359466e-06,
+      "loss": 1.0969,
+      "step": 4114
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.445355656498284e-06,
+      "loss": 1.1551,
+      "step": 4115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.44503702150935e-06,
+      "loss": 1.1135,
+      "step": 4116
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.444718300398829e-06,
+      "loss": 0.8676,
+      "step": 4117
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.444399493172898e-06,
+      "loss": 1.065,
+      "step": 4118
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.444080599837737e-06,
+      "loss": 0.8989,
+      "step": 4119
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443761620399524e-06,
+      "loss": 1.0423,
+      "step": 4120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443442554864434e-06,
+      "loss": 1.038,
+      "step": 4121
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443123403238655e-06,
+      "loss": 1.1184,
+      "step": 4122
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.442804165528366e-06,
+      "loss": 1.0474,
+      "step": 4123
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.442484841739755e-06,
+      "loss": 1.0271,
+      "step": 4124
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.442165431879006e-06,
+      "loss": 0.9507,
+      "step": 4125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.441845935952313e-06,
+      "loss": 1.0965,
+      "step": 4126
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.44152635396586e-06,
+      "loss": 1.019,
+      "step": 4127
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.441206685925843e-06,
+      "loss": 1.063,
+      "step": 4128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.440886931838453e-06,
+      "loss": 1.0367,
+      "step": 4129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.440567091709886e-06,
+      "loss": 1.0924,
+      "step": 4130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.440247165546339e-06,
+      "loss": 1.0525,
+      "step": 4131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43992715335401e-06,
+      "loss": 1.0303,
+      "step": 4132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4396070551391e-06,
+      "loss": 1.0661,
+      "step": 4133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43928687090781e-06,
+      "loss": 1.0097,
+      "step": 4134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438966600666344e-06,
+      "loss": 0.9142,
+      "step": 4135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438646244420909e-06,
+      "loss": 0.9924,
+      "step": 4136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438325802177708e-06,
+      "loss": 0.982,
+      "step": 4137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438005273942951e-06,
+      "loss": 1.041,
+      "step": 4138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43768465972285e-06,
+      "loss": 0.8016,
+      "step": 4139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.437363959523613e-06,
+      "loss": 0.7656,
+      "step": 4140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.437043173351458e-06,
+      "loss": 1.129,
+      "step": 4141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.436722301212598e-06,
+      "loss": 0.9593,
+      "step": 4142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.436401343113248e-06,
+      "loss": 0.9885,
+      "step": 4143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.436080299059629e-06,
+      "loss": 0.9791,
+      "step": 4144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43575916905796e-06,
+      "loss": 0.9167,
+      "step": 4145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.435437953114467e-06,
+      "loss": 0.7264,
+      "step": 4146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.435116651235365e-06,
+      "loss": 1.008,
+      "step": 4147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.434795263426887e-06,
+      "loss": 0.9776,
+      "step": 4148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.434473789695256e-06,
+      "loss": 1.0699,
+      "step": 4149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4341522300467e-06,
+      "loss": 0.8339,
+      "step": 4150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.433830584487452e-06,
+      "loss": 0.8396,
+      "step": 4151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.433508853023741e-06,
+      "loss": 0.9412,
+      "step": 4152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.433187035661802e-06,
+      "loss": 1.0785,
+      "step": 4153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43286513240787e-06,
+      "loss": 0.9589,
+      "step": 4154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.432543143268182e-06,
+      "loss": 1.2065,
+      "step": 4155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.432221068248975e-06,
+      "loss": 0.8846,
+      "step": 4156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.431898907356492e-06,
+      "loss": 0.8123,
+      "step": 4157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.431576660596971e-06,
+      "loss": 0.996,
+      "step": 4158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.431254327976658e-06,
+      "loss": 0.8431,
+      "step": 4159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.430931909501798e-06,
+      "loss": 1.0462,
+      "step": 4160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.430609405178636e-06,
+      "loss": 0.9053,
+      "step": 4161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.430286815013423e-06,
+      "loss": 1.361,
+      "step": 4162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.429964139012409e-06,
+      "loss": 0.8827,
+      "step": 4163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.429641377181843e-06,
+      "loss": 0.9911,
+      "step": 4164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42931852952798e-06,
+      "loss": 1.0459,
+      "step": 4165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428995596057077e-06,
+      "loss": 0.9979,
+      "step": 4166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428672576775388e-06,
+      "loss": 0.9546,
+      "step": 4167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428349471689172e-06,
+      "loss": 1.0235,
+      "step": 4168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42802628080469e-06,
+      "loss": 1.056,
+      "step": 4169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427703004128204e-06,
+      "loss": 0.813,
+      "step": 4170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427379641665975e-06,
+      "loss": 1.0518,
+      "step": 4171
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427056193424271e-06,
+      "loss": 0.9377,
+      "step": 4172
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426732659409357e-06,
+      "loss": 0.8438,
+      "step": 4173
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426409039627501e-06,
+      "loss": 1.0592,
+      "step": 4174
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426085334084975e-06,
+      "loss": 1.0406,
+      "step": 4175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425761542788049e-06,
+      "loss": 0.9237,
+      "step": 4176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425437665742998e-06,
+      "loss": 0.7877,
+      "step": 4177
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425113702956095e-06,
+      "loss": 1.0973,
+      "step": 4178
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42478965443362e-06,
+      "loss": 0.9553,
+      "step": 4179
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424465520181847e-06,
+      "loss": 0.8702,
+      "step": 4180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424141300207058e-06,
+      "loss": 1.1247,
+      "step": 4181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423816994515535e-06,
+      "loss": 0.9536,
+      "step": 4182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423492603113563e-06,
+      "loss": 1.1425,
+      "step": 4183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423168126007422e-06,
+      "loss": 0.9758,
+      "step": 4184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422843563203406e-06,
+      "loss": 0.9995,
+      "step": 4185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422518914707797e-06,
+      "loss": 1.0288,
+      "step": 4186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422194180526887e-06,
+      "loss": 0.9107,
+      "step": 4187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421869360666968e-06,
+      "loss": 1.0274,
+      "step": 4188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421544455134333e-06,
+      "loss": 1.0057,
+      "step": 4189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421219463935279e-06,
+      "loss": 1.0039,
+      "step": 4190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420894387076098e-06,
+      "loss": 0.9038,
+      "step": 4191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420569224563092e-06,
+      "loss": 0.8908,
+      "step": 4192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42024397640256e-06,
+      "loss": 0.872,
+      "step": 4193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419918642600805e-06,
+      "loss": 1.1277,
+      "step": 4194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419593223164128e-06,
+      "loss": 0.9919,
+      "step": 4195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419267718098835e-06,
+      "loss": 1.119,
+      "step": 4196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418942127411232e-06,
+      "loss": 1.1006,
+      "step": 4197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418616451107629e-06,
+      "loss": 1.2226,
+      "step": 4198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418290689194333e-06,
+      "loss": 0.9576,
+      "step": 4199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417964841677658e-06,
+      "loss": 1.0675,
+      "step": 4200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417638908563916e-06,
+      "loss": 0.8547,
+      "step": 4201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417312889859421e-06,
+      "loss": 0.9847,
+      "step": 4202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416986785570492e-06,
+      "loss": 0.9598,
+      "step": 4203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416660595703447e-06,
+      "loss": 0.8592,
+      "step": 4204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416334320264604e-06,
+      "loss": 0.9994,
+      "step": 4205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416007959260285e-06,
+      "loss": 1.0871,
+      "step": 4206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415681512696812e-06,
+      "loss": 0.8844,
+      "step": 4207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415354980580513e-06,
+      "loss": 1.1094,
+      "step": 4208
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415028362917713e-06,
+      "loss": 1.0778,
+      "step": 4209
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41470165971474e-06,
+      "loss": 0.8355,
+      "step": 4210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414374870977923e-06,
+      "loss": 1.0251,
+      "step": 4211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414047996713595e-06,
+      "loss": 0.9512,
+      "step": 4212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413721036928089e-06,
+      "loss": 1.0001,
+      "step": 4213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413393991627737e-06,
+      "loss": 1.102,
+      "step": 4214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41306686081888e-06,
+      "loss": 1.0444,
+      "step": 4215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412739644507851e-06,
+      "loss": 0.7394,
+      "step": 4216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412412342700993e-06,
+      "loss": 1.124,
+      "step": 4217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412084955404649e-06,
+      "loss": 0.9697,
+      "step": 4218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41175748262516e-06,
+      "loss": 0.8303,
+      "step": 4219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411429924368867e-06,
+      "loss": 0.7862,
+      "step": 4220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411102280642121e-06,
+      "loss": 0.9411,
+      "step": 4221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41077455145127e-06,
+      "loss": 0.9558,
+      "step": 4222
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410446736802663e-06,
+      "loss": 1.0859,
+      "step": 4223
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41011883670265e-06,
+      "loss": 0.7747,
+      "step": 4224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409790851157585e-06,
+      "loss": 1.2481,
+      "step": 4225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409462780173823e-06,
+      "loss": 0.8752,
+      "step": 4226
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409134623757718e-06,
+      "loss": 1.0185,
+      "step": 4227
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408806381915631e-06,
+      "loss": 0.9276,
+      "step": 4228
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408478054653923e-06,
+      "loss": 0.9707,
+      "step": 4229
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40814964197895e-06,
+      "loss": 0.9885,
+      "step": 4230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407821143897078e-06,
+      "loss": 0.9917,
+      "step": 4231
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407492560414671e-06,
+      "loss": 1.0472,
+      "step": 4232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407163891538095e-06,
+      "loss": 0.824,
+      "step": 4233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40683513727372e-06,
+      "loss": 0.8665,
+      "step": 4234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406506297627913e-06,
+      "loss": 1.1652,
+      "step": 4235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406177372607046e-06,
+      "loss": 0.9063,
+      "step": 4236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40584836221749e-06,
+      "loss": 0.9717,
+      "step": 4237
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405519266465625e-06,
+      "loss": 1.2245,
+      "step": 4238
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405190085357821e-06,
+      "loss": 0.9001,
+      "step": 4239
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404860818900458e-06,
+      "loss": 0.9612,
+      "step": 4240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404531467099918e-06,
+      "loss": 1.0025,
+      "step": 4241
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404202029962579e-06,
+      "loss": 0.7994,
+      "step": 4242
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403872507494825e-06,
+      "loss": 1.1027,
+      "step": 4243
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40354289970304e-06,
+      "loss": 1.0618,
+      "step": 4244
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403213206593608e-06,
+      "loss": 1.0656,
+      "step": 4245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402883428172922e-06,
+      "loss": 0.9363,
+      "step": 4246
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402553564447367e-06,
+      "loss": 0.9637,
+      "step": 4247
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402223615423335e-06,
+      "loss": 0.8699,
+      "step": 4248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401893581107221e-06,
+      "loss": 0.8614,
+      "step": 4249
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401563461505415e-06,
+      "loss": 1.0578,
+      "step": 4250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401233256624318e-06,
+      "loss": 1.0026,
+      "step": 4251
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400902966470323e-06,
+      "loss": 0.9892,
+      "step": 4252
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40057259104983e-06,
+      "loss": 1.0985,
+      "step": 4253
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400242130369243e-06,
+      "loss": 1.0425,
+      "step": 4254
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399911584434964e-06,
+      "loss": 0.9054,
+      "step": 4255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399580953253395e-06,
+      "loss": 0.8259,
+      "step": 4256
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399250236830944e-06,
+      "loss": 0.9927,
+      "step": 4257
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398919435174015e-06,
+      "loss": 0.9631,
+      "step": 4258
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398588548289022e-06,
+      "loss": 1.0615,
+      "step": 4259
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398257576182372e-06,
+      "loss": 1.0465,
+      "step": 4260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39792651886048e-06,
+      "loss": 0.925,
+      "step": 4261
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39759537632976e-06,
+      "loss": 1.1073,
+      "step": 4262
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397264148596627e-06,
+      "loss": 0.9198,
+      "step": 4263
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396932835667499e-06,
+      "loss": 1.0534,
+      "step": 4264
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396601437548794e-06,
+      "loss": 1.2337,
+      "step": 4265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396269954246932e-06,
+      "loss": 0.981,
+      "step": 4266
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395938385768338e-06,
+      "loss": 0.9114,
+      "step": 4267
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395606732119435e-06,
+      "loss": 1.1092,
+      "step": 4268
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395274993306649e-06,
+      "loss": 0.9871,
+      "step": 4269
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.394943169336407e-06,
+      "loss": 1.0441,
+      "step": 4270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.394611260215138e-06,
+      "loss": 0.9741,
+      "step": 4271
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.394279265949272e-06,
+      "loss": 1.1192,
+      "step": 4272
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.393947186545243e-06,
+      "loss": 0.8358,
+      "step": 4273
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.393615022009484e-06,
+      "loss": 0.9637,
+      "step": 4274
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.393282772348429e-06,
+      "loss": 0.8547,
+      "step": 4275
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.392950437568518e-06,
+      "loss": 1.0059,
+      "step": 4276
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39261801767619e-06,
+      "loss": 1.049,
+      "step": 4277
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.392285512677883e-06,
+      "loss": 0.8583,
+      "step": 4278
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39195292258004e-06,
+      "loss": 0.876,
+      "step": 4279
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.391620247389106e-06,
+      "loss": 1.0194,
+      "step": 4280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.391287487111525e-06,
+      "loss": 0.9239,
+      "step": 4281
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.390954641753746e-06,
+      "loss": 0.9185,
+      "step": 4282
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.390621711322216e-06,
+      "loss": 1.0096,
+      "step": 4283
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.390288695823388e-06,
+      "loss": 1.0211,
+      "step": 4284
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.389955595263712e-06,
+      "loss": 1.107,
+      "step": 4285
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38962240964964e-06,
+      "loss": 1.0905,
+      "step": 4286
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.389289138987632e-06,
+      "loss": 1.0165,
+      "step": 4287
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.388955783284141e-06,
+      "loss": 0.9146,
+      "step": 4288
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.388622342545628e-06,
+      "loss": 1.0671,
+      "step": 4289
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.388288816778553e-06,
+      "loss": 1.0831,
+      "step": 4290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.387955205989379e-06,
+      "loss": 1.1049,
+      "step": 4291
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.387621510184565e-06,
+      "loss": 1.0327,
+      "step": 4292
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.387287729370583e-06,
+      "loss": 0.9067,
+      "step": 4293
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.386953863553894e-06,
+      "loss": 0.9896,
+      "step": 4294
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.386619912740969e-06,
+      "loss": 0.977,
+      "step": 4295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.386285876938278e-06,
+      "loss": 1.0312,
+      "step": 4296
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.385951756152294e-06,
+      "loss": 1.0816,
+      "step": 4297
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38561755038949e-06,
+      "loss": 1.0673,
+      "step": 4298
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.385283259656339e-06,
+      "loss": 0.8108,
+      "step": 4299
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.384948883959322e-06,
+      "loss": 1.0059,
+      "step": 4300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.384614423304913e-06,
+      "loss": 0.9614,
+      "step": 4301
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.384279877699596e-06,
+      "loss": 1.0065,
+      "step": 4302
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38394524714985e-06,
+      "loss": 0.9984,
+      "step": 4303
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.383610531662159e-06,
+      "loss": 1.0293,
+      "step": 4304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.383275731243008e-06,
+      "loss": 1.0883,
+      "step": 4305
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382940845898884e-06,
+      "loss": 0.9795,
+      "step": 4306
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382605875636277e-06,
+      "loss": 0.9775,
+      "step": 4307
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382270820461674e-06,
+      "loss": 1.1164,
+      "step": 4308
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381935680381568e-06,
+      "loss": 0.9726,
+      "step": 4309
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.381600455402451e-06,
+      "loss": 0.992,
+      "step": 4310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38126514553082e-06,
+      "loss": 0.9416,
+      "step": 4311
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380929750773172e-06,
+      "loss": 0.9598,
+      "step": 4312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380594271136004e-06,
+      "loss": 0.9329,
+      "step": 4313
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380258706625814e-06,
+      "loss": 1.0052,
+      "step": 4314
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379923057249105e-06,
+      "loss": 0.8864,
+      "step": 4315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379587323012381e-06,
+      "loss": 0.98,
+      "step": 4316
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379251503922146e-06,
+      "loss": 0.9242,
+      "step": 4317
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378915599984906e-06,
+      "loss": 0.947,
+      "step": 4318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37857961120717e-06,
+      "loss": 0.856,
+      "step": 4319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378243537595449e-06,
+      "loss": 1.1178,
+      "step": 4320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37790737915625e-06,
+      "loss": 0.8276,
+      "step": 4321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377571135896088e-06,
+      "loss": 1.013,
+      "step": 4322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37723480782148e-06,
+      "loss": 0.8695,
+      "step": 4323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376898394938942e-06,
+      "loss": 0.8265,
+      "step": 4324
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376561897254987e-06,
+      "loss": 0.7747,
+      "step": 4325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37622531477614e-06,
+      "loss": 1.0229,
+      "step": 4326
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37588864750892e-06,
+      "loss": 0.98,
+      "step": 4327
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375551895459852e-06,
+      "loss": 1.1843,
+      "step": 4328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375215058635457e-06,
+      "loss": 1.047,
+      "step": 4329
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37487813704226e-06,
+      "loss": 0.9129,
+      "step": 4330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374541130686794e-06,
+      "loss": 0.9469,
+      "step": 4331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374204039575587e-06,
+      "loss": 0.9979,
+      "step": 4332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373866863715167e-06,
+      "loss": 0.9767,
+      "step": 4333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373529603112068e-06,
+      "loss": 1.0075,
+      "step": 4334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373192257772827e-06,
+      "loss": 1.0215,
+      "step": 4335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37285482770398e-06,
+      "loss": 0.9626,
+      "step": 4336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372517312912059e-06,
+      "loss": 1.1361,
+      "step": 4337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372179713403607e-06,
+      "loss": 0.978,
+      "step": 4338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371842029185164e-06,
+      "loss": 0.9453,
+      "step": 4339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371504260263278e-06,
+      "loss": 0.9331,
+      "step": 4340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371166406644485e-06,
+      "loss": 1.047,
+      "step": 4341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370828468335335e-06,
+      "loss": 0.9813,
+      "step": 4342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370490445342375e-06,
+      "loss": 1.0796,
+      "step": 4343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370152337672154e-06,
+      "loss": 0.9444,
+      "step": 4344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369814145331223e-06,
+      "loss": 0.8775,
+      "step": 4345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369475868326135e-06,
+      "loss": 1.1085,
+      "step": 4346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369137506663441e-06,
+      "loss": 0.9455,
+      "step": 4347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3687990603497e-06,
+      "loss": 0.9337,
+      "step": 4348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368460529391468e-06,
+      "loss": 0.8085,
+      "step": 4349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368121913795305e-06,
+      "loss": 0.9618,
+      "step": 4350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36778321356777e-06,
+      "loss": 0.9211,
+      "step": 4351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367444428715426e-06,
+      "loss": 0.9718,
+      "step": 4352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367105559244837e-06,
+      "loss": 0.8632,
+      "step": 4353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366766605162569e-06,
+      "loss": 0.9255,
+      "step": 4354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366427566475189e-06,
+      "loss": 1.0595,
+      "step": 4355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366088443189263e-06,
+      "loss": 0.8582,
+      "step": 4356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365749235311367e-06,
+      "loss": 0.9382,
+      "step": 4357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365409942848067e-06,
+      "loss": 0.9755,
+      "step": 4358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365070565805941e-06,
+      "loss": 0.9779,
+      "step": 4359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.364731104191563e-06,
+      "loss": 0.8176,
+      "step": 4360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36439155801151e-06,
+      "loss": 0.9637,
+      "step": 4361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36405192727236e-06,
+      "loss": 0.9803,
+      "step": 4362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363712211980694e-06,
+      "loss": 0.9854,
+      "step": 4363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363372412143096e-06,
+      "loss": 0.9245,
+      "step": 4364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363032527766144e-06,
+      "loss": 1.048,
+      "step": 4365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362692558856427e-06,
+      "loss": 1.1597,
+      "step": 4366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362352505420532e-06,
+      "loss": 0.9688,
+      "step": 4367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362012367465048e-06,
+      "loss": 0.9394,
+      "step": 4368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.361672144996562e-06,
+      "loss": 0.9819,
+      "step": 4369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.361331838021668e-06,
+      "loss": 0.9374,
+      "step": 4370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.360991446546958e-06,
+      "loss": 1.0038,
+      "step": 4371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36065097057903e-06,
+      "loss": 1.0744,
+      "step": 4372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.360310410124478e-06,
+      "loss": 0.9887,
+      "step": 4373
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3599697651899e-06,
+      "loss": 0.8753,
+      "step": 4374
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.359629035781897e-06,
+      "loss": 0.9422,
+      "step": 4375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.359288221907071e-06,
+      "loss": 0.8735,
+      "step": 4376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.358947323572025e-06,
+      "loss": 0.9502,
+      "step": 4377
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.358606340783363e-06,
+      "loss": 0.9367,
+      "step": 4378
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35826527354769e-06,
+      "loss": 0.8357,
+      "step": 4379
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.357924121871618e-06,
+      "loss": 1.0405,
+      "step": 4380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.357582885761754e-06,
+      "loss": 0.8613,
+      "step": 4381
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.357241565224711e-06,
+      "loss": 0.9939,
+      "step": 4382
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.356900160267099e-06,
+      "loss": 0.8079,
+      "step": 4383
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.356558670895537e-06,
+      "loss": 0.9038,
+      "step": 4384
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.356217097116637e-06,
+      "loss": 1.0529,
+      "step": 4385
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.355875438937021e-06,
+      "loss": 0.9978,
+      "step": 4386
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.355533696363305e-06,
+      "loss": 0.8309,
+      "step": 4387
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.355191869402112e-06,
+      "loss": 0.9569,
+      "step": 4388
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.354849958060064e-06,
+      "loss": 0.9628,
+      "step": 4389
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.354507962343787e-06,
+      "loss": 0.9102,
+      "step": 4390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.354165882259904e-06,
+      "loss": 0.8221,
+      "step": 4391
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.353823717815047e-06,
+      "loss": 1.0806,
+      "step": 4392
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.353481469015842e-06,
+      "loss": 0.7666,
+      "step": 4393
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35313913586892e-06,
+      "loss": 0.9193,
+      "step": 4394
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.352796718380916e-06,
+      "loss": 0.8775,
+      "step": 4395
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.352454216558464e-06,
+      "loss": 1.0512,
+      "step": 4396
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.352111630408198e-06,
+      "loss": 1.1694,
+      "step": 4397
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.351768959936755e-06,
+      "loss": 0.9118,
+      "step": 4398
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.351426205150778e-06,
+      "loss": 1.0466,
+      "step": 4399
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.351083366056904e-06,
+      "loss": 0.8488,
+      "step": 4400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.350740442661776e-06,
+      "loss": 1.0104,
+      "step": 4401
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35039743497204e-06,
+      "loss": 0.8375,
+      "step": 4402
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35005434299434e-06,
+      "loss": 1.087,
+      "step": 4403
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.349711166735322e-06,
+      "loss": 0.9052,
+      "step": 4404
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34936790620164e-06,
+      "loss": 0.9734,
+      "step": 4405
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34902456139994e-06,
+      "loss": 0.9209,
+      "step": 4406
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.348681132336876e-06,
+      "loss": 0.9181,
+      "step": 4407
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3483376190191e-06,
+      "loss": 1.045,
+      "step": 4408
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.347994021453272e-06,
+      "loss": 0.848,
+      "step": 4409
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.347650339646045e-06,
+      "loss": 0.9313,
+      "step": 4410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.347306573604079e-06,
+      "loss": 0.9468,
+      "step": 4411
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.346962723334033e-06,
+      "loss": 1.0728,
+      "step": 4412
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34661878884257e-06,
+      "loss": 0.9877,
+      "step": 4413
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.346274770136355e-06,
+      "loss": 0.8301,
+      "step": 4414
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.345930667222052e-06,
+      "loss": 1.0163,
+      "step": 4415
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.345586480106327e-06,
+      "loss": 1.0747,
+      "step": 4416
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.345242208795852e-06,
+      "loss": 0.9116,
+      "step": 4417
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.344897853297294e-06,
+      "loss": 0.8197,
+      "step": 4418
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.344553413617325e-06,
+      "loss": 1.024,
+      "step": 4419
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34420888976262e-06,
+      "loss": 0.9511,
+      "step": 4420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.343864281739854e-06,
+      "loss": 1.0384,
+      "step": 4421
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3435195895557e-06,
+      "loss": 0.9531,
+      "step": 4422
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34317481321684e-06,
+      "loss": 0.9719,
+      "step": 4423
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.342829952729954e-06,
+      "loss": 0.7555,
+      "step": 4424
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.342485008101724e-06,
+      "loss": 0.9124,
+      "step": 4425
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.342139979338831e-06,
+      "loss": 1.0593,
+      "step": 4426
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34179486644796e-06,
+      "loss": 1.002,
+      "step": 4427
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3414496694358e-06,
+      "loss": 0.9706,
+      "step": 4428
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.341104388309036e-06,
+      "loss": 0.9638,
+      "step": 4429
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34075902307436e-06,
+      "loss": 1.0215,
+      "step": 4430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.340413573738463e-06,
+      "loss": 1.137,
+      "step": 4431
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.340068040308036e-06,
+      "loss": 1.0527,
+      "step": 4432
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.339722422789777e-06,
+      "loss": 0.9911,
+      "step": 4433
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.33937672119038e-06,
+      "loss": 0.865,
+      "step": 4434
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.339030935516542e-06,
+      "loss": 1.0042,
+      "step": 4435
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.338685065774964e-06,
+      "loss": 1.0606,
+      "step": 4436
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.338339111972348e-06,
+      "loss": 1.1545,
+      "step": 4437
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.337993074115397e-06,
+      "loss": 0.9864,
+      "step": 4438
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.337646952210813e-06,
+      "loss": 1.1408,
+      "step": 4439
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.337300746265303e-06,
+      "loss": 1.0114,
+      "step": 4440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.336954456285574e-06,
+      "loss": 0.8848,
+      "step": 4441
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.336608082278338e-06,
+      "loss": 1.0502,
+      "step": 4442
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.336261624250303e-06,
+      "loss": 1.0038,
+      "step": 4443
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.335915082208184e-06,
+      "loss": 1.2256,
+      "step": 4444
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.335568456158692e-06,
+      "loss": 0.9871,
+      "step": 4445
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.335221746108545e-06,
+      "loss": 0.8964,
+      "step": 4446
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.334874952064462e-06,
+      "loss": 0.9043,
+      "step": 4447
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.334528074033158e-06,
+      "loss": 1.1247,
+      "step": 4448
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.334181112021357e-06,
+      "loss": 1.1564,
+      "step": 4449
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.333834066035778e-06,
+      "loss": 1.0921,
+      "step": 4450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.333486936083149e-06,
+      "loss": 0.948,
+      "step": 4451
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.333139722170193e-06,
+      "loss": 0.9135,
+      "step": 4452
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.332792424303638e-06,
+      "loss": 0.987,
+      "step": 4453
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.332445042490212e-06,
+      "loss": 0.9511,
+      "step": 4454
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.332097576736648e-06,
+      "loss": 1.185,
+      "step": 4455
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.331750027049675e-06,
+      "loss": 0.8797,
+      "step": 4456
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.331402393436027e-06,
+      "loss": 0.9843,
+      "step": 4457
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.331054675902441e-06,
+      "loss": 0.9057,
+      "step": 4458
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.330706874455653e-06,
+      "loss": 0.9142,
+      "step": 4459
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.330358989102402e-06,
+      "loss": 0.8053,
+      "step": 4460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.330011019849428e-06,
+      "loss": 1.0129,
+      "step": 4461
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.329662966703472e-06,
+      "loss": 0.9774,
+      "step": 4462
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.329314829671278e-06,
+      "loss": 0.8278,
+      "step": 4463
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.328966608759591e-06,
+      "loss": 1.0045,
+      "step": 4464
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32861830397516e-06,
+      "loss": 1.0582,
+      "step": 4465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.328269915324728e-06,
+      "loss": 0.7669,
+      "step": 4466
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32792144281505e-06,
+      "loss": 0.9913,
+      "step": 4467
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.327572886452875e-06,
+      "loss": 0.9939,
+      "step": 4468
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32722424624496e-06,
+      "loss": 1.1008,
+      "step": 4469
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.326875522198054e-06,
+      "loss": 1.0448,
+      "step": 4470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.326526714318916e-06,
+      "loss": 0.8617,
+      "step": 4471
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.326177822614305e-06,
+      "loss": 1.0328,
+      "step": 4472
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32582884709098e-06,
+      "loss": 0.778,
+      "step": 4473
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.325479787755703e-06,
+      "loss": 0.9649,
+      "step": 4474
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.325130644615236e-06,
+      "loss": 1.1136,
+      "step": 4475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.324781417676343e-06,
+      "loss": 0.9206,
+      "step": 4476
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.324432106945791e-06,
+      "loss": 0.9352,
+      "step": 4477
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.324082712430349e-06,
+      "loss": 1.0473,
+      "step": 4478
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.323733234136785e-06,
+      "loss": 0.9935,
+      "step": 4479
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.323383672071871e-06,
+      "loss": 0.9751,
+      "step": 4480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.323034026242378e-06,
+      "loss": 1.0273,
+      "step": 4481
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.322684296655083e-06,
+      "loss": 1.1409,
+      "step": 4482
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.322334483316758e-06,
+      "loss": 1.0358,
+      "step": 4483
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.321984586234182e-06,
+      "loss": 1.1477,
+      "step": 4484
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.321634605414138e-06,
+      "loss": 0.8991,
+      "step": 4485
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.321284540863403e-06,
+      "loss": 1.0414,
+      "step": 4486
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32093439258876e-06,
+      "loss": 0.883,
+      "step": 4487
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.320584160596993e-06,
+      "loss": 1.0367,
+      "step": 4488
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.320233844894888e-06,
+      "loss": 1.0721,
+      "step": 4489
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.319883445489232e-06,
+      "loss": 1.0095,
+      "step": 4490
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.319532962386816e-06,
+      "loss": 0.998,
+      "step": 4491
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.319182395594427e-06,
+      "loss": 0.991,
+      "step": 4492
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.318831745118859e-06,
+      "loss": 0.9852,
+      "step": 4493
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.318481010966907e-06,
+      "loss": 0.9103,
+      "step": 4494
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.318130193145362e-06,
+      "loss": 0.9723,
+      "step": 4495
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.317779291661027e-06,
+      "loss": 1.1774,
+      "step": 4496
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.317428306520696e-06,
+      "loss": 1.1312,
+      "step": 4497
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.317077237731172e-06,
+      "loss": 0.8323,
+      "step": 4498
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.316726085299255e-06,
+      "loss": 0.9606,
+      "step": 4499
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.316374849231748e-06,
+      "loss": 1.0082,
+      "step": 4500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.31602352953546e-06,
+      "loss": 1.0598,
+      "step": 4501
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.315672126217194e-06,
+      "loss": 1.0865,
+      "step": 4502
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.31532063928376e-06,
+      "loss": 1.0919,
+      "step": 4503
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.314969068741967e-06,
+      "loss": 0.9307,
+      "step": 4504
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.314617414598627e-06,
+      "loss": 0.9882,
+      "step": 4505
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.314265676860553e-06,
+      "loss": 1.0959,
+      "step": 4506
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.31391385553456e-06,
+      "loss": 0.9319,
+      "step": 4507
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.313561950627465e-06,
+      "loss": 1.0112,
+      "step": 4508
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.313209962146087e-06,
+      "loss": 0.9302,
+      "step": 4509
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.312857890097243e-06,
+      "loss": 0.8671,
+      "step": 4510
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.312505734487755e-06,
+      "loss": 1.025,
+      "step": 4511
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.312153495324449e-06,
+      "loss": 0.8545,
+      "step": 4512
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.311801172614146e-06,
+      "loss": 1.0528,
+      "step": 4513
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.311448766363674e-06,
+      "loss": 0.9784,
+      "step": 4514
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.311096276579858e-06,
+      "loss": 0.7871,
+      "step": 4515
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.310743703269531e-06,
+      "loss": 1.0602,
+      "step": 4516
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.310391046439522e-06,
+      "loss": 1.0611,
+      "step": 4517
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.310038306096665e-06,
+      "loss": 1.0241,
+      "step": 4518
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.309685482247794e-06,
+      "loss": 0.853,
+      "step": 4519
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.309332574899744e-06,
+      "loss": 1.0896,
+      "step": 4520
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.30897958405935e-06,
+      "loss": 0.8974,
+      "step": 4521
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.308626509733459e-06,
+      "loss": 0.9221,
+      "step": 4522
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.308273351928903e-06,
+      "loss": 1.0678,
+      "step": 4523
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.30792011065253e-06,
+      "loss": 0.9671,
+      "step": 4524
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.30756678591118e-06,
+      "loss": 0.9545,
+      "step": 4525
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.307213377711702e-06,
+      "loss": 0.9267,
+      "step": 4526
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.30685988606094e-06,
+      "loss": 0.8253,
+      "step": 4527
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.306506310965749e-06,
+      "loss": 0.8745,
+      "step": 4528
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.306152652432972e-06,
+      "loss": 0.8794,
+      "step": 4529
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.305798910469465e-06,
+      "loss": 0.8766,
+      "step": 4530
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.305445085082082e-06,
+      "loss": 0.8756,
+      "step": 4531
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.305091176277674e-06,
+      "loss": 1.044,
+      "step": 4532
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.304737184063103e-06,
+      "loss": 1.0423,
+      "step": 4533
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.304383108445225e-06,
+      "loss": 0.8506,
+      "step": 4534
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.304028949430902e-06,
+      "loss": 1.1732,
+      "step": 4535
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.303674707026993e-06,
+      "loss": 0.9542,
+      "step": 4536
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.303320381240362e-06,
+      "loss": 1.0303,
+      "step": 4537
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.302965972077877e-06,
+      "loss": 1.0059,
+      "step": 4538
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.3026114795464e-06,
+      "loss": 0.8077,
+      "step": 4539
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.302256903652804e-06,
+      "loss": 1.1634,
+      "step": 4540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.301902244403956e-06,
+      "loss": 0.886,
+      "step": 4541
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.301547501806725e-06,
+      "loss": 0.9638,
+      "step": 4542
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.30119267586799e-06,
+      "loss": 0.773,
+      "step": 4543
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.300837766594621e-06,
+      "loss": 1.0669,
+      "step": 4544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.300482773993497e-06,
+      "loss": 0.8381,
+      "step": 4545
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.300127698071493e-06,
+      "loss": 1.1243,
+      "step": 4546
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.299772538835492e-06,
+      "loss": 1.0732,
+      "step": 4547
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.299417296292373e-06,
+      "loss": 1.0609,
+      "step": 4548
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29906197044902e-06,
+      "loss": 0.869,
+      "step": 4549
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.298706561312315e-06,
+      "loss": 1.1383,
+      "step": 4550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.298351068889147e-06,
+      "loss": 1.177,
+      "step": 4551
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2979954931864e-06,
+      "loss": 0.8756,
+      "step": 4552
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.297639834210968e-06,
+      "loss": 1.0417,
+      "step": 4553
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.297284091969736e-06,
+      "loss": 0.9243,
+      "step": 4554
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.296928266469602e-06,
+      "loss": 0.9053,
+      "step": 4555
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.296572357717457e-06,
+      "loss": 0.9702,
+      "step": 4556
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.296216365720195e-06,
+      "loss": 0.8412,
+      "step": 4557
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.295860290484717e-06,
+      "loss": 1.0012,
+      "step": 4558
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29550413201792e-06,
+      "loss": 1.0274,
+      "step": 4559
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.295147890326706e-06,
+      "loss": 0.8776,
+      "step": 4560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.294791565417975e-06,
+      "loss": 0.976,
+      "step": 4561
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.294435157298632e-06,
+      "loss": 1.0158,
+      "step": 4562
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.294078665975582e-06,
+      "loss": 0.9325,
+      "step": 4563
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.293722091455732e-06,
+      "loss": 0.9871,
+      "step": 4564
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.293365433745991e-06,
+      "loss": 1.0433,
+      "step": 4565
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.293008692853267e-06,
+      "loss": 0.8299,
+      "step": 4566
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.292651868784475e-06,
+      "loss": 0.973,
+      "step": 4567
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.292294961546526e-06,
+      "loss": 1.01,
+      "step": 4568
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.291937971146335e-06,
+      "loss": 1.0177,
+      "step": 4569
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29158089759082e-06,
+      "loss": 0.8856,
+      "step": 4570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2912237408869e-06,
+      "loss": 1.0634,
+      "step": 4571
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29086650104149e-06,
+      "loss": 0.8517,
+      "step": 4572
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29050917806152e-06,
+      "loss": 0.9166,
+      "step": 4573
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.290151771953904e-06,
+      "loss": 1.0966,
+      "step": 4574
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28979428272557e-06,
+      "loss": 0.9974,
+      "step": 4575
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.289436710383447e-06,
+      "loss": 1.1484,
+      "step": 4576
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28907905493446e-06,
+      "loss": 0.8836,
+      "step": 4577
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.288721316385541e-06,
+      "loss": 1.0472,
+      "step": 4578
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.288363494743618e-06,
+      "loss": 1.1724,
+      "step": 4579
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.288005590015623e-06,
+      "loss": 0.8745,
+      "step": 4580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.287647602208495e-06,
+      "loss": 1.0565,
+      "step": 4581
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.287289531329167e-06,
+      "loss": 0.8241,
+      "step": 4582
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.286931377384576e-06,
+      "loss": 0.99,
+      "step": 4583
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.286573140381663e-06,
+      "loss": 0.9868,
+      "step": 4584
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.286214820327366e-06,
+      "loss": 0.8581,
+      "step": 4585
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.285856417228631e-06,
+      "loss": 1.0012,
+      "step": 4586
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2854979310924e-06,
+      "loss": 0.9596,
+      "step": 4587
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.285139361925617e-06,
+      "loss": 0.9623,
+      "step": 4588
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.284780709735232e-06,
+      "loss": 0.9356,
+      "step": 4589
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.284421974528194e-06,
+      "loss": 0.9906,
+      "step": 4590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.284063156311448e-06,
+      "loss": 0.917,
+      "step": 4591
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.283704255091954e-06,
+      "loss": 0.9023,
+      "step": 4592
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.283345270876662e-06,
+      "loss": 1.0189,
+      "step": 4593
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.282986203672525e-06,
+      "loss": 0.9945,
+      "step": 4594
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.282627053486502e-06,
+      "loss": 0.9638,
+      "step": 4595
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.282267820325552e-06,
+      "loss": 0.9261,
+      "step": 4596
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.281908504196635e-06,
+      "loss": 0.9274,
+      "step": 4597
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28154910510671e-06,
+      "loss": 0.9582,
+      "step": 4598
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.281189623062745e-06,
+      "loss": 0.9736,
+      "step": 4599
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2808300580717e-06,
+      "loss": 1.0128,
+      "step": 4600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.280470410140546e-06,
+      "loss": 0.9794,
+      "step": 4601
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.280110679276246e-06,
+      "loss": 1.0201,
+      "step": 4602
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.279750865485772e-06,
+      "loss": 0.8713,
+      "step": 4603
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.279390968776099e-06,
+      "loss": 0.833,
+      "step": 4604
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.279030989154196e-06,
+      "loss": 0.9238,
+      "step": 4605
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.278670926627035e-06,
+      "loss": 0.9928,
+      "step": 4606
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.278310781201599e-06,
+      "loss": 0.9929,
+      "step": 4607
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27795055288486e-06,
+      "loss": 0.926,
+      "step": 4608
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2775902416838e-06,
+      "loss": 1.0964,
+      "step": 4609
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2772298476054e-06,
+      "loss": 1.0932,
+      "step": 4610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.276869370656643e-06,
+      "loss": 0.8692,
+      "step": 4611
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27650881084451e-06,
+      "loss": 1.1611,
+      "step": 4612
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.276148168175988e-06,
+      "loss": 0.9894,
+      "step": 4613
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.275787442658068e-06,
+      "loss": 1.2723,
+      "step": 4614
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.275426634297735e-06,
+      "loss": 1.0894,
+      "step": 4615
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27506574310198e-06,
+      "loss": 1.0116,
+      "step": 4616
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.274704769077798e-06,
+      "loss": 1.229,
+      "step": 4617
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.274343712232178e-06,
+      "loss": 0.9522,
+      "step": 4618
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.273982572572123e-06,
+      "loss": 0.8874,
+      "step": 4619
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27362135010462e-06,
+      "loss": 1.1972,
+      "step": 4620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.273260044836675e-06,
+      "loss": 0.9466,
+      "step": 4621
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.272898656775285e-06,
+      "loss": 0.7934,
+      "step": 4622
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.272537185927453e-06,
+      "loss": 1.1056,
+      "step": 4623
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.272175632300184e-06,
+      "loss": 0.9944,
+      "step": 4624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.271813995900479e-06,
+      "loss": 0.9689,
+      "step": 4625
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.271452276735348e-06,
+      "loss": 0.9103,
+      "step": 4626
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.271090474811797e-06,
+      "loss": 0.8614,
+      "step": 4627
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.270728590136839e-06,
+      "loss": 0.8778,
+      "step": 4628
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.270366622717481e-06,
+      "loss": 1.0519,
+      "step": 4629
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27000457256074e-06,
+      "loss": 0.7252,
+      "step": 4630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.269642439673627e-06,
+      "loss": 1.1265,
+      "step": 4631
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.269280224063162e-06,
+      "loss": 0.8768,
+      "step": 4632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26891792573636e-06,
+      "loss": 0.9575,
+      "step": 4633
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.268555544700242e-06,
+      "loss": 1.0226,
+      "step": 4634
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26819308096183e-06,
+      "loss": 0.9676,
+      "step": 4635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.267830534528144e-06,
+      "loss": 1.0495,
+      "step": 4636
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26746790540621e-06,
+      "loss": 1.0019,
+      "step": 4637
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.267105193603051e-06,
+      "loss": 0.9786,
+      "step": 4638
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2667423991257e-06,
+      "loss": 0.8969,
+      "step": 4639
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.266379521981181e-06,
+      "loss": 0.9686,
+      "step": 4640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.266016562176527e-06,
+      "loss": 0.9645,
+      "step": 4641
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26565351971877e-06,
+      "loss": 0.9877,
+      "step": 4642
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.265290394614942e-06,
+      "loss": 0.9567,
+      "step": 4643
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.264927186872084e-06,
+      "loss": 1.108,
+      "step": 4644
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.264563896497227e-06,
+      "loss": 0.8965,
+      "step": 4645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26420052349741e-06,
+      "loss": 0.9322,
+      "step": 4646
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.263837067879678e-06,
+      "loss": 0.9072,
+      "step": 4647
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.263473529651069e-06,
+      "loss": 0.868,
+      "step": 4648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.263109908818628e-06,
+      "loss": 0.8302,
+      "step": 4649
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.262746205389399e-06,
+      "loss": 0.9797,
+      "step": 4650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.262382419370427e-06,
+      "loss": 0.8999,
+      "step": 4651
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.262018550768766e-06,
+      "loss": 0.8121,
+      "step": 4652
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.261654599591461e-06,
+      "loss": 1.0018,
+      "step": 4653
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.261290565845566e-06,
+      "loss": 1.0265,
+      "step": 4654
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26092644953813e-06,
+      "loss": 1.1107,
+      "step": 4655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.260562250676213e-06,
+      "loss": 0.84,
+      "step": 4656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26019796926687e-06,
+      "loss": 0.838,
+      "step": 4657
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.259833605317155e-06,
+      "loss": 0.9208,
+      "step": 4658
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.259469158834131e-06,
+      "loss": 1.0745,
+      "step": 4659
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.259104629824858e-06,
+      "loss": 1.1755,
+      "step": 4660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.258740018296399e-06,
+      "loss": 0.9234,
+      "step": 4661
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25837532425582e-06,
+      "loss": 0.6845,
+      "step": 4662
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.258010547710182e-06,
+      "loss": 0.874,
+      "step": 4663
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.257645688666557e-06,
+      "loss": 1.0818,
+      "step": 4664
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.257280747132013e-06,
+      "loss": 1.0196,
+      "step": 4665
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25691572311362e-06,
+      "loss": 1.0676,
+      "step": 4666
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25655061661845e-06,
+      "loss": 0.9094,
+      "step": 4667
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.256185427653578e-06,
+      "loss": 1.0822,
+      "step": 4668
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25582015622608e-06,
+      "loss": 1.1009,
+      "step": 4669
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.255454802343029e-06,
+      "loss": 0.9077,
+      "step": 4670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.255089366011507e-06,
+      "loss": 0.7883,
+      "step": 4671
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.254723847238596e-06,
+      "loss": 1.2487,
+      "step": 4672
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.254358246031374e-06,
+      "loss": 0.9393,
+      "step": 4673
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.253992562396925e-06,
+      "loss": 0.8904,
+      "step": 4674
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.253626796342337e-06,
+      "loss": 0.8913,
+      "step": 4675
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.253260947874694e-06,
+      "loss": 1.01,
+      "step": 4676
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.252895017001085e-06,
+      "loss": 0.9916,
+      "step": 4677
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2525290037286e-06,
+      "loss": 0.9512,
+      "step": 4678
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.252162908064331e-06,
+      "loss": 0.8948,
+      "step": 4679
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.251796730015368e-06,
+      "loss": 0.843,
+      "step": 4680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25143046958881e-06,
+      "loss": 0.8705,
+      "step": 4681
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.251064126791749e-06,
+      "loss": 0.9115,
+      "step": 4682
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.250697701631286e-06,
+      "loss": 0.9515,
+      "step": 4683
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.250331194114518e-06,
+      "loss": 0.9886,
+      "step": 4684
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.249964604248549e-06,
+      "loss": 0.9433,
+      "step": 4685
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24959793204048e-06,
+      "loss": 1.098,
+      "step": 4686
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.249231177497414e-06,
+      "loss": 1.0137,
+      "step": 4687
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.248864340626458e-06,
+      "loss": 0.8456,
+      "step": 4688
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.248497421434719e-06,
+      "loss": 1.108,
+      "step": 4689
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.248130419929307e-06,
+      "loss": 0.8038,
+      "step": 4690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24776333611733e-06,
+      "loss": 0.8554,
+      "step": 4691
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.247396170005904e-06,
+      "loss": 0.9378,
+      "step": 4692
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24702892160214e-06,
+      "loss": 0.9922,
+      "step": 4693
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.246661590913153e-06,
+      "loss": 1.2164,
+      "step": 4694
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.246294177946062e-06,
+      "loss": 0.8587,
+      "step": 4695
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.245926682707985e-06,
+      "loss": 0.9571,
+      "step": 4696
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24555910520604e-06,
+      "loss": 1.0757,
+      "step": 4697
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24519144544735e-06,
+      "loss": 0.8673,
+      "step": 4698
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.244823703439041e-06,
+      "loss": 0.9929,
+      "step": 4699
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.244455879188235e-06,
+      "loss": 0.9607,
+      "step": 4700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.244087972702058e-06,
+      "loss": 0.8886,
+      "step": 4701
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24371998398764e-06,
+      "loss": 0.9439,
+      "step": 4702
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.243351913052111e-06,
+      "loss": 0.8787,
+      "step": 4703
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2429837599026e-06,
+      "loss": 0.9751,
+      "step": 4704
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.242615524546242e-06,
+      "loss": 1.0869,
+      "step": 4705
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24224720699017e-06,
+      "loss": 1.0997,
+      "step": 4706
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24187880724152e-06,
+      "loss": 1.0588,
+      "step": 4707
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.241510325307432e-06,
+      "loss": 0.8658,
+      "step": 4708
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.241141761195043e-06,
+      "loss": 0.9478,
+      "step": 4709
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.240773114911494e-06,
+      "loss": 0.959,
+      "step": 4710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.240404386463925e-06,
+      "loss": 1.044,
+      "step": 4711
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.240035575859486e-06,
+      "loss": 1.0284,
+      "step": 4712
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.23966668310532e-06,
+      "loss": 1.0178,
+      "step": 4713
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.239297708208574e-06,
+      "loss": 0.9337,
+      "step": 4714
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.238928651176396e-06,
+      "loss": 1.0182,
+      "step": 4715
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.238559512015936e-06,
+      "loss": 1.1474,
+      "step": 4716
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.238190290734347e-06,
+      "loss": 0.9582,
+      "step": 4717
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.237820987338784e-06,
+      "loss": 0.9955,
+      "step": 4718
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.237451601836398e-06,
+      "loss": 1.135,
+      "step": 4719
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.23708213423435e-06,
+      "loss": 0.9833,
+      "step": 4720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.236712584539797e-06,
+      "loss": 1.048,
+      "step": 4721
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2363429527599e-06,
+      "loss": 0.8024,
+      "step": 4722
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.235973238901816e-06,
+      "loss": 0.8864,
+      "step": 4723
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.235603442972714e-06,
+      "loss": 1.011,
+      "step": 4724
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.235233564979756e-06,
+      "loss": 0.8113,
+      "step": 4725
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.234863604930107e-06,
+      "loss": 0.9227,
+      "step": 4726
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.234493562830937e-06,
+      "loss": 1.1752,
+      "step": 4727
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.234123438689414e-06,
+      "loss": 1.0006,
+      "step": 4728
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.23375323251271e-06,
+      "loss": 1.0694,
+      "step": 4729
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.233382944307998e-06,
+      "loss": 0.8967,
+      "step": 4730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.233012574082452e-06,
+      "loss": 1.0893,
+      "step": 4731
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.232642121843247e-06,
+      "loss": 0.9353,
+      "step": 4732
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.232271587597562e-06,
+      "loss": 0.8304,
+      "step": 4733
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.231900971352575e-06,
+      "loss": 0.8116,
+      "step": 4734
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.231530273115465e-06,
+      "loss": 0.911,
+      "step": 4735
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.231159492893417e-06,
+      "loss": 1.187,
+      "step": 4736
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.230788630693614e-06,
+      "loss": 0.9824,
+      "step": 4737
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.230417686523242e-06,
+      "loss": 0.995,
+      "step": 4738
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.230046660389485e-06,
+      "loss": 0.9427,
+      "step": 4739
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.229675552299533e-06,
+      "loss": 0.8392,
+      "step": 4740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.229304362260576e-06,
+      "loss": 1.1268,
+      "step": 4741
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.22893309027981e-06,
+      "loss": 1.1573,
+      "step": 4742
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.228561736364422e-06,
+      "loss": 1.0278,
+      "step": 4743
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.22819030052161e-06,
+      "loss": 1.0701,
+      "step": 4744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.227818782758569e-06,
+      "loss": 1.0463,
+      "step": 4745
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2274471830825e-06,
+      "loss": 0.8363,
+      "step": 4746
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2270755015006e-06,
+      "loss": 0.973,
+      "step": 4747
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.22670373802007e-06,
+      "loss": 1.0271,
+      "step": 4748
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.226331892648114e-06,
+      "loss": 0.9709,
+      "step": 4749
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.225959965391938e-06,
+      "loss": 1.0385,
+      "step": 4750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.225587956258745e-06,
+      "loss": 0.9113,
+      "step": 4751
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.225215865255745e-06,
+      "loss": 1.1325,
+      "step": 4752
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.224843692390147e-06,
+      "loss": 0.852,
+      "step": 4753
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.224471437669158e-06,
+      "loss": 1.044,
+      "step": 4754
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.224099101099993e-06,
+      "loss": 0.8342,
+      "step": 4755
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.223726682689869e-06,
+      "loss": 0.8459,
+      "step": 4756
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.223354182445996e-06,
+      "loss": 0.9815,
+      "step": 4757
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.222981600375595e-06,
+      "loss": 1.0132,
+      "step": 4758
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.222608936485883e-06,
+      "loss": 1.0019,
+      "step": 4759
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.222236190784081e-06,
+      "loss": 1.0216,
+      "step": 4760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.221863363277411e-06,
+      "loss": 0.8782,
+      "step": 4761
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.221490453973094e-06,
+      "loss": 0.8678,
+      "step": 4762
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.22111746287836e-06,
+      "loss": 0.8755,
+      "step": 4763
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.220744390000432e-06,
+      "loss": 1.096,
+      "step": 4764
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.220371235346537e-06,
+      "loss": 0.9147,
+      "step": 4765
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.219997998923907e-06,
+      "loss": 1.1778,
+      "step": 4766
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.219624680739774e-06,
+      "loss": 0.9453,
+      "step": 4767
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.21925128080137e-06,
+      "loss": 0.918,
+      "step": 4768
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.218877799115929e-06,
+      "loss": 0.8888,
+      "step": 4769
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.218504235690688e-06,
+      "loss": 0.6911,
+      "step": 4770
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.218130590532885e-06,
+      "loss": 1.1576,
+      "step": 4771
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.217756863649756e-06,
+      "loss": 0.8599,
+      "step": 4772
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.217383055048547e-06,
+      "loss": 1.1065,
+      "step": 4773
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.217009164736497e-06,
+      "loss": 0.968,
+      "step": 4774
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.216635192720853e-06,
+      "loss": 1.0996,
+      "step": 4775
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.216261139008856e-06,
+      "loss": 1.0398,
+      "step": 4776
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.215887003607758e-06,
+      "loss": 0.9562,
+      "step": 4777
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.215512786524803e-06,
+      "loss": 1.0527,
+      "step": 4778
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.215138487767246e-06,
+      "loss": 0.9204,
+      "step": 4779
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.214764107342336e-06,
+      "loss": 0.9731,
+      "step": 4780
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.214389645257328e-06,
+      "loss": 1.0189,
+      "step": 4781
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.214015101519477e-06,
+      "loss": 0.9803,
+      "step": 4782
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.21364047613604e-06,
+      "loss": 1.1897,
+      "step": 4783
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.213265769114273e-06,
+      "loss": 0.9997,
+      "step": 4784
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.212890980461437e-06,
+      "loss": 1.2078,
+      "step": 4785
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.212516110184794e-06,
+      "loss": 1.0851,
+      "step": 4786
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.212141158291608e-06,
+      "loss": 1.0478,
+      "step": 4787
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.211766124789143e-06,
+      "loss": 0.8683,
+      "step": 4788
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.211391009684663e-06,
+      "loss": 0.9251,
+      "step": 4789
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.21101581298544e-06,
+      "loss": 0.9122,
+      "step": 4790
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.210640534698739e-06,
+      "loss": 0.9299,
+      "step": 4791
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.210265174831832e-06,
+      "loss": 1.1131,
+      "step": 4792
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.209889733391993e-06,
+      "loss": 0.9201,
+      "step": 4793
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.209514210386496e-06,
+      "loss": 1.0593,
+      "step": 4794
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.209138605822616e-06,
+      "loss": 0.9797,
+      "step": 4795
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20876291970763e-06,
+      "loss": 1.0436,
+      "step": 4796
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.208387152048817e-06,
+      "loss": 0.9565,
+      "step": 4797
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.208011302853458e-06,
+      "loss": 0.9036,
+      "step": 4798
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.207635372128833e-06,
+      "loss": 1.2451,
+      "step": 4799
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.207259359882231e-06,
+      "loss": 0.9234,
+      "step": 4800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20688326612093e-06,
+      "loss": 1.0307,
+      "step": 4801
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.206507090852218e-06,
+      "loss": 1.0892,
+      "step": 4802
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20613083408339e-06,
+      "loss": 0.8277,
+      "step": 4803
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.205754495821728e-06,
+      "loss": 1.0863,
+      "step": 4804
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20537807607453e-06,
+      "loss": 1.0833,
+      "step": 4805
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.205001574849081e-06,
+      "loss": 0.9718,
+      "step": 4806
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.204624992152684e-06,
+      "loss": 1.0646,
+      "step": 4807
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.204248327992632e-06,
+      "loss": 0.7639,
+      "step": 4808
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20387158237622e-06,
+      "loss": 0.9851,
+      "step": 4809
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.203494755310749e-06,
+      "loss": 0.9537,
+      "step": 4810
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.203117846803523e-06,
+      "loss": 0.9491,
+      "step": 4811
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.202740856861842e-06,
+      "loss": 0.9986,
+      "step": 4812
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20236378549301e-06,
+      "loss": 0.9461,
+      "step": 4813
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.201986632704332e-06,
+      "loss": 1.0775,
+      "step": 4814
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.201609398503114e-06,
+      "loss": 0.9731,
+      "step": 4815
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.201232082896672e-06,
+      "loss": 0.7696,
+      "step": 4816
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20085468589231e-06,
+      "loss": 0.9233,
+      "step": 4817
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.200477207497337e-06,
+      "loss": 0.998,
+      "step": 4818
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.200099647719077e-06,
+      "loss": 1.0001,
+      "step": 4819
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.199722006564835e-06,
+      "loss": 0.8435,
+      "step": 4820
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.199344284041934e-06,
+      "loss": 0.9943,
+      "step": 4821
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.198966480157689e-06,
+      "loss": 0.8869,
+      "step": 4822
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19858859491942e-06,
+      "loss": 1.0234,
+      "step": 4823
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.198210628334451e-06,
+      "loss": 1.0951,
+      "step": 4824
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.197832580410102e-06,
+      "loss": 0.8881,
+      "step": 4825
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1974544511537e-06,
+      "loss": 0.9845,
+      "step": 4826
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.197076240572571e-06,
+      "loss": 0.854,
+      "step": 4827
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.196697948674041e-06,
+      "loss": 0.9295,
+      "step": 4828
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19631957546544e-06,
+      "loss": 1.0493,
+      "step": 4829
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.195941120954097e-06,
+      "loss": 1.0123,
+      "step": 4830
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.195562585147351e-06,
+      "loss": 0.8425,
+      "step": 4831
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.195183968052529e-06,
+      "loss": 1.0243,
+      "step": 4832
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19480526967697e-06,
+      "loss": 1.0058,
+      "step": 4833
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19442649002801e-06,
+      "loss": 1.0593,
+      "step": 4834
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.194047629112987e-06,
+      "loss": 0.9246,
+      "step": 4835
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.193668686939245e-06,
+      "loss": 1.0385,
+      "step": 4836
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.193289663514122e-06,
+      "loss": 0.9005,
+      "step": 4837
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.192910558844963e-06,
+      "loss": 0.9221,
+      "step": 4838
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.192531372939113e-06,
+      "loss": 1.0598,
+      "step": 4839
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.192152105803919e-06,
+      "loss": 0.9525,
+      "step": 4840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.191772757446729e-06,
+      "loss": 1.0551,
+      "step": 4841
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.191393327874894e-06,
+      "loss": 1.018,
+      "step": 4842
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.191013817095762e-06,
+      "loss": 0.8873,
+      "step": 4843
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.190634225116688e-06,
+      "loss": 1.0438,
+      "step": 4844
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.190254551945029e-06,
+      "loss": 1.0536,
+      "step": 4845
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.189874797588135e-06,
+      "loss": 0.8378,
+      "step": 4846
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18949496205337e-06,
+      "loss": 1.0442,
+      "step": 4847
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.189115045348088e-06,
+      "loss": 1.022,
+      "step": 4848
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.188735047479654e-06,
+      "loss": 0.9243,
+      "step": 4849
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.188354968455428e-06,
+      "loss": 1.1833,
+      "step": 4850
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.187974808282775e-06,
+      "loss": 1.0584,
+      "step": 4851
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.187594566969059e-06,
+      "loss": 1.1042,
+      "step": 4852
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18721424452165e-06,
+      "loss": 0.9087,
+      "step": 4853
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.186833840947914e-06,
+      "loss": 0.972,
+      "step": 4854
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.186453356255221e-06,
+      "loss": 0.9259,
+      "step": 4855
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.186072790450943e-06,
+      "loss": 0.9259,
+      "step": 4856
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.185692143542458e-06,
+      "loss": 1.2123,
+      "step": 4857
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.185311415537135e-06,
+      "loss": 0.9475,
+      "step": 4858
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.184930606442352e-06,
+      "loss": 1.0607,
+      "step": 4859
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18454971626549e-06,
+      "loss": 1.1036,
+      "step": 4860
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.184168745013926e-06,
+      "loss": 0.8531,
+      "step": 4861
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.183787692695038e-06,
+      "loss": 0.8557,
+      "step": 4862
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.183406559316217e-06,
+      "loss": 1.0011,
+      "step": 4863
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.183025344884842e-06,
+      "loss": 1.1496,
+      "step": 4864
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.182644049408299e-06,
+      "loss": 1.0621,
+      "step": 4865
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.182262672893976e-06,
+      "loss": 1.0751,
+      "step": 4866
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.181881215349262e-06,
+      "loss": 0.9293,
+      "step": 4867
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.181499676781548e-06,
+      "loss": 1.0275,
+      "step": 4868
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.181118057198228e-06,
+      "loss": 0.8641,
+      "step": 4869
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.180736356606693e-06,
+      "loss": 1.0882,
+      "step": 4870
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.180354575014339e-06,
+      "loss": 0.9634,
+      "step": 4871
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.179972712428563e-06,
+      "loss": 1.0469,
+      "step": 4872
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.179590768856767e-06,
+      "loss": 0.836,
+      "step": 4873
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.179208744306345e-06,
+      "loss": 0.9553,
+      "step": 4874
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.178826638784704e-06,
+      "loss": 1.0602,
+      "step": 4875
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.178444452299241e-06,
+      "loss": 0.9198,
+      "step": 4876
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.178062184857369e-06,
+      "loss": 0.9596,
+      "step": 4877
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.177679836466487e-06,
+      "loss": 0.8328,
+      "step": 4878
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.177297407134005e-06,
+      "loss": 0.9125,
+      "step": 4879
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.176914896867335e-06,
+      "loss": 0.9369,
+      "step": 4880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.176532305673886e-06,
+      "loss": 0.8702,
+      "step": 4881
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17614963356107e-06,
+      "loss": 0.8556,
+      "step": 4882
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.175766880536305e-06,
+      "loss": 1.043,
+      "step": 4883
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.175384046607004e-06,
+      "loss": 0.9987,
+      "step": 4884
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.175001131780582e-06,
+      "loss": 1.0376,
+      "step": 4885
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17461813606446e-06,
+      "loss": 1.1598,
+      "step": 4886
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.174235059466058e-06,
+      "loss": 1.0588,
+      "step": 4887
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1738519019928e-06,
+      "loss": 0.7815,
+      "step": 4888
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.173468663652107e-06,
+      "loss": 0.8376,
+      "step": 4889
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.173085344451407e-06,
+      "loss": 0.9585,
+      "step": 4890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.172701944398125e-06,
+      "loss": 0.9446,
+      "step": 4891
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.172318463499688e-06,
+      "loss": 1.0809,
+      "step": 4892
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.171934901763528e-06,
+      "loss": 0.9782,
+      "step": 4893
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.171551259197076e-06,
+      "loss": 1.0583,
+      "step": 4894
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.171167535807763e-06,
+      "loss": 0.888,
+      "step": 4895
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.170783731603028e-06,
+      "loss": 0.9957,
+      "step": 4896
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.170399846590301e-06,
+      "loss": 1.1251,
+      "step": 4897
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.170015880777025e-06,
+      "loss": 1.094,
+      "step": 4898
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.169631834170638e-06,
+      "loss": 1.083,
+      "step": 4899
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.169247706778578e-06,
+      "loss": 0.912,
+      "step": 4900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16886349860829e-06,
+      "loss": 0.9186,
+      "step": 4901
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.168479209667219e-06,
+      "loss": 1.1092,
+      "step": 4902
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.168094839962807e-06,
+      "loss": 0.9751,
+      "step": 4903
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.167710389502503e-06,
+      "loss": 1.0456,
+      "step": 4904
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.167325858293757e-06,
+      "loss": 0.8883,
+      "step": 4905
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.166941246344016e-06,
+      "loss": 0.9317,
+      "step": 4906
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.166556553660736e-06,
+      "loss": 1.0288,
+      "step": 4907
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.166171780251365e-06,
+      "loss": 0.943,
+      "step": 4908
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16578692612336e-06,
+      "loss": 0.9864,
+      "step": 4909
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.165401991284183e-06,
+      "loss": 0.848,
+      "step": 4910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.165016975741284e-06,
+      "loss": 0.8765,
+      "step": 4911
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.164631879502126e-06,
+      "loss": 1.0425,
+      "step": 4912
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16424670257417e-06,
+      "loss": 1.0733,
+      "step": 4913
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16386144496488e-06,
+      "loss": 1.0255,
+      "step": 4914
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.163476106681718e-06,
+      "loss": 0.8507,
+      "step": 4915
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.163090687732152e-06,
+      "loss": 1.0577,
+      "step": 4916
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.162705188123647e-06,
+      "loss": 1.0427,
+      "step": 4917
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.162319607863675e-06,
+      "loss": 1.0034,
+      "step": 4918
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.161933946959702e-06,
+      "loss": 1.0272,
+      "step": 4919
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.161548205419205e-06,
+      "loss": 0.9385,
+      "step": 4920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.161162383249656e-06,
+      "loss": 0.9171,
+      "step": 4921
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16077648045853e-06,
+      "loss": 0.9646,
+      "step": 4922
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1603904970533e-06,
+      "loss": 0.9627,
+      "step": 4923
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16000443304145e-06,
+      "loss": 0.9685,
+      "step": 4924
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.159618288430459e-06,
+      "loss": 0.8398,
+      "step": 4925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.159232063227806e-06,
+      "loss": 0.8306,
+      "step": 4926
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.158845757440974e-06,
+      "loss": 1.102,
+      "step": 4927
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15845937107745e-06,
+      "loss": 1.0796,
+      "step": 4928
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15807290414472e-06,
+      "loss": 0.9274,
+      "step": 4929
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.157686356650269e-06,
+      "loss": 0.8649,
+      "step": 4930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.157299728601588e-06,
+      "loss": 1.206,
+      "step": 4931
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.156913020006169e-06,
+      "loss": 1.0615,
+      "step": 4932
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.156526230871504e-06,
+      "loss": 0.891,
+      "step": 4933
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.156139361205083e-06,
+      "loss": 0.9786,
+      "step": 4934
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.155752411014406e-06,
+      "loss": 0.9715,
+      "step": 4935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.155365380306971e-06,
+      "loss": 1.0464,
+      "step": 4936
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.154978269090271e-06,
+      "loss": 0.7917,
+      "step": 4937
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.154591077371812e-06,
+      "loss": 1.0532,
+      "step": 4938
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.154203805159092e-06,
+      "loss": 1.0484,
+      "step": 4939
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.153816452459616e-06,
+      "loss": 0.8994,
+      "step": 4940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.153429019280888e-06,
+      "loss": 1.0631,
+      "step": 4941
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.153041505630415e-06,
+      "loss": 1.1048,
+      "step": 4942
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.152653911515705e-06,
+      "loss": 0.9606,
+      "step": 4943
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15226623694427e-06,
+      "loss": 0.6816,
+      "step": 4944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.151878481923615e-06,
+      "loss": 0.9654,
+      "step": 4945
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.151490646461257e-06,
+      "loss": 0.81,
+      "step": 4946
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15110273056471e-06,
+      "loss": 0.8617,
+      "step": 4947
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15071473424149e-06,
+      "loss": 1.0416,
+      "step": 4948
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.150326657499112e-06,
+      "loss": 0.9548,
+      "step": 4949
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.149938500345097e-06,
+      "loss": 0.8189,
+      "step": 4950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.149550262786965e-06,
+      "loss": 0.7193,
+      "step": 4951
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.149161944832239e-06,
+      "loss": 1.036,
+      "step": 4952
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.148773546488441e-06,
+      "loss": 0.9453,
+      "step": 4953
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.148385067763094e-06,
+      "loss": 0.9696,
+      "step": 4954
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14799650866373e-06,
+      "loss": 0.9739,
+      "step": 4955
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.147607869197874e-06,
+      "loss": 0.8714,
+      "step": 4956
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.147219149373057e-06,
+      "loss": 0.8842,
+      "step": 4957
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.146830349196809e-06,
+      "loss": 0.9464,
+      "step": 4958
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.146441468676664e-06,
+      "loss": 1.0255,
+      "step": 4959
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.146052507820156e-06,
+      "loss": 1.1571,
+      "step": 4960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.145663466634821e-06,
+      "loss": 1.0995,
+      "step": 4961
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.145274345128197e-06,
+      "loss": 1.0347,
+      "step": 4962
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.144885143307822e-06,
+      "loss": 0.9675,
+      "step": 4963
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.144495861181238e-06,
+      "loss": 0.9496,
+      "step": 4964
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.144106498755987e-06,
+      "loss": 0.9692,
+      "step": 4965
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.143717056039613e-06,
+      "loss": 0.9184,
+      "step": 4966
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.143327533039661e-06,
+      "loss": 0.983,
+      "step": 4967
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.142937929763678e-06,
+      "loss": 0.9756,
+      "step": 4968
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.142548246219212e-06,
+      "loss": 0.9033,
+      "step": 4969
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.142158482413812e-06,
+      "loss": 1.0208,
+      "step": 4970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.141768638355032e-06,
+      "loss": 0.8773,
+      "step": 4971
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.141378714050426e-06,
+      "loss": 0.8398,
+      "step": 4972
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.140988709507546e-06,
+      "loss": 1.0216,
+      "step": 4973
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.140598624733948e-06,
+      "loss": 0.8855,
+      "step": 4974
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14020845973719e-06,
+      "loss": 0.9652,
+      "step": 4975
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.139818214524834e-06,
+      "loss": 0.995,
+      "step": 4976
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13942788910444e-06,
+      "loss": 0.8036,
+      "step": 4977
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.139037483483567e-06,
+      "loss": 0.9214,
+      "step": 4978
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.138646997669783e-06,
+      "loss": 0.9833,
+      "step": 4979
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.138256431670653e-06,
+      "loss": 0.9952,
+      "step": 4980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.137865785493742e-06,
+      "loss": 0.9356,
+      "step": 4981
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.137475059146622e-06,
+      "loss": 0.9366,
+      "step": 4982
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.137084252636859e-06,
+      "loss": 0.926,
+      "step": 4983
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.136693365972028e-06,
+      "loss": 0.9076,
+      "step": 4984
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1363023991597e-06,
+      "loss": 0.7984,
+      "step": 4985
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.135911352207453e-06,
+      "loss": 1.0583,
+      "step": 4986
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.135520225122862e-06,
+      "loss": 0.7563,
+      "step": 4987
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.135129017913504e-06,
+      "loss": 0.8288,
+      "step": 4988
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13473773058696e-06,
+      "loss": 0.9807,
+      "step": 4989
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13434636315081e-06,
+      "loss": 0.8591,
+      "step": 4990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.133954915612635e-06,
+      "loss": 1.0057,
+      "step": 4991
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.133563387980023e-06,
+      "loss": 0.8455,
+      "step": 4992
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.133171780260559e-06,
+      "loss": 0.9847,
+      "step": 4993
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.132780092461829e-06,
+      "loss": 1.2232,
+      "step": 4994
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.132388324591421e-06,
+      "loss": 1.1013,
+      "step": 4995
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.131996476656928e-06,
+      "loss": 1.0246,
+      "step": 4996
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.131604548665937e-06,
+      "loss": 0.9254,
+      "step": 4997
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.131212540626049e-06,
+      "loss": 0.7955,
+      "step": 4998
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.130820452544854e-06,
+      "loss": 0.9873,
+      "step": 4999
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13042828442995e-06,
+      "loss": 1.0461,
+      "step": 5000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.130036036288934e-06,
+      "loss": 0.9302,
+      "step": 5001
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.129643708129408e-06,
+      "loss": 0.9268,
+      "step": 5002
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12925129995897e-06,
+      "loss": 1.0004,
+      "step": 5003
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.128858811785226e-06,
+      "loss": 0.8922,
+      "step": 5004
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.128466243615779e-06,
+      "loss": 0.9413,
+      "step": 5005
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.128073595458235e-06,
+      "loss": 0.9284,
+      "step": 5006
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.1276808673202e-06,
+      "loss": 0.9016,
+      "step": 5007
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.127288059209287e-06,
+      "loss": 1.0431,
+      "step": 5008
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.126895171133104e-06,
+      "loss": 0.7023,
+      "step": 5009
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.126502203099262e-06,
+      "loss": 0.975,
+      "step": 5010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.126109155115377e-06,
+      "loss": 0.8979,
+      "step": 5011
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.125716027189062e-06,
+      "loss": 1.062,
+      "step": 5012
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.125322819327935e-06,
+      "loss": 1.1394,
+      "step": 5013
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.124929531539614e-06,
+      "loss": 1.0847,
+      "step": 5014
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.124536163831719e-06,
+      "loss": 1.1467,
+      "step": 5015
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.124142716211873e-06,
+      "loss": 1.1009,
+      "step": 5016
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.123749188687693e-06,
+      "loss": 1.1583,
+      "step": 5017
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.123355581266812e-06,
+      "loss": 0.9038,
+      "step": 5018
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.122961893956852e-06,
+      "loss": 0.8212,
+      "step": 5019
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.122568126765439e-06,
+      "loss": 1.0631,
+      "step": 5020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.122174279700204e-06,
+      "loss": 0.992,
+      "step": 5021
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.121780352768777e-06,
+      "loss": 0.9657,
+      "step": 5022
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.121386345978792e-06,
+      "loss": 1.0705,
+      "step": 5023
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12099225933788e-06,
+      "loss": 1.1708,
+      "step": 5024
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.120598092853678e-06,
+      "loss": 1.0536,
+      "step": 5025
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.120203846533824e-06,
+      "loss": 0.9537,
+      "step": 5026
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.119809520385953e-06,
+      "loss": 1.0835,
+      "step": 5027
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.119415114417709e-06,
+      "loss": 1.1499,
+      "step": 5028
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.119020628636732e-06,
+      "loss": 0.9092,
+      "step": 5029
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.118626063050661e-06,
+      "loss": 0.8759,
+      "step": 5030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.118231417667147e-06,
+      "loss": 0.9167,
+      "step": 5031
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.117836692493834e-06,
+      "loss": 0.9918,
+      "step": 5032
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.117441887538368e-06,
+      "loss": 1.1017,
+      "step": 5033
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.1170470028084e-06,
+      "loss": 1.1,
+      "step": 5034
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.11665203831158e-06,
+      "loss": 0.9002,
+      "step": 5035
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.116256994055562e-06,
+      "loss": 1.0137,
+      "step": 5036
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.115861870047995e-06,
+      "loss": 1.0383,
+      "step": 5037
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.115466666296541e-06,
+      "loss": 1.0905,
+      "step": 5038
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.115071382808852e-06,
+      "loss": 0.949,
+      "step": 5039
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.114676019592589e-06,
+      "loss": 1.1543,
+      "step": 5040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.114280576655413e-06,
+      "loss": 0.8696,
+      "step": 5041
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.113885054004983e-06,
+      "loss": 0.7446,
+      "step": 5042
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.113489451648962e-06,
+      "loss": 1.1595,
+      "step": 5043
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.113093769595018e-06,
+      "loss": 1.0532,
+      "step": 5044
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.112698007850812e-06,
+      "loss": 1.1219,
+      "step": 5045
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.112302166424018e-06,
+      "loss": 0.8604,
+      "step": 5046
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.111906245322303e-06,
+      "loss": 1.1001,
+      "step": 5047
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.111510244553335e-06,
+      "loss": 0.9618,
+      "step": 5048
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.11111416412479e-06,
+      "loss": 0.8561,
+      "step": 5049
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.110718004044342e-06,
+      "loss": 0.9804,
+      "step": 5050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.11032176431966e-06,
+      "loss": 1.0056,
+      "step": 5051
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.109925444958432e-06,
+      "loss": 0.8469,
+      "step": 5052
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.109529045968326e-06,
+      "loss": 1.1136,
+      "step": 5053
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.10913256735703e-06,
+      "loss": 0.8809,
+      "step": 5054
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.108736009132221e-06,
+      "loss": 1.0905,
+      "step": 5055
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.108339371301586e-06,
+      "loss": 0.995,
+      "step": 5056
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.107942653872806e-06,
+      "loss": 0.9294,
+      "step": 5057
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.10754585685357e-06,
+      "loss": 0.894,
+      "step": 5058
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.107148980251563e-06,
+      "loss": 0.9982,
+      "step": 5059
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.106752024074478e-06,
+      "loss": 0.8654,
+      "step": 5060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.106354988330004e-06,
+      "loss": 0.8599,
+      "step": 5061
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.105957873025832e-06,
+      "loss": 0.8178,
+      "step": 5062
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.105560678169659e-06,
+      "loss": 0.8946,
+      "step": 5063
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.105163403769177e-06,
+      "loss": 1.1026,
+      "step": 5064
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.104766049832088e-06,
+      "loss": 0.9578,
+      "step": 5065
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.104368616366085e-06,
+      "loss": 1.0061,
+      "step": 5066
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.103971103378874e-06,
+      "loss": 0.969,
+      "step": 5067
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.103573510878152e-06,
+      "loss": 0.9858,
+      "step": 5068
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.103175838871625e-06,
+      "loss": 0.8614,
+      "step": 5069
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.102778087366997e-06,
+      "loss": 0.9553,
+      "step": 5070
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.102380256371973e-06,
+      "loss": 1.0659,
+      "step": 5071
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.101982345894264e-06,
+      "loss": 0.9994,
+      "step": 5072
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.101584355941575e-06,
+      "loss": 1.0474,
+      "step": 5073
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.101186286521622e-06,
+      "loss": 1.0214,
+      "step": 5074
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.100788137642113e-06,
+      "loss": 0.8327,
+      "step": 5075
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.100389909310767e-06,
+      "loss": 1.0055,
+      "step": 5076
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.099991601535294e-06,
+      "loss": 0.9524,
+      "step": 5077
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.099593214323414e-06,
+      "loss": 0.8802,
+      "step": 5078
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.099194747682846e-06,
+      "loss": 1.0122,
+      "step": 5079
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09879620162131e-06,
+      "loss": 0.8677,
+      "step": 5080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.098397576146525e-06,
+      "loss": 0.8611,
+      "step": 5081
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.097998871266219e-06,
+      "loss": 0.9646,
+      "step": 5082
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.097600086988114e-06,
+      "loss": 0.9675,
+      "step": 5083
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.097201223319936e-06,
+      "loss": 0.9062,
+      "step": 5084
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.096802280269414e-06,
+      "loss": 1.1449,
+      "step": 5085
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.096403257844278e-06,
+      "loss": 1.0468,
+      "step": 5086
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.096004156052258e-06,
+      "loss": 0.9277,
+      "step": 5087
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.095604974901087e-06,
+      "loss": 1.1794,
+      "step": 5088
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.095205714398497e-06,
+      "loss": 0.944,
+      "step": 5089
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.094806374552226e-06,
+      "loss": 0.9449,
+      "step": 5090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09440695537001e-06,
+      "loss": 1.0739,
+      "step": 5091
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.094007456859589e-06,
+      "loss": 0.9464,
+      "step": 5092
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0936078790287e-06,
+      "loss": 0.9851,
+      "step": 5093
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09320822188509e-06,
+      "loss": 0.9535,
+      "step": 5094
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.092808485436497e-06,
+      "loss": 0.8783,
+      "step": 5095
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.092408669690669e-06,
+      "loss": 0.9238,
+      "step": 5096
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.092008774655351e-06,
+      "loss": 0.9089,
+      "step": 5097
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.091608800338292e-06,
+      "loss": 1.0216,
+      "step": 5098
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09120874674724e-06,
+      "loss": 1.0355,
+      "step": 5099
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.090808613889945e-06,
+      "loss": 0.8455,
+      "step": 5100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.090408401774163e-06,
+      "loss": 0.9827,
+      "step": 5101
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.090008110407646e-06,
+      "loss": 0.9823,
+      "step": 5102
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08960773979815e-06,
+      "loss": 1.1867,
+      "step": 5103
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08920728995343e-06,
+      "loss": 1.1309,
+      "step": 5104
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.088806760881248e-06,
+      "loss": 0.9108,
+      "step": 5105
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.088406152589363e-06,
+      "loss": 1.0484,
+      "step": 5106
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.088005465085534e-06,
+      "loss": 0.8707,
+      "step": 5107
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.087604698377528e-06,
+      "loss": 0.8199,
+      "step": 5108
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.087203852473106e-06,
+      "loss": 1.0362,
+      "step": 5109
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08680292738004e-06,
+      "loss": 0.8502,
+      "step": 5110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.086401923106092e-06,
+      "loss": 0.96,
+      "step": 5111
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.086000839659033e-06,
+      "loss": 1.0893,
+      "step": 5112
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.085599677046637e-06,
+      "loss": 1.0354,
+      "step": 5113
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.085198435276673e-06,
+      "loss": 0.9315,
+      "step": 5114
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.084797114356915e-06,
+      "loss": 1.0267,
+      "step": 5115
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08439571429514e-06,
+      "loss": 0.9097,
+      "step": 5116
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.083994235099126e-06,
+      "loss": 0.7755,
+      "step": 5117
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.083592676776647e-06,
+      "loss": 0.9327,
+      "step": 5118
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.083191039335488e-06,
+      "loss": 0.9638,
+      "step": 5119
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08278932278343e-06,
+      "loss": 1.0127,
+      "step": 5120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.082387527128253e-06,
+      "loss": 0.9976,
+      "step": 5121
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.081985652377745e-06,
+      "loss": 0.8674,
+      "step": 5122
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08158369853969e-06,
+      "loss": 0.9857,
+      "step": 5123
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.081181665621878e-06,
+      "loss": 0.9351,
+      "step": 5124
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.080779553632097e-06,
+      "loss": 1.2443,
+      "step": 5125
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.080377362578137e-06,
+      "loss": 0.9625,
+      "step": 5126
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.07997509246779e-06,
+      "loss": 0.8802,
+      "step": 5127
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.079572743308854e-06,
+      "loss": 0.9498,
+      "step": 5128
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.079170315109118e-06,
+      "loss": 1.2198,
+      "step": 5129
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.078767807876386e-06,
+      "loss": 1.1109,
+      "step": 5130
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.078365221618451e-06,
+      "loss": 0.8905,
+      "step": 5131
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.077962556343115e-06,
+      "loss": 0.9552,
+      "step": 5132
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.077559812058178e-06,
+      "loss": 1.0349,
+      "step": 5133
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.077156988771447e-06,
+      "loss": 0.9333,
+      "step": 5134
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.076754086490724e-06,
+      "loss": 0.9624,
+      "step": 5135
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.076351105223816e-06,
+      "loss": 0.877,
+      "step": 5136
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.075948044978529e-06,
+      "loss": 0.9426,
+      "step": 5137
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.075544905762671e-06,
+      "loss": 1.0876,
+      "step": 5138
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.075141687584056e-06,
+      "loss": 1.004,
+      "step": 5139
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.074738390450498e-06,
+      "loss": 1.0222,
+      "step": 5140
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.074335014369807e-06,
+      "loss": 0.9449,
+      "step": 5141
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.073931559349799e-06,
+      "loss": 0.9766,
+      "step": 5142
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.073528025398288e-06,
+      "loss": 0.9832,
+      "step": 5143
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0731244125231e-06,
+      "loss": 0.8811,
+      "step": 5144
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.072720720732048e-06,
+      "loss": 1.1166,
+      "step": 5145
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.072316950032955e-06,
+      "loss": 0.9517,
+      "step": 5146
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.071913100433646e-06,
+      "loss": 1.0064,
+      "step": 5147
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.071509171941946e-06,
+      "loss": 0.8006,
+      "step": 5148
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.071105164565678e-06,
+      "loss": 1.1434,
+      "step": 5149
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.070701078312672e-06,
+      "loss": 0.9414,
+      "step": 5150
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.070296913190755e-06,
+      "loss": 0.875,
+      "step": 5151
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.069892669207757e-06,
+      "loss": 1.067,
+      "step": 5152
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.069488346371514e-06,
+      "loss": 0.8838,
+      "step": 5153
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.069083944689857e-06,
+      "loss": 0.9669,
+      "step": 5154
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.06867946417062e-06,
+      "loss": 1.061,
+      "step": 5155
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.068274904821643e-06,
+      "loss": 0.9675,
+      "step": 5156
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.067870266650764e-06,
+      "loss": 1.181,
+      "step": 5157
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.067465549665819e-06,
+      "loss": 1.1221,
+      "step": 5158
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.067060753874653e-06,
+      "loss": 0.926,
+      "step": 5159
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.066655879285105e-06,
+      "loss": 0.8899,
+      "step": 5160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.066250925905023e-06,
+      "loss": 1.0806,
+      "step": 5161
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.065845893742254e-06,
+      "loss": 1.0285,
+      "step": 5162
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.06544078280464e-06,
+      "loss": 1.0894,
+      "step": 5163
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.065035593100033e-06,
+      "loss": 0.9143,
+      "step": 5164
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.064630324636285e-06,
+      "loss": 0.9067,
+      "step": 5165
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.064224977421248e-06,
+      "loss": 1.0368,
+      "step": 5166
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.063819551462772e-06,
+      "loss": 0.9871,
+      "step": 5167
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.063414046768714e-06,
+      "loss": 0.914,
+      "step": 5168
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.06300846334693e-06,
+      "loss": 0.9711,
+      "step": 5169
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.06260280120528e-06,
+      "loss": 1.0156,
+      "step": 5170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.062197060351622e-06,
+      "loss": 1.0158,
+      "step": 5171
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.061791240793817e-06,
+      "loss": 0.9547,
+      "step": 5172
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.061385342539727e-06,
+      "loss": 1.1277,
+      "step": 5173
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.06097936559722e-06,
+      "loss": 0.9183,
+      "step": 5174
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.060573309974157e-06,
+      "loss": 1.0211,
+      "step": 5175
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.060167175678407e-06,
+      "loss": 0.9911,
+      "step": 5176
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05976096271784e-06,
+      "loss": 0.8415,
+      "step": 5177
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.059354671100324e-06,
+      "loss": 0.8902,
+      "step": 5178
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.058948300833735e-06,
+      "loss": 1.0315,
+      "step": 5179
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05854185192594e-06,
+      "loss": 0.8257,
+      "step": 5180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05813532438482e-06,
+      "loss": 1.0056,
+      "step": 5181
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.057728718218246e-06,
+      "loss": 1.1059,
+      "step": 5182
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.057322033434099e-06,
+      "loss": 0.9517,
+      "step": 5183
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.056915270040259e-06,
+      "loss": 0.843,
+      "step": 5184
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.056508428044606e-06,
+      "loss": 1.0827,
+      "step": 5185
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.056101507455022e-06,
+      "loss": 1.049,
+      "step": 5186
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.055694508279392e-06,
+      "loss": 1.0023,
+      "step": 5187
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0552874305256e-06,
+      "loss": 0.9583,
+      "step": 5188
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.054880274201533e-06,
+      "loss": 1.036,
+      "step": 5189
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.054473039315082e-06,
+      "loss": 0.9887,
+      "step": 5190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.054065725874137e-06,
+      "loss": 1.0349,
+      "step": 5191
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.053658333886586e-06,
+      "loss": 0.6514,
+      "step": 5192
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.053250863360324e-06,
+      "loss": 0.8608,
+      "step": 5193
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.052843314303247e-06,
+      "loss": 1.0246,
+      "step": 5194
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05243568672325e-06,
+      "loss": 0.9978,
+      "step": 5195
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05202798062823e-06,
+      "loss": 1.0833,
+      "step": 5196
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.051620196026087e-06,
+      "loss": 1.0131,
+      "step": 5197
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.051212332924723e-06,
+      "loss": 1.0398,
+      "step": 5198
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.050804391332038e-06,
+      "loss": 0.9488,
+      "step": 5199
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.050396371255938e-06,
+      "loss": 0.9534,
+      "step": 5200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.049988272704326e-06,
+      "loss": 0.8336,
+      "step": 5201
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.049580095685112e-06,
+      "loss": 0.9425,
+      "step": 5202
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0491718402062e-06,
+      "loss": 1.0327,
+      "step": 5203
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.048763506275502e-06,
+      "loss": 0.9956,
+      "step": 5204
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.048355093900929e-06,
+      "loss": 0.8645,
+      "step": 5205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.047946603090397e-06,
+      "loss": 0.9266,
+      "step": 5206
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.047538033851814e-06,
+      "loss": 1.0072,
+      "step": 5207
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.047129386193103e-06,
+      "loss": 1.0308,
+      "step": 5208
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.046720660122177e-06,
+      "loss": 1.0069,
+      "step": 5209
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.046311855646956e-06,
+      "loss": 0.9409,
+      "step": 5210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04590297277536e-06,
+      "loss": 0.7533,
+      "step": 5211
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.045494011515315e-06,
+      "loss": 0.9993,
+      "step": 5212
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.045084971874738e-06,
+      "loss": 0.8248,
+      "step": 5213
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.044675853861557e-06,
+      "loss": 0.9649,
+      "step": 5214
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.044266657483702e-06,
+      "loss": 0.948,
+      "step": 5215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.043857382749095e-06,
+      "loss": 0.929,
+      "step": 5216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.043448029665669e-06,
+      "loss": 0.9778,
+      "step": 5217
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.043038598241355e-06,
+      "loss": 0.9178,
+      "step": 5218
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.042629088484085e-06,
+      "loss": 0.956,
+      "step": 5219
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.042219500401795e-06,
+      "loss": 1.0058,
+      "step": 5220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.041809834002416e-06,
+      "loss": 0.8358,
+      "step": 5221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.041400089293891e-06,
+      "loss": 1.1301,
+      "step": 5222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.040990266284155e-06,
+      "loss": 1.1949,
+      "step": 5223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.040580364981149e-06,
+      "loss": 1.1554,
+      "step": 5224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.040170385392815e-06,
+      "loss": 1.0687,
+      "step": 5225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.039760327527097e-06,
+      "loss": 1.0565,
+      "step": 5226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.039350191391938e-06,
+      "loss": 1.1184,
+      "step": 5227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.038939976995287e-06,
+      "loss": 0.8907,
+      "step": 5228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.038529684345087e-06,
+      "loss": 1.0171,
+      "step": 5229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.038119313449293e-06,
+      "loss": 0.9679,
+      "step": 5230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.037708864315853e-06,
+      "loss": 0.9259,
+      "step": 5231
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.03729833695272e-06,
+      "loss": 0.9658,
+      "step": 5232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.036887731367846e-06,
+      "loss": 1.0014,
+      "step": 5233
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.03647704756919e-06,
+      "loss": 0.9265,
+      "step": 5234
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.036066285564705e-06,
+      "loss": 0.9674,
+      "step": 5235
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.035655445362352e-06,
+      "loss": 1.1532,
+      "step": 5236
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.035244526970091e-06,
+      "loss": 1.0404,
+      "step": 5237
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.034833530395881e-06,
+      "loss": 1.198,
+      "step": 5238
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.034422455647687e-06,
+      "loss": 0.8782,
+      "step": 5239
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.034011302733474e-06,
+      "loss": 1.0034,
+      "step": 5240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.033600071661206e-06,
+      "loss": 0.8834,
+      "step": 5241
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.033188762438853e-06,
+      "loss": 1.1106,
+      "step": 5242
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.032777375074381e-06,
+      "loss": 1.1356,
+      "step": 5243
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.032365909575764e-06,
+      "loss": 1.0139,
+      "step": 5244
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.03195436595097e-06,
+      "loss": 0.8889,
+      "step": 5245
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.031542744207977e-06,
+      "loss": 0.9255,
+      "step": 5246
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.031131044354755e-06,
+      "loss": 1.0212,
+      "step": 5247
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.030719266399285e-06,
+      "loss": 1.0055,
+      "step": 5248
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.030307410349544e-06,
+      "loss": 0.8882,
+      "step": 5249
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.02989547621351e-06,
+      "loss": 1.0263,
+      "step": 5250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.029483463999167e-06,
+      "loss": 0.7834,
+      "step": 5251
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.029071373714496e-06,
+      "loss": 0.935,
+      "step": 5252
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.02865920536748e-06,
+      "loss": 0.8623,
+      "step": 5253
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.028246958966105e-06,
+      "loss": 1.0554,
+      "step": 5254
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.027834634518359e-06,
+      "loss": 1.2228,
+      "step": 5255
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.027422232032232e-06,
+      "loss": 1.0885,
+      "step": 5256
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.027009751515711e-06,
+      "loss": 0.633,
+      "step": 5257
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.026597192976792e-06,
+      "loss": 0.9318,
+      "step": 5258
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.026184556423464e-06,
+      "loss": 1.0555,
+      "step": 5259
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.025771841863724e-06,
+      "loss": 0.9904,
+      "step": 5260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.025359049305569e-06,
+      "loss": 1.1091,
+      "step": 5261
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.024946178756994e-06,
+      "loss": 1.0069,
+      "step": 5262
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.024533230226e-06,
+      "loss": 1.1014,
+      "step": 5263
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.024120203720587e-06,
+      "loss": 1.0121,
+      "step": 5264
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.02370709924876e-06,
+      "loss": 0.848,
+      "step": 5265
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.02329391681852e-06,
+      "loss": 1.066,
+      "step": 5266
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.022880656437871e-06,
+      "loss": 0.9503,
+      "step": 5267
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.022467318114823e-06,
+      "loss": 1.1283,
+      "step": 5268
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.022053901857383e-06,
+      "loss": 0.9574,
+      "step": 5269
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.02164040767356e-06,
+      "loss": 0.8431,
+      "step": 5270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.021226835571369e-06,
+      "loss": 1.0076,
+      "step": 5271
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.020813185558819e-06,
+      "loss": 0.9872,
+      "step": 5272
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.020399457643924e-06,
+      "loss": 1.0771,
+      "step": 5273
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.019985651834703e-06,
+      "loss": 0.93,
+      "step": 5274
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.019571768139172e-06,
+      "loss": 0.8944,
+      "step": 5275
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.019157806565351e-06,
+      "loss": 0.8887,
+      "step": 5276
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.018743767121257e-06,
+      "loss": 0.9993,
+      "step": 5277
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.018329649814914e-06,
+      "loss": 1.0265,
+      "step": 5278
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.017915454654347e-06,
+      "loss": 0.8504,
+      "step": 5279
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.017501181647578e-06,
+      "loss": 1.0513,
+      "step": 5280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.017086830802638e-06,
+      "loss": 0.9571,
+      "step": 5281
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.016672402127548e-06,
+      "loss": 1.0834,
+      "step": 5282
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.016257895630342e-06,
+      "loss": 1.0187,
+      "step": 5283
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.015843311319052e-06,
+      "loss": 0.7573,
+      "step": 5284
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01542864920171e-06,
+      "loss": 0.8968,
+      "step": 5285
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.015013909286345e-06,
+      "loss": 0.9156,
+      "step": 5286
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.014599091581e-06,
+      "loss": 0.7818,
+      "step": 5287
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.014184196093705e-06,
+      "loss": 1.0707,
+      "step": 5288
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.013769222832503e-06,
+      "loss": 0.805,
+      "step": 5289
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.013354171805434e-06,
+      "loss": 0.8927,
+      "step": 5290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.012939043020537e-06,
+      "loss": 1.1508,
+      "step": 5291
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.012523836485857e-06,
+      "loss": 0.984,
+      "step": 5292
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.012108552209438e-06,
+      "loss": 1.1478,
+      "step": 5293
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.011693190199323e-06,
+      "loss": 0.995,
+      "step": 5294
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.011277750463566e-06,
+      "loss": 0.9516,
+      "step": 5295
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.010862233010212e-06,
+      "loss": 0.9367,
+      "step": 5296
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01044663784731e-06,
+      "loss": 0.9794,
+      "step": 5297
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.010030964982916e-06,
+      "loss": 0.953,
+      "step": 5298
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.009615214425083e-06,
+      "loss": 0.9911,
+      "step": 5299
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.009199386181862e-06,
+      "loss": 1.0125,
+      "step": 5300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.008783480261312e-06,
+      "loss": 0.9864,
+      "step": 5301
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.008367496671494e-06,
+      "loss": 0.9932,
+      "step": 5302
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.007951435420463e-06,
+      "loss": 1.0354,
+      "step": 5303
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.007535296516283e-06,
+      "loss": 0.9031,
+      "step": 5304
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.007119079967016e-06,
+      "loss": 1.0614,
+      "step": 5305
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.006702785780725e-06,
+      "loss": 0.9926,
+      "step": 5306
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.006286413965476e-06,
+      "loss": 0.8566,
+      "step": 5307
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.005869964529337e-06,
+      "loss": 0.9622,
+      "step": 5308
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.005453437480378e-06,
+      "loss": 0.9141,
+      "step": 5309
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.005036832826664e-06,
+      "loss": 1.1536,
+      "step": 5310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.004620150576273e-06,
+      "loss": 1.0054,
+      "step": 5311
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.004203390737272e-06,
+      "loss": 0.9892,
+      "step": 5312
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.003786553317739e-06,
+      "loss": 1.0642,
+      "step": 5313
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.003369638325751e-06,
+      "loss": 0.9192,
+      "step": 5314
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.002952645769385e-06,
+      "loss": 1.0054,
+      "step": 5315
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00253557565672e-06,
+      "loss": 0.8115,
+      "step": 5316
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.002118427995834e-06,
+      "loss": 1.1292,
+      "step": 5317
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00170120279481e-06,
+      "loss": 1.0601,
+      "step": 5318
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.001283900061736e-06,
+      "loss": 0.8881,
+      "step": 5319
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.000866519804695e-06,
+      "loss": 1.0561,
+      "step": 5320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.000449062031769e-06,
+      "loss": 0.8748,
+      "step": 5321
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.000031526751052e-06,
+      "loss": 0.9669,
+      "step": 5322
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99961391397063e-06,
+      "loss": 0.903,
+      "step": 5323
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.999196223698599e-06,
+      "loss": 0.9809,
+      "step": 5324
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.998778455943045e-06,
+      "loss": 0.9386,
+      "step": 5325
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.998360610712067e-06,
+      "loss": 0.9177,
+      "step": 5326
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.997942688013759e-06,
+      "loss": 1.0518,
+      "step": 5327
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.997524687856219e-06,
+      "loss": 0.8758,
+      "step": 5328
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.997106610247542e-06,
+      "loss": 0.9351,
+      "step": 5329
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.996688455195835e-06,
+      "loss": 1.0061,
+      "step": 5330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.996270222709194e-06,
+      "loss": 1.058,
+      "step": 5331
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.995851912795724e-06,
+      "loss": 0.8047,
+      "step": 5332
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.995433525463528e-06,
+      "loss": 1.0042,
+      "step": 5333
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.995015060720716e-06,
+      "loss": 0.8858,
+      "step": 5334
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.994596518575393e-06,
+      "loss": 0.9948,
+      "step": 5335
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.994177899035667e-06,
+      "loss": 0.7724,
+      "step": 5336
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.993759202109654e-06,
+      "loss": 1.0333,
+      "step": 5337
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.993340427805457e-06,
+      "loss": 0.7971,
+      "step": 5338
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.9929215761312e-06,
+      "loss": 0.9349,
+      "step": 5339
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99250264709499e-06,
+      "loss": 1.0796,
+      "step": 5340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.992083640704949e-06,
+      "loss": 0.8855,
+      "step": 5341
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99166455696919e-06,
+      "loss": 0.8648,
+      "step": 5342
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99124539589584e-06,
+      "loss": 1.1456,
+      "step": 5343
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.990826157493013e-06,
+      "loss": 0.7707,
+      "step": 5344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.990406841768835e-06,
+      "loss": 1.0094,
+      "step": 5345
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98998744873143e-06,
+      "loss": 0.866,
+      "step": 5346
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.989567978388922e-06,
+      "loss": 1.1126,
+      "step": 5347
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98914843074944e-06,
+      "loss": 0.9016,
+      "step": 5348
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.988728805821112e-06,
+      "loss": 1.0421,
+      "step": 5349
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98830910361207e-06,
+      "loss": 1.0824,
+      "step": 5350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.987889324130443e-06,
+      "loss": 1.1645,
+      "step": 5351
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.987469467384363e-06,
+      "loss": 0.917,
+      "step": 5352
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98704953338197e-06,
+      "loss": 0.9084,
+      "step": 5353
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.986629522131392e-06,
+      "loss": 1.0236,
+      "step": 5354
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.986209433640774e-06,
+      "loss": 0.9259,
+      "step": 5355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.985789267918252e-06,
+      "loss": 1.0836,
+      "step": 5356
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.985369024971968e-06,
+      "loss": 0.9168,
+      "step": 5357
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.984948704810063e-06,
+      "loss": 0.9532,
+      "step": 5358
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.984528307440681e-06,
+      "loss": 1.0211,
+      "step": 5359
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.984107832871966e-06,
+      "loss": 0.8534,
+      "step": 5360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.983687281112066e-06,
+      "loss": 0.8394,
+      "step": 5361
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.983266652169128e-06,
+      "loss": 1.0305,
+      "step": 5362
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.982845946051302e-06,
+      "loss": 0.9281,
+      "step": 5363
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98242516276674e-06,
+      "loss": 1.0034,
+      "step": 5364
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.982004302323593e-06,
+      "loss": 1.036,
+      "step": 5365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.981583364730018e-06,
+      "loss": 0.9685,
+      "step": 5366
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.981162349994165e-06,
+      "loss": 0.9483,
+      "step": 5367
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.980741258124198e-06,
+      "loss": 1.1321,
+      "step": 5368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98032008912827e-06,
+      "loss": 0.9918,
+      "step": 5369
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.979898843014545e-06,
+      "loss": 0.9535,
+      "step": 5370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.97947751979118e-06,
+      "loss": 0.9045,
+      "step": 5371
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.979056119466343e-06,
+      "loss": 0.9631,
+      "step": 5372
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.978634642048194e-06,
+      "loss": 1.0162,
+      "step": 5373
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.978213087544904e-06,
+      "loss": 0.9002,
+      "step": 5374
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.977791455964637e-06,
+      "loss": 1.0432,
+      "step": 5375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.977369747315563e-06,
+      "loss": 0.9426,
+      "step": 5376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.976947961605852e-06,
+      "loss": 1.022,
+      "step": 5377
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.976526098843677e-06,
+      "loss": 0.9393,
+      "step": 5378
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.976104159037208e-06,
+      "loss": 0.828,
+      "step": 5379
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.975682142194626e-06,
+      "loss": 1.0817,
+      "step": 5380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.975260048324103e-06,
+      "loss": 1.0807,
+      "step": 5381
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.974837877433819e-06,
+      "loss": 0.9434,
+      "step": 5382
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.974415629531952e-06,
+      "loss": 1.2629,
+      "step": 5383
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.973993304626685e-06,
+      "loss": 1.0065,
+      "step": 5384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.973570902726198e-06,
+      "loss": 0.9162,
+      "step": 5385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.973148423838675e-06,
+      "loss": 1.0892,
+      "step": 5386
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.972725867972304e-06,
+      "loss": 0.9529,
+      "step": 5387
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.97230323513527e-06,
+      "loss": 1.0202,
+      "step": 5388
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.971880525335763e-06,
+      "loss": 0.8701,
+      "step": 5389
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.971457738581972e-06,
+      "loss": 1.1035,
+      "step": 5390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.971034874882087e-06,
+      "loss": 0.8658,
+      "step": 5391
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.970611934244302e-06,
+      "loss": 1.0075,
+      "step": 5392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.970188916676812e-06,
+      "loss": 1.1155,
+      "step": 5393
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.969765822187812e-06,
+      "loss": 0.9777,
+      "step": 5394
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.9693426507855e-06,
+      "loss": 0.9307,
+      "step": 5395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.968919402478076e-06,
+      "loss": 1.04,
+      "step": 5396
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.968496077273736e-06,
+      "loss": 0.9132,
+      "step": 5397
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.968072675180686e-06,
+      "loss": 1.052,
+      "step": 5398
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.967649196207128e-06,
+      "loss": 1.0844,
+      "step": 5399
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.967225640361268e-06,
+      "loss": 0.9203,
+      "step": 5400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96680200765131e-06,
+      "loss": 1.0817,
+      "step": 5401
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.966378298085462e-06,
+      "loss": 0.9552,
+      "step": 5402
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.965954511671935e-06,
+      "loss": 1.0722,
+      "step": 5403
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.965530648418938e-06,
+      "loss": 1.0015,
+      "step": 5404
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.965106708334687e-06,
+      "loss": 1.1129,
+      "step": 5405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96468269142739e-06,
+      "loss": 0.8511,
+      "step": 5406
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.964258597705267e-06,
+      "loss": 0.988,
+      "step": 5407
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.963834427176532e-06,
+      "loss": 0.8617,
+      "step": 5408
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.963410179849405e-06,
+      "loss": 0.9196,
+      "step": 5409
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.962985855732105e-06,
+      "loss": 0.9466,
+      "step": 5410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.962561454832852e-06,
+      "loss": 1.1807,
+      "step": 5411
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96213697715987e-06,
+      "loss": 0.9806,
+      "step": 5412
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.961712422721383e-06,
+      "loss": 1.016,
+      "step": 5413
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.961287791525616e-06,
+      "loss": 0.9485,
+      "step": 5414
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.960863083580798e-06,
+      "loss": 0.9763,
+      "step": 5415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.960438298895154e-06,
+      "loss": 0.9847,
+      "step": 5416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.960013437476918e-06,
+      "loss": 0.8674,
+      "step": 5417
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95958849933432e-06,
+      "loss": 1.0384,
+      "step": 5418
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.959163484475591e-06,
+      "loss": 0.8915,
+      "step": 5419
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.958738392908969e-06,
+      "loss": 1.0973,
+      "step": 5420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95831322464269e-06,
+      "loss": 0.8651,
+      "step": 5421
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.957887979684987e-06,
+      "loss": 0.91,
+      "step": 5422
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.957462658044104e-06,
+      "loss": 0.9335,
+      "step": 5423
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.957037259728281e-06,
+      "loss": 1.0674,
+      "step": 5424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.956611784745756e-06,
+      "loss": 0.7549,
+      "step": 5425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.956186233104777e-06,
+      "loss": 0.9151,
+      "step": 5426
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.955760604813587e-06,
+      "loss": 0.9497,
+      "step": 5427
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.955334899880432e-06,
+      "loss": 1.1698,
+      "step": 5428
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.954909118313561e-06,
+      "loss": 0.8016,
+      "step": 5429
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.954483260121223e-06,
+      "loss": 1.0281,
+      "step": 5430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.954057325311669e-06,
+      "loss": 0.9238,
+      "step": 5431
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95363131389315e-06,
+      "loss": 1.1421,
+      "step": 5432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.953205225873922e-06,
+      "loss": 0.9374,
+      "step": 5433
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95277906126224e-06,
+      "loss": 0.9836,
+      "step": 5434
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.952352820066359e-06,
+      "loss": 1.1078,
+      "step": 5435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95192650229454e-06,
+      "loss": 0.9288,
+      "step": 5436
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.951500107955041e-06,
+      "loss": 0.9417,
+      "step": 5437
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.951073637056125e-06,
+      "loss": 0.8934,
+      "step": 5438
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95064708960605e-06,
+      "loss": 0.856,
+      "step": 5439
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.950220465613087e-06,
+      "loss": 0.9315,
+      "step": 5440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.949793765085497e-06,
+      "loss": 1.0833,
+      "step": 5441
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94936698803155e-06,
+      "loss": 1.0515,
+      "step": 5442
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.948940134459512e-06,
+      "loss": 1.1621,
+      "step": 5443
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.948513204377657e-06,
+      "loss": 0.688,
+      "step": 5444
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.948086197794253e-06,
+      "loss": 0.7781,
+      "step": 5445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.947659114717572e-06,
+      "loss": 1.0075,
+      "step": 5446
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.947231955155894e-06,
+      "loss": 0.8956,
+      "step": 5447
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.946804719117492e-06,
+      "loss": 1.0262,
+      "step": 5448
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.946377406610645e-06,
+      "loss": 0.8236,
+      "step": 5449
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.945950017643629e-06,
+      "loss": 1.0115,
+      "step": 5450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.945522552224727e-06,
+      "loss": 0.8144,
+      "step": 5451
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94509501036222e-06,
+      "loss": 1.1503,
+      "step": 5452
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.944667392064393e-06,
+      "loss": 0.9386,
+      "step": 5453
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94423969733953e-06,
+      "loss": 0.9374,
+      "step": 5454
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.943811926195917e-06,
+      "loss": 0.9022,
+      "step": 5455
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.943384078641843e-06,
+      "loss": 1.0402,
+      "step": 5456
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.942956154685596e-06,
+      "loss": 0.9058,
+      "step": 5457
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.942528154335469e-06,
+      "loss": 1.0976,
+      "step": 5458
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.942100077599753e-06,
+      "loss": 0.9576,
+      "step": 5459
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.941671924486743e-06,
+      "loss": 1.0438,
+      "step": 5460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.941243695004734e-06,
+      "loss": 1.1472,
+      "step": 5461
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94081538916202e-06,
+      "loss": 0.9859,
+      "step": 5462
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.940387006966901e-06,
+      "loss": 1.0634,
+      "step": 5463
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.939958548427681e-06,
+      "loss": 0.8287,
+      "step": 5464
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.939530013552658e-06,
+      "loss": 0.9555,
+      "step": 5465
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.939101402350132e-06,
+      "loss": 1.0191,
+      "step": 5466
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.938672714828412e-06,
+      "loss": 0.8768,
+      "step": 5467
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.938243950995799e-06,
+      "loss": 0.9836,
+      "step": 5468
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.937815110860604e-06,
+      "loss": 0.8837,
+      "step": 5469
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.937386194431135e-06,
+      "loss": 0.9147,
+      "step": 5470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.9369572017157e-06,
+      "loss": 0.9325,
+      "step": 5471
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.936528132722616e-06,
+      "loss": 1.0278,
+      "step": 5472
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.936098987460188e-06,
+      "loss": 1.0684,
+      "step": 5473
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.935669765936739e-06,
+      "loss": 0.9889,
+      "step": 5474
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.935240468160578e-06,
+      "loss": 1.016,
+      "step": 5475
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.934811094140027e-06,
+      "loss": 0.9491,
+      "step": 5476
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.934381643883403e-06,
+      "loss": 1.0232,
+      "step": 5477
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.933952117399027e-06,
+      "loss": 0.9627,
+      "step": 5478
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.933522514695224e-06,
+      "loss": 0.9447,
+      "step": 5479
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.93309283578031e-06,
+      "loss": 1.1473,
+      "step": 5480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.932663080662618e-06,
+      "loss": 0.9393,
+      "step": 5481
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.932233249350468e-06,
+      "loss": 1.1102,
+      "step": 5482
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.931803341852192e-06,
+      "loss": 1.088,
+      "step": 5483
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.931373358176118e-06,
+      "loss": 1.0178,
+      "step": 5484
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.93094329833058e-06,
+      "loss": 0.9176,
+      "step": 5485
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.930513162323902e-06,
+      "loss": 1.0695,
+      "step": 5486
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.930082950164427e-06,
+      "loss": 0.8335,
+      "step": 5487
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.929652661860483e-06,
+      "loss": 0.8312,
+      "step": 5488
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.929222297420413e-06,
+      "loss": 0.9179,
+      "step": 5489
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.928791856852551e-06,
+      "loss": 0.8644,
+      "step": 5490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.928361340165239e-06,
+      "loss": 1.0138,
+      "step": 5491
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.927930747366816e-06,
+      "loss": 0.9979,
+      "step": 5492
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.927500078465628e-06,
+      "loss": 1.1181,
+      "step": 5493
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.927069333470014e-06,
+      "loss": 0.9668,
+      "step": 5494
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.926638512388323e-06,
+      "loss": 0.9851,
+      "step": 5495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.926207615228902e-06,
+      "loss": 1.0598,
+      "step": 5496
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.925776642000099e-06,
+      "loss": 1.0189,
+      "step": 5497
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.925345592710265e-06,
+      "loss": 0.9903,
+      "step": 5498
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.924914467367749e-06,
+      "loss": 0.992,
+      "step": 5499
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.924483265980908e-06,
+      "loss": 0.944,
+      "step": 5500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.924051988558091e-06,
+      "loss": 1.0414,
+      "step": 5501
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.923620635107659e-06,
+      "loss": 0.9727,
+      "step": 5502
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.923189205637966e-06,
+      "loss": 0.9502,
+      "step": 5503
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.922757700157371e-06,
+      "loss": 1.0117,
+      "step": 5504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.922326118674238e-06,
+      "loss": 0.9325,
+      "step": 5505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.921894461196924e-06,
+      "loss": 0.9704,
+      "step": 5506
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.921462727733797e-06,
+      "loss": 0.9581,
+      "step": 5507
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.921030918293216e-06,
+      "loss": 0.9954,
+      "step": 5508
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.920599032883553e-06,
+      "loss": 0.9155,
+      "step": 5509
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.920167071513173e-06,
+      "loss": 1.0551,
+      "step": 5510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.919735034190444e-06,
+      "loss": 0.867,
+      "step": 5511
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.919302920923738e-06,
+      "loss": 0.9366,
+      "step": 5512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.918870731721427e-06,
+      "loss": 1.0308,
+      "step": 5513
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.918438466591885e-06,
+      "loss": 1.1388,
+      "step": 5514
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.918006125543486e-06,
+      "loss": 0.9447,
+      "step": 5515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.917573708584606e-06,
+      "loss": 1.0992,
+      "step": 5516
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.917141215723625e-06,
+      "loss": 1.1261,
+      "step": 5517
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.916708646968924e-06,
+      "loss": 0.9633,
+      "step": 5518
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.916276002328879e-06,
+      "loss": 0.9237,
+      "step": 5519
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.915843281811874e-06,
+      "loss": 1.0178,
+      "step": 5520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.915410485426295e-06,
+      "loss": 1.0125,
+      "step": 5521
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.914977613180526e-06,
+      "loss": 1.1006,
+      "step": 5522
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.914544665082953e-06,
+      "loss": 1.078,
+      "step": 5523
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.914111641141965e-06,
+      "loss": 1.0365,
+      "step": 5524
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.913678541365954e-06,
+      "loss": 0.903,
+      "step": 5525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.913245365763309e-06,
+      "loss": 0.8385,
+      "step": 5526
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.912812114342419e-06,
+      "loss": 0.941,
+      "step": 5527
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.912378787111684e-06,
+      "loss": 0.9917,
+      "step": 5528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.911945384079497e-06,
+      "loss": 1.1957,
+      "step": 5529
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.911511905254256e-06,
+      "loss": 1.0669,
+      "step": 5530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.91107835064436e-06,
+      "loss": 1.0653,
+      "step": 5531
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.910644720258208e-06,
+      "loss": 1.0463,
+      "step": 5532
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.9102110141042e-06,
+      "loss": 0.9901,
+      "step": 5533
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.909777232190741e-06,
+      "loss": 0.9453,
+      "step": 5534
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.909343374526237e-06,
+      "loss": 1.0138,
+      "step": 5535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90890944111909e-06,
+      "loss": 0.8052,
+      "step": 5536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90847543197771e-06,
+      "loss": 1.0324,
+      "step": 5537
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.908041347110506e-06,
+      "loss": 0.9825,
+      "step": 5538
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.907607186525884e-06,
+      "loss": 1.0079,
+      "step": 5539
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.907172950232262e-06,
+      "loss": 1.0779,
+      "step": 5540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90673863823805e-06,
+      "loss": 0.8796,
+      "step": 5541
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.906304250551662e-06,
+      "loss": 0.9836,
+      "step": 5542
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.905869787181517e-06,
+      "loss": 1.1433,
+      "step": 5543
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90543524813603e-06,
+      "loss": 1.0585,
+      "step": 5544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.905000633423622e-06,
+      "loss": 0.9267,
+      "step": 5545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90456594305271e-06,
+      "loss": 0.828,
+      "step": 5546
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90413117703172e-06,
+      "loss": 1.1776,
+      "step": 5547
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.903696335369073e-06,
+      "loss": 0.8893,
+      "step": 5548
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.903261418073198e-06,
+      "loss": 0.9574,
+      "step": 5549
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.902826425152516e-06,
+      "loss": 0.9296,
+      "step": 5550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.902391356615458e-06,
+      "loss": 0.9977,
+      "step": 5551
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.901956212470453e-06,
+      "loss": 0.9937,
+      "step": 5552
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90152099272593e-06,
+      "loss": 0.8766,
+      "step": 5553
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.901085697390324e-06,
+      "loss": 1.0658,
+      "step": 5554
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.900650326472067e-06,
+      "loss": 0.9802,
+      "step": 5555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.900214879979596e-06,
+      "loss": 1.0086,
+      "step": 5556
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.899779357921346e-06,
+      "loss": 0.9359,
+      "step": 5557
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.899343760305754e-06,
+      "loss": 0.8554,
+      "step": 5558
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.898908087141263e-06,
+      "loss": 0.9955,
+      "step": 5559
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89847233843631e-06,
+      "loss": 0.8524,
+      "step": 5560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.898036514199342e-06,
+      "loss": 0.9797,
+      "step": 5561
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.897600614438799e-06,
+      "loss": 1.1288,
+      "step": 5562
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.897164639163129e-06,
+      "loss": 1.0717,
+      "step": 5563
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89672858838078e-06,
+      "loss": 0.9652,
+      "step": 5564
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.896292462100196e-06,
+      "loss": 0.9158,
+      "step": 5565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.895856260329832e-06,
+      "loss": 0.9067,
+      "step": 5566
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.895419983078134e-06,
+      "loss": 0.8751,
+      "step": 5567
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.894983630353558e-06,
+      "loss": 0.9648,
+      "step": 5568
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.894547202164558e-06,
+      "loss": 1.0816,
+      "step": 5569
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.894110698519588e-06,
+      "loss": 1.0072,
+      "step": 5570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.893674119427109e-06,
+      "loss": 0.9814,
+      "step": 5571
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.893237464895576e-06,
+      "loss": 1.0088,
+      "step": 5572
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.892800734933452e-06,
+      "loss": 0.9383,
+      "step": 5573
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.892363929549195e-06,
+      "loss": 0.8734,
+      "step": 5574
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89192704875127e-06,
+      "loss": 1.0962,
+      "step": 5575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89149009254814e-06,
+      "loss": 0.9367,
+      "step": 5576
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.891053060948274e-06,
+      "loss": 0.8666,
+      "step": 5577
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.890615953960137e-06,
+      "loss": 0.8947,
+      "step": 5578
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.890178771592198e-06,
+      "loss": 1.0467,
+      "step": 5579
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.889741513852928e-06,
+      "loss": 1.0583,
+      "step": 5580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.889304180750799e-06,
+      "loss": 0.8897,
+      "step": 5581
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.888866772294285e-06,
+      "loss": 0.8995,
+      "step": 5582
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.888429288491857e-06,
+      "loss": 0.9843,
+      "step": 5583
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.887991729351994e-06,
+      "loss": 0.854,
+      "step": 5584
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.887554094883174e-06,
+      "loss": 1.0944,
+      "step": 5585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.887116385093875e-06,
+      "loss": 0.7922,
+      "step": 5586
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.886678599992577e-06,
+      "loss": 1.1305,
+      "step": 5587
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.886240739587764e-06,
+      "loss": 1.0604,
+      "step": 5588
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.885802803887917e-06,
+      "loss": 0.8826,
+      "step": 5589
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.885364792901522e-06,
+      "loss": 1.1542,
+      "step": 5590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.884926706637067e-06,
+      "loss": 1.0958,
+      "step": 5591
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.884488545103039e-06,
+      "loss": 0.8308,
+      "step": 5592
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.884050308307924e-06,
+      "loss": 1.0624,
+      "step": 5593
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.883611996260218e-06,
+      "loss": 0.8212,
+      "step": 5594
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.883173608968408e-06,
+      "loss": 0.9428,
+      "step": 5595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.88273514644099e-06,
+      "loss": 0.9149,
+      "step": 5596
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.882296608686462e-06,
+      "loss": 0.9075,
+      "step": 5597
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.881857995713316e-06,
+      "loss": 0.9987,
+      "step": 5598
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.881419307530053e-06,
+      "loss": 1.0532,
+      "step": 5599
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.88098054414517e-06,
+      "loss": 1.0012,
+      "step": 5600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.880541705567169e-06,
+      "loss": 1.0406,
+      "step": 5601
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.880102791804555e-06,
+      "loss": 0.8701,
+      "step": 5602
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.879663802865828e-06,
+      "loss": 0.9405,
+      "step": 5603
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.879224738759496e-06,
+      "loss": 0.9874,
+      "step": 5604
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.878785599494061e-06,
+      "loss": 0.916,
+      "step": 5605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.878346385078039e-06,
+      "loss": 1.0615,
+      "step": 5606
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.877907095519934e-06,
+      "loss": 0.9037,
+      "step": 5607
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.877467730828259e-06,
+      "loss": 0.9108,
+      "step": 5608
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.877028291011527e-06,
+      "loss": 0.8624,
+      "step": 5609
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87658877607825e-06,
+      "loss": 0.9976,
+      "step": 5610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.876149186036945e-06,
+      "loss": 0.9434,
+      "step": 5611
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87570952089613e-06,
+      "loss": 0.9483,
+      "step": 5612
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.875269780664319e-06,
+      "loss": 1.1482,
+      "step": 5613
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.874829965350038e-06,
+      "loss": 0.7733,
+      "step": 5614
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.874390074961805e-06,
+      "loss": 1.0051,
+      "step": 5615
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.873950109508143e-06,
+      "loss": 1.1517,
+      "step": 5616
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.873510068997576e-06,
+      "loss": 0.9794,
+      "step": 5617
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87306995343863e-06,
+      "loss": 0.9308,
+      "step": 5618
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.872629762839832e-06,
+      "loss": 1.0598,
+      "step": 5619
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.872189497209712e-06,
+      "loss": 0.8722,
+      "step": 5620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.871749156556799e-06,
+      "loss": 0.9244,
+      "step": 5621
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.871308740889622e-06,
+      "loss": 1.0641,
+      "step": 5622
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.870868250216718e-06,
+      "loss": 0.9776,
+      "step": 5623
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87042768454662e-06,
+      "loss": 0.9236,
+      "step": 5624
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.869987043887863e-06,
+      "loss": 0.9586,
+      "step": 5625
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.869546328248985e-06,
+      "loss": 1.1118,
+      "step": 5626
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.869105537638525e-06,
+      "loss": 0.921,
+      "step": 5627
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.868664672065021e-06,
+      "loss": 0.9786,
+      "step": 5628
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.868223731537019e-06,
+      "loss": 1.045,
+      "step": 5629
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.867782716063059e-06,
+      "loss": 0.7972,
+      "step": 5630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.867341625651685e-06,
+      "loss": 0.9675,
+      "step": 5631
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.866900460311446e-06,
+      "loss": 0.9865,
+      "step": 5632
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.866459220050888e-06,
+      "loss": 0.956,
+      "step": 5633
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.866017904878556e-06,
+      "loss": 0.969,
+      "step": 5634
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.865576514803006e-06,
+      "loss": 1.0136,
+      "step": 5635
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86513504983279e-06,
+      "loss": 0.8708,
+      "step": 5636
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.864693509976458e-06,
+      "loss": 1.0677,
+      "step": 5637
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.864251895242564e-06,
+      "loss": 1.086,
+      "step": 5638
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86381020563967e-06,
+      "loss": 0.8937,
+      "step": 5639
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.863368441176326e-06,
+      "loss": 1.0376,
+      "step": 5640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.862926601861098e-06,
+      "loss": 0.8898,
+      "step": 5641
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86248468770254e-06,
+      "loss": 1.131,
+      "step": 5642
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.862042698709219e-06,
+      "loss": 1.0245,
+      "step": 5643
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.861600634889697e-06,
+      "loss": 0.916,
+      "step": 5644
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.861158496252537e-06,
+      "loss": 1.0363,
+      "step": 5645
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86071628280631e-06,
+      "loss": 0.862,
+      "step": 5646
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.860273994559578e-06,
+      "loss": 1.0541,
+      "step": 5647
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.859831631520914e-06,
+      "loss": 0.9673,
+      "step": 5648
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.859389193698888e-06,
+      "loss": 0.9407,
+      "step": 5649
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.858946681102071e-06,
+      "loss": 0.8777,
+      "step": 5650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.858504093739038e-06,
+      "loss": 0.8774,
+      "step": 5651
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.858061431618365e-06,
+      "loss": 0.9567,
+      "step": 5652
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.857618694748625e-06,
+      "loss": 0.9789,
+      "step": 5653
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.857175883138398e-06,
+      "loss": 0.9419,
+      "step": 5654
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.856732996796262e-06,
+      "loss": 1.1941,
+      "step": 5655
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.856290035730802e-06,
+      "loss": 1.2606,
+      "step": 5656
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.855846999950595e-06,
+      "loss": 1.0147,
+      "step": 5657
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.855403889464229e-06,
+      "loss": 0.9142,
+      "step": 5658
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.854960704280285e-06,
+      "loss": 1.0525,
+      "step": 5659
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.854517444407353e-06,
+      "loss": 1.0558,
+      "step": 5660
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85407410985402e-06,
+      "loss": 0.8078,
+      "step": 5661
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.853630700628874e-06,
+      "loss": 0.9173,
+      "step": 5662
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85318721674051e-06,
+      "loss": 0.9814,
+      "step": 5663
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.852743658197516e-06,
+      "loss": 0.934,
+      "step": 5664
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.852300025008487e-06,
+      "loss": 1.0232,
+      "step": 5665
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85185631718202e-06,
+      "loss": 0.9635,
+      "step": 5666
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85141253472671e-06,
+      "loss": 0.9572,
+      "step": 5667
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.850968677651155e-06,
+      "loss": 1.0061,
+      "step": 5668
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.850524745963958e-06,
+      "loss": 1.0997,
+      "step": 5669
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.850080739673714e-06,
+      "loss": 0.9164,
+      "step": 5670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.849636658789032e-06,
+      "loss": 1.0102,
+      "step": 5671
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.849192503318512e-06,
+      "loss": 1.0785,
+      "step": 5672
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.848748273270759e-06,
+      "loss": 0.8732,
+      "step": 5673
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.848303968654384e-06,
+      "loss": 0.9232,
+      "step": 5674
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84785958947799e-06,
+      "loss": 1.0086,
+      "step": 5675
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84741513575019e-06,
+      "loss": 0.9801,
+      "step": 5676
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.846970607479595e-06,
+      "loss": 0.9891,
+      "step": 5677
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.846526004674817e-06,
+      "loss": 0.948,
+      "step": 5678
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.846081327344471e-06,
+      "loss": 0.8568,
+      "step": 5679
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84563657549717e-06,
+      "loss": 0.9839,
+      "step": 5680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.845191749141532e-06,
+      "loss": 0.9799,
+      "step": 5681
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84474684828618e-06,
+      "loss": 0.8409,
+      "step": 5682
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.844301872939726e-06,
+      "loss": 0.9198,
+      "step": 5683
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.843856823110797e-06,
+      "loss": 0.8099,
+      "step": 5684
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.843411698808015e-06,
+      "loss": 0.9426,
+      "step": 5685
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.842966500040002e-06,
+      "loss": 0.9807,
+      "step": 5686
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.842521226815384e-06,
+      "loss": 1.0006,
+      "step": 5687
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84207587914279e-06,
+      "loss": 0.9334,
+      "step": 5688
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.841630457030847e-06,
+      "loss": 0.8999,
+      "step": 5689
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.841184960488185e-06,
+      "loss": 0.8994,
+      "step": 5690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.840739389523435e-06,
+      "loss": 1.1152,
+      "step": 5691
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.840293744145232e-06,
+      "loss": 0.9728,
+      "step": 5692
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.83984802436221e-06,
+      "loss": 1.0402,
+      "step": 5693
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.839402230183e-06,
+      "loss": 0.9208,
+      "step": 5694
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.838956361616247e-06,
+      "loss": 0.8833,
+      "step": 5695
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.838510418670582e-06,
+      "loss": 1.0844,
+      "step": 5696
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.838064401354649e-06,
+      "loss": 0.9215,
+      "step": 5697
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.83761830967709e-06,
+      "loss": 0.9376,
+      "step": 5698
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.837172143646545e-06,
+      "loss": 0.9285,
+      "step": 5699
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.83672590327166e-06,
+      "loss": 0.887,
+      "step": 5700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.836279588561084e-06,
+      "loss": 0.88,
+      "step": 5701
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.835833199523459e-06,
+      "loss": 0.9171,
+      "step": 5702
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.835386736167436e-06,
+      "loss": 1.0495,
+      "step": 5703
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.834940198501665e-06,
+      "loss": 0.9687,
+      "step": 5704
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.834493586534797e-06,
+      "loss": 1.0057,
+      "step": 5705
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.834046900275486e-06,
+      "loss": 1.0901,
+      "step": 5706
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.833600139732387e-06,
+      "loss": 1.1665,
+      "step": 5707
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.833153304914153e-06,
+      "loss": 0.947,
+      "step": 5708
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.832706395829444e-06,
+      "loss": 1.0598,
+      "step": 5709
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.832259412486918e-06,
+      "loss": 0.9074,
+      "step": 5710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.831812354895238e-06,
+      "loss": 0.9445,
+      "step": 5711
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.83136522306306e-06,
+      "loss": 0.8936,
+      "step": 5712
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.83091801699905e-06,
+      "loss": 0.9194,
+      "step": 5713
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.830470736711872e-06,
+      "loss": 0.8537,
+      "step": 5714
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.830023382210193e-06,
+      "loss": 1.101,
+      "step": 5715
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82957595350268e-06,
+      "loss": 0.7357,
+      "step": 5716
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.829128450598e-06,
+      "loss": 0.9476,
+      "step": 5717
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.828680873504827e-06,
+      "loss": 0.9199,
+      "step": 5718
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82823322223183e-06,
+      "loss": 1.0671,
+      "step": 5719
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.827785496787683e-06,
+      "loss": 0.778,
+      "step": 5720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82733769718106e-06,
+      "loss": 0.8797,
+      "step": 5721
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.826889823420637e-06,
+      "loss": 1.0686,
+      "step": 5722
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.826441875515092e-06,
+      "loss": 1.0629,
+      "step": 5723
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.825993853473102e-06,
+      "loss": 1.0102,
+      "step": 5724
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82554575730335e-06,
+      "loss": 1.0061,
+      "step": 5725
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82509758701452e-06,
+      "loss": 1.1661,
+      "step": 5726
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.824649342615289e-06,
+      "loss": 1.1044,
+      "step": 5727
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.824201024114344e-06,
+      "loss": 0.9076,
+      "step": 5728
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.823752631520373e-06,
+      "loss": 0.9446,
+      "step": 5729
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.823304164842061e-06,
+      "loss": 0.8104,
+      "step": 5730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.822855624088099e-06,
+      "loss": 0.7958,
+      "step": 5731
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.822407009267175e-06,
+      "loss": 0.9877,
+      "step": 5732
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.821958320387983e-06,
+      "loss": 0.9706,
+      "step": 5733
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.821509557459216e-06,
+      "loss": 1.2527,
+      "step": 5734
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.821060720489569e-06,
+      "loss": 1.0429,
+      "step": 5735
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.820611809487736e-06,
+      "loss": 0.8352,
+      "step": 5736
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.820162824462414e-06,
+      "loss": 1.054,
+      "step": 5737
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.819713765422307e-06,
+      "loss": 1.0496,
+      "step": 5738
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.819264632376111e-06,
+      "loss": 1.1133,
+      "step": 5739
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81881542533253e-06,
+      "loss": 1.0588,
+      "step": 5740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.818366144300266e-06,
+      "loss": 0.9266,
+      "step": 5741
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.817916789288024e-06,
+      "loss": 0.9835,
+      "step": 5742
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.817467360304508e-06,
+      "loss": 0.9894,
+      "step": 5743
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.817017857358432e-06,
+      "loss": 0.7488,
+      "step": 5744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.816568280458498e-06,
+      "loss": 0.917,
+      "step": 5745
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81611862961342e-06,
+      "loss": 0.9057,
+      "step": 5746
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81566890483191e-06,
+      "loss": 0.7782,
+      "step": 5747
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81521910612268e-06,
+      "loss": 0.7438,
+      "step": 5748
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.814769233494445e-06,
+      "loss": 0.9387,
+      "step": 5749
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81431928695592e-06,
+      "loss": 1.0088,
+      "step": 5750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.813869266515826e-06,
+      "loss": 0.9597,
+      "step": 5751
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81341917218288e-06,
+      "loss": 0.9666,
+      "step": 5752
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.812969003965802e-06,
+      "loss": 0.9357,
+      "step": 5753
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.812518761873314e-06,
+      "loss": 1.0259,
+      "step": 5754
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.812068445914141e-06,
+      "loss": 0.9689,
+      "step": 5755
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.811618056097006e-06,
+      "loss": 0.9679,
+      "step": 5756
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.811167592430636e-06,
+      "loss": 1.1928,
+      "step": 5757
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.810717054923758e-06,
+      "loss": 0.9542,
+      "step": 5758
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.810266443585101e-06,
+      "loss": 1.0388,
+      "step": 5759
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.809815758423397e-06,
+      "loss": 0.888,
+      "step": 5760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.809364999447377e-06,
+      "loss": 0.946,
+      "step": 5761
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.808914166665773e-06,
+      "loss": 1.0046,
+      "step": 5762
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.808463260087322e-06,
+      "loss": 0.9938,
+      "step": 5763
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.80801227972076e-06,
+      "loss": 1.0126,
+      "step": 5764
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.807561225574822e-06,
+      "loss": 0.8473,
+      "step": 5765
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.80711009765825e-06,
+      "loss": 0.8416,
+      "step": 5766
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.806658895979785e-06,
+      "loss": 1.0132,
+      "step": 5767
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.806207620548165e-06,
+      "loss": 0.9019,
+      "step": 5768
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.805756271372135e-06,
+      "loss": 0.9304,
+      "step": 5769
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.805304848460443e-06,
+      "loss": 0.9763,
+      "step": 5770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.804853351821832e-06,
+      "loss": 0.852,
+      "step": 5771
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.80440178146505e-06,
+      "loss": 0.8642,
+      "step": 5772
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.803950137398846e-06,
+      "loss": 0.8709,
+      "step": 5773
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.803498419631971e-06,
+      "loss": 0.96,
+      "step": 5774
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.803046628173177e-06,
+      "loss": 1.0257,
+      "step": 5775
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.802594763031219e-06,
+      "loss": 0.9302,
+      "step": 5776
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.802142824214847e-06,
+      "loss": 0.9925,
+      "step": 5777
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.801690811732824e-06,
+      "loss": 0.9393,
+      "step": 5778
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.801238725593902e-06,
+      "loss": 0.9014,
+      "step": 5779
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.800786565806842e-06,
+      "loss": 1.0298,
+      "step": 5780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.800334332380405e-06,
+      "loss": 0.8228,
+      "step": 5781
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79988202532335e-06,
+      "loss": 0.9113,
+      "step": 5782
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.799429644644445e-06,
+      "loss": 1.0664,
+      "step": 5783
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79897719035245e-06,
+      "loss": 1.0047,
+      "step": 5784
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.798524662456138e-06,
+      "loss": 0.9956,
+      "step": 5785
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79807206096427e-06,
+      "loss": 0.9954,
+      "step": 5786
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.797619385885616e-06,
+      "loss": 0.8977,
+      "step": 5787
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79716663722895e-06,
+      "loss": 0.9156,
+      "step": 5788
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79671381500304e-06,
+      "loss": 1.2498,
+      "step": 5789
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.796260919216663e-06,
+      "loss": 1.1351,
+      "step": 5790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79580794987859e-06,
+      "loss": 0.7759,
+      "step": 5791
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7953549069976e-06,
+      "loss": 0.7283,
+      "step": 5792
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.79490179058247e-06,
+      "loss": 0.9752,
+      "step": 5793
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.794448600641977e-06,
+      "loss": 0.9636,
+      "step": 5794
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.793995337184904e-06,
+      "loss": 0.9428,
+      "step": 5795
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.793542000220032e-06,
+      "loss": 0.9412,
+      "step": 5796
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.793088589756145e-06,
+      "loss": 0.9739,
+      "step": 5797
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.792635105802025e-06,
+      "loss": 0.9289,
+      "step": 5798
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.792181548366464e-06,
+      "loss": 0.9801,
+      "step": 5799
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.791727917458241e-06,
+      "loss": 0.9006,
+      "step": 5800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.791274213086153e-06,
+      "loss": 1.0174,
+      "step": 5801
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.790820435258985e-06,
+      "loss": 1.0095,
+      "step": 5802
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.790366583985532e-06,
+      "loss": 1.0939,
+      "step": 5803
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.789912659274586e-06,
+      "loss": 0.9372,
+      "step": 5804
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.789458661134943e-06,
+      "loss": 1.1088,
+      "step": 5805
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.789004589575397e-06,
+      "loss": 0.8573,
+      "step": 5806
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.788550444604747e-06,
+      "loss": 1.0153,
+      "step": 5807
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78809622623179e-06,
+      "loss": 0.8988,
+      "step": 5808
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78764193446533e-06,
+      "loss": 1.016,
+      "step": 5809
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.787187569314167e-06,
+      "loss": 1.0229,
+      "step": 5810
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.786733130787102e-06,
+      "loss": 0.9707,
+      "step": 5811
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.786278618892943e-06,
+      "loss": 1.0183,
+      "step": 5812
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.785824033640493e-06,
+      "loss": 1.0163,
+      "step": 5813
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.785369375038561e-06,
+      "loss": 1.0524,
+      "step": 5814
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.784914643095958e-06,
+      "loss": 0.9128,
+      "step": 5815
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.784459837821493e-06,
+      "loss": 1.0693,
+      "step": 5816
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.784004959223974e-06,
+      "loss": 0.981,
+      "step": 5817
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.783550007312219e-06,
+      "loss": 0.9318,
+      "step": 5818
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78309498209504e-06,
+      "loss": 0.9582,
+      "step": 5819
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.782639883581255e-06,
+      "loss": 0.9473,
+      "step": 5820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78218471177968e-06,
+      "loss": 1.1237,
+      "step": 5821
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.781729466699133e-06,
+      "loss": 0.9229,
+      "step": 5822
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.781274148348438e-06,
+      "loss": 0.9469,
+      "step": 5823
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.780818756736411e-06,
+      "loss": 0.9422,
+      "step": 5824
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78036329187188e-06,
+      "loss": 0.8633,
+      "step": 5825
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.779907753763667e-06,
+      "loss": 0.9825,
+      "step": 5826
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7794521424206e-06,
+      "loss": 0.9631,
+      "step": 5827
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.778996457851506e-06,
+      "loss": 0.8492,
+      "step": 5828
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77854070006521e-06,
+      "loss": 0.8536,
+      "step": 5829
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.778084869070547e-06,
+      "loss": 0.9936,
+      "step": 5830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.777628964876345e-06,
+      "loss": 0.9101,
+      "step": 5831
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.777172987491438e-06,
+      "loss": 0.9424,
+      "step": 5832
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.776716936924663e-06,
+      "loss": 0.8994,
+      "step": 5833
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.776260813184855e-06,
+      "loss": 0.9543,
+      "step": 5834
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77580461628085e-06,
+      "loss": 0.9215,
+      "step": 5835
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.775348346221485e-06,
+      "loss": 1.0494,
+      "step": 5836
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.774892003015604e-06,
+      "loss": 0.9783,
+      "step": 5837
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.774435586672046e-06,
+      "loss": 0.8563,
+      "step": 5838
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.773979097199653e-06,
+      "loss": 0.9097,
+      "step": 5839
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.773522534607276e-06,
+      "loss": 0.9701,
+      "step": 5840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77306589890375e-06,
+      "loss": 0.986,
+      "step": 5841
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.772609190097932e-06,
+      "loss": 1.0114,
+      "step": 5842
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.772152408198665e-06,
+      "loss": 0.7984,
+      "step": 5843
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.771695553214801e-06,
+      "loss": 1.0883,
+      "step": 5844
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.771238625155191e-06,
+      "loss": 1.0183,
+      "step": 5845
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77078162402869e-06,
+      "loss": 1.016,
+      "step": 5846
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.770324549844149e-06,
+      "loss": 1.0245,
+      "step": 5847
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.769867402610425e-06,
+      "loss": 1.0195,
+      "step": 5848
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.769410182336375e-06,
+      "loss": 1.0027,
+      "step": 5849
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.768952889030859e-06,
+      "loss": 0.996,
+      "step": 5850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.768495522702733e-06,
+      "loss": 0.9009,
+      "step": 5851
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.768038083360863e-06,
+      "loss": 1.0855,
+      "step": 5852
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.767580571014111e-06,
+      "loss": 0.8623,
+      "step": 5853
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.767122985671339e-06,
+      "loss": 0.9525,
+      "step": 5854
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.766665327341413e-06,
+      "loss": 1.089,
+      "step": 5855
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7662075960332e-06,
+      "loss": 1.0171,
+      "step": 5856
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76574979175557e-06,
+      "loss": 1.0687,
+      "step": 5857
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.765291914517395e-06,
+      "loss": 1.0418,
+      "step": 5858
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.764833964327539e-06,
+      "loss": 1.0856,
+      "step": 5859
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.764375941194881e-06,
+      "loss": 1.0478,
+      "step": 5860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.763917845128293e-06,
+      "loss": 1.0645,
+      "step": 5861
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76345967613665e-06,
+      "loss": 0.947,
+      "step": 5862
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76300143422883e-06,
+      "loss": 0.9821,
+      "step": 5863
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76254311941371e-06,
+      "loss": 1.0555,
+      "step": 5864
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.762084731700172e-06,
+      "loss": 0.9241,
+      "step": 5865
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.761626271097095e-06,
+      "loss": 1.0881,
+      "step": 5866
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.761167737613361e-06,
+      "loss": 0.9518,
+      "step": 5867
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.760709131257855e-06,
+      "loss": 0.8916,
+      "step": 5868
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.760250452039464e-06,
+      "loss": 0.9318,
+      "step": 5869
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.759791699967074e-06,
+      "loss": 0.8043,
+      "step": 5870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.759332875049572e-06,
+      "loss": 0.9573,
+      "step": 5871
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.758873977295848e-06,
+      "loss": 0.9635,
+      "step": 5872
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.758415006714794e-06,
+      "loss": 0.9254,
+      "step": 5873
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7579559633153e-06,
+      "loss": 0.9475,
+      "step": 5874
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.757496847106264e-06,
+      "loss": 0.9057,
+      "step": 5875
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.757037658096579e-06,
+      "loss": 0.9072,
+      "step": 5876
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.756578396295142e-06,
+      "loss": 1.0275,
+      "step": 5877
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.756119061710849e-06,
+      "loss": 0.8753,
+      "step": 5878
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.755659654352599e-06,
+      "loss": 1.1708,
+      "step": 5879
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.755200174229299e-06,
+      "loss": 1.031,
+      "step": 5880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.754740621349847e-06,
+      "loss": 1.0747,
+      "step": 5881
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.754280995723146e-06,
+      "loss": 1.0249,
+      "step": 5882
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.753821297358102e-06,
+      "loss": 0.8204,
+      "step": 5883
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.753361526263622e-06,
+      "loss": 0.9402,
+      "step": 5884
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.752901682448615e-06,
+      "loss": 0.8857,
+      "step": 5885
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.752441765921989e-06,
+      "loss": 1.1368,
+      "step": 5886
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.751981776692655e-06,
+      "loss": 0.8714,
+      "step": 5887
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.751521714769525e-06,
+      "loss": 1.0218,
+      "step": 5888
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.75106158016151e-06,
+      "loss": 1.0356,
+      "step": 5889
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.750601372877532e-06,
+      "loss": 1.0346,
+      "step": 5890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7501410929265e-06,
+      "loss": 1.0193,
+      "step": 5891
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.749680740317337e-06,
+      "loss": 1.0914,
+      "step": 5892
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.749220315058959e-06,
+      "loss": 0.9517,
+      "step": 5893
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.74875981716029e-06,
+      "loss": 0.8424,
+      "step": 5894
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.748299246630247e-06,
+      "loss": 1.0879,
+      "step": 5895
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.747838603477758e-06,
+      "loss": 0.8376,
+      "step": 5896
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.747377887711746e-06,
+      "loss": 1.1095,
+      "step": 5897
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.746917099341138e-06,
+      "loss": 0.9728,
+      "step": 5898
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.746456238374861e-06,
+      "loss": 0.8844,
+      "step": 5899
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.745995304821844e-06,
+      "loss": 1.0715,
+      "step": 5900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.745534298691016e-06,
+      "loss": 0.9661,
+      "step": 5901
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.745073219991312e-06,
+      "loss": 0.9974,
+      "step": 5902
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.744612068731666e-06,
+      "loss": 0.8749,
+      "step": 5903
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.744150844921006e-06,
+      "loss": 0.9066,
+      "step": 5904
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.743689548568277e-06,
+      "loss": 1.0851,
+      "step": 5905
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.74322817968241e-06,
+      "loss": 1.1069,
+      "step": 5906
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.742766738272347e-06,
+      "loss": 0.7665,
+      "step": 5907
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.742305224347028e-06,
+      "loss": 1.0465,
+      "step": 5908
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.741843637915392e-06,
+      "loss": 0.8663,
+      "step": 5909
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.741381978986385e-06,
+      "loss": 1.0405,
+      "step": 5910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.740920247568951e-06,
+      "loss": 0.8342,
+      "step": 5911
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.740458443672038e-06,
+      "loss": 0.8262,
+      "step": 5912
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.739996567304589e-06,
+      "loss": 0.9809,
+      "step": 5913
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.739534618475555e-06,
+      "loss": 1.0284,
+      "step": 5914
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.739072597193885e-06,
+      "loss": 0.8957,
+      "step": 5915
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.738610503468534e-06,
+      "loss": 1.1894,
+      "step": 5916
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.73814833730845e-06,
+      "loss": 1.0226,
+      "step": 5917
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.737686098722592e-06,
+      "loss": 1.0571,
+      "step": 5918
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.737223787719914e-06,
+      "loss": 0.8166,
+      "step": 5919
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.73676140430937e-06,
+      "loss": 0.8067,
+      "step": 5920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.736298948499924e-06,
+      "loss": 1.024,
+      "step": 5921
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.735836420300532e-06,
+      "loss": 1.0502,
+      "step": 5922
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.735373819720159e-06,
+      "loss": 1.0735,
+      "step": 5923
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.734911146767764e-06,
+      "loss": 1.0144,
+      "step": 5924
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.73444840145231e-06,
+      "loss": 1.0974,
+      "step": 5925
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.73398558378277e-06,
+      "loss": 0.8161,
+      "step": 5926
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.733522693768104e-06,
+      "loss": 1.0231,
+      "step": 5927
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.733059731417284e-06,
+      "loss": 0.9614,
+      "step": 5928
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.732596696739278e-06,
+      "loss": 0.8797,
+      "step": 5929
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.732133589743057e-06,
+      "loss": 0.9153,
+      "step": 5930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.731670410437593e-06,
+      "loss": 1.0355,
+      "step": 5931
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.731207158831865e-06,
+      "loss": 0.9739,
+      "step": 5932
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.730743834934842e-06,
+      "loss": 0.7561,
+      "step": 5933
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.730280438755505e-06,
+      "loss": 1.0364,
+      "step": 5934
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.729816970302829e-06,
+      "loss": 0.8953,
+      "step": 5935
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.729353429585796e-06,
+      "loss": 0.9651,
+      "step": 5936
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.728889816613386e-06,
+      "loss": 0.8742,
+      "step": 5937
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.728426131394582e-06,
+      "loss": 0.9814,
+      "step": 5938
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.727962373938368e-06,
+      "loss": 1.0909,
+      "step": 5939
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.727498544253728e-06,
+      "loss": 0.9303,
+      "step": 5940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.727034642349647e-06,
+      "loss": 0.9848,
+      "step": 5941
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.726570668235119e-06,
+      "loss": 1.1392,
+      "step": 5942
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.726106621919128e-06,
+      "loss": 1.1902,
+      "step": 5943
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.725642503410666e-06,
+      "loss": 0.9553,
+      "step": 5944
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.725178312718727e-06,
+      "loss": 1.1392,
+      "step": 5945
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.724714049852302e-06,
+      "loss": 1.0965,
+      "step": 5946
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.724249714820387e-06,
+      "loss": 0.9573,
+      "step": 5947
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.723785307631978e-06,
+      "loss": 1.0313,
+      "step": 5948
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.723320828296075e-06,
+      "loss": 0.9964,
+      "step": 5949
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.722856276821675e-06,
+      "loss": 0.925,
+      "step": 5950
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.72239165321778e-06,
+      "loss": 1.1381,
+      "step": 5951
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.72192695749339e-06,
+      "loss": 0.8435,
+      "step": 5952
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.72146218965751e-06,
+      "loss": 0.9752,
+      "step": 5953
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.720997349719145e-06,
+      "loss": 1.0091,
+      "step": 5954
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.720532437687299e-06,
+      "loss": 1.1112,
+      "step": 5955
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.72006745357098e-06,
+      "loss": 0.8306,
+      "step": 5956
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.7196023973792e-06,
+      "loss": 1.0553,
+      "step": 5957
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.719137269120967e-06,
+      "loss": 0.9914,
+      "step": 5958
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.718672068805291e-06,
+      "loss": 1.0454,
+      "step": 5959
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71820679644119e-06,
+      "loss": 0.8878,
+      "step": 5960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.717741452037674e-06,
+      "loss": 0.9284,
+      "step": 5961
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71727603560376e-06,
+      "loss": 1.0279,
+      "step": 5962
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71681054714847e-06,
+      "loss": 0.7147,
+      "step": 5963
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.716344986680817e-06,
+      "loss": 0.9513,
+      "step": 5964
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71587935420982e-06,
+      "loss": 0.9795,
+      "step": 5965
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.715413649744507e-06,
+      "loss": 1.0423,
+      "step": 5966
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.714947873293895e-06,
+      "loss": 1.0417,
+      "step": 5967
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71448202486701e-06,
+      "loss": 0.8868,
+      "step": 5968
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71401610447288e-06,
+      "loss": 1.2206,
+      "step": 5969
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71355011212053e-06,
+      "loss": 0.9365,
+      "step": 5970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71308404781899e-06,
+      "loss": 0.9319,
+      "step": 5971
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.712617911577288e-06,
+      "loss": 1.1402,
+      "step": 5972
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.712151703404456e-06,
+      "loss": 1.0609,
+      "step": 5973
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.711685423309525e-06,
+      "loss": 1.1143,
+      "step": 5974
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.711219071301534e-06,
+      "loss": 0.9713,
+      "step": 5975
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.710752647389514e-06,
+      "loss": 1.0359,
+      "step": 5976
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.710286151582503e-06,
+      "loss": 1.1279,
+      "step": 5977
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.709819583889541e-06,
+      "loss": 1.1292,
+      "step": 5978
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.709352944319664e-06,
+      "loss": 1.2335,
+      "step": 5979
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.708886232881916e-06,
+      "loss": 1.0706,
+      "step": 5980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.708419449585339e-06,
+      "loss": 0.8543,
+      "step": 5981
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.707952594438975e-06,
+      "loss": 0.9473,
+      "step": 5982
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70748566745187e-06,
+      "loss": 1.0093,
+      "step": 5983
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.707018668633074e-06,
+      "loss": 1.0183,
+      "step": 5984
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.706551597991632e-06,
+      "loss": 0.7872,
+      "step": 5985
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.706084455536592e-06,
+      "loss": 0.9374,
+      "step": 5986
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.705617241277006e-06,
+      "loss": 0.8881,
+      "step": 5987
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.705149955221928e-06,
+      "loss": 1.0118,
+      "step": 5988
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70468259738041e-06,
+      "loss": 0.8828,
+      "step": 5989
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.704215167761506e-06,
+      "loss": 0.9557,
+      "step": 5990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.703747666374273e-06,
+      "loss": 1.0492,
+      "step": 5991
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70328009322777e-06,
+      "loss": 0.8842,
+      "step": 5992
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.702812448331055e-06,
+      "loss": 1.1255,
+      "step": 5993
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.702344731693189e-06,
+      "loss": 0.801,
+      "step": 5994
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.701876943323234e-06,
+      "loss": 1.0783,
+      "step": 5995
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.701409083230253e-06,
+      "loss": 0.847,
+      "step": 5996
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70094115142331e-06,
+      "loss": 1.0753,
+      "step": 5997
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.700473147911472e-06,
+      "loss": 0.955,
+      "step": 5998
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.700005072703806e-06,
+      "loss": 0.9615,
+      "step": 5999
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.69953692580938e-06,
+      "loss": 0.89,
+      "step": 6000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.699068707237268e-06,
+      "loss": 0.9011,
+      "step": 6001
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.698600416996536e-06,
+      "loss": 0.9748,
+      "step": 6002
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.698132055096262e-06,
+      "loss": 0.9384,
+      "step": 6003
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.697663621545516e-06,
+      "loss": 1.0825,
+      "step": 6004
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.697195116353379e-06,
+      "loss": 0.9074,
+      "step": 6005
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.696726539528924e-06,
+      "loss": 1.1129,
+      "step": 6006
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.69625789108123e-06,
+      "loss": 1.0755,
+      "step": 6007
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.695789171019378e-06,
+      "loss": 0.9627,
+      "step": 6008
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.69532037935245e-06,
+      "loss": 0.9211,
+      "step": 6009
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.694851516089526e-06,
+      "loss": 0.9167,
+      "step": 6010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.694382581239695e-06,
+      "loss": 0.9633,
+      "step": 6011
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.693913574812039e-06,
+      "loss": 0.9778,
+      "step": 6012
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.693444496815645e-06,
+      "loss": 1.0541,
+      "step": 6013
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.6929753472596e-06,
+      "loss": 0.9258,
+      "step": 6014
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.692506126152996e-06,
+      "loss": 0.8574,
+      "step": 6015
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.692036833504924e-06,
+      "loss": 1.0496,
+      "step": 6016
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.691567469324476e-06,
+      "loss": 0.9158,
+      "step": 6017
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.691098033620743e-06,
+      "loss": 0.8824,
+      "step": 6018
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.690628526402827e-06,
+      "loss": 1.131,
+      "step": 6019
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.690158947679818e-06,
+      "loss": 1.0208,
+      "step": 6020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.689689297460818e-06,
+      "loss": 1.0437,
+      "step": 6021
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.689219575754923e-06,
+      "loss": 1.026,
+      "step": 6022
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.688749782571237e-06,
+      "loss": 0.8958,
+      "step": 6023
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68827991791886e-06,
+      "loss": 0.8911,
+      "step": 6024
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.687809981806897e-06,
+      "loss": 0.979,
+      "step": 6025
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.687339974244452e-06,
+      "loss": 1.1933,
+      "step": 6026
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.686869895240631e-06,
+      "loss": 0.794,
+      "step": 6027
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.686399744804544e-06,
+      "loss": 0.9416,
+      "step": 6028
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.685929522945296e-06,
+      "loss": 0.9528,
+      "step": 6029
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.685459229671999e-06,
+      "loss": 1.0313,
+      "step": 6030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.684988864993767e-06,
+      "loss": 0.9315,
+      "step": 6031
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.684518428919711e-06,
+      "loss": 1.0833,
+      "step": 6032
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.684047921458946e-06,
+      "loss": 0.8125,
+      "step": 6033
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.683577342620588e-06,
+      "loss": 0.9644,
+      "step": 6034
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.683106692413754e-06,
+      "loss": 0.9985,
+      "step": 6035
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.682635970847566e-06,
+      "loss": 1.0712,
+      "step": 6036
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.682165177931138e-06,
+      "loss": 0.8684,
+      "step": 6037
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.681694313673597e-06,
+      "loss": 0.9873,
+      "step": 6038
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.681223378084065e-06,
+      "loss": 1.0023,
+      "step": 6039
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.680752371171664e-06,
+      "loss": 1.0057,
+      "step": 6040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68028129294552e-06,
+      "loss": 0.9398,
+      "step": 6041
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67981014341476e-06,
+      "loss": 1.0785,
+      "step": 6042
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.679338922588516e-06,
+      "loss": 0.9806,
+      "step": 6043
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.678867630475912e-06,
+      "loss": 0.9701,
+      "step": 6044
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.678396267086084e-06,
+      "loss": 0.9464,
+      "step": 6045
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.677924832428162e-06,
+      "loss": 0.8261,
+      "step": 6046
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67745332651128e-06,
+      "loss": 1.0742,
+      "step": 6047
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.676981749344575e-06,
+      "loss": 0.8542,
+      "step": 6048
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.676510100937184e-06,
+      "loss": 1.0638,
+      "step": 6049
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.676038381298241e-06,
+      "loss": 1.1625,
+      "step": 6050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67556659043689e-06,
+      "loss": 0.9681,
+      "step": 6051
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67509472836227e-06,
+      "loss": 0.9245,
+      "step": 6052
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.674622795083522e-06,
+      "loss": 1.0627,
+      "step": 6053
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.674150790609793e-06,
+      "loss": 0.982,
+      "step": 6054
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.673678714950224e-06,
+      "loss": 0.9621,
+      "step": 6055
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.673206568113963e-06,
+      "loss": 0.9009,
+      "step": 6056
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67273435011016e-06,
+      "loss": 1.0305,
+      "step": 6057
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67226206094796e-06,
+      "loss": 0.9543,
+      "step": 6058
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.671789700636517e-06,
+      "loss": 0.8785,
+      "step": 6059
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67131726918498e-06,
+      "loss": 1.0737,
+      "step": 6060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.670844766602506e-06,
+      "loss": 0.9429,
+      "step": 6061
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.670372192898245e-06,
+      "loss": 1.1807,
+      "step": 6062
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.669899548081356e-06,
+      "loss": 0.8381,
+      "step": 6063
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.669426832160997e-06,
+      "loss": 1.0846,
+      "step": 6064
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.668954045146325e-06,
+      "loss": 0.9657,
+      "step": 6065
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.6684811870465e-06,
+      "loss": 1.0037,
+      "step": 6066
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.668008257870684e-06,
+      "loss": 0.9729,
+      "step": 6067
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.66753525762804e-06,
+      "loss": 0.8625,
+      "step": 6068
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.667062186327733e-06,
+      "loss": 1.1213,
+      "step": 6069
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.666589043978927e-06,
+      "loss": 1.0203,
+      "step": 6070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.666115830590792e-06,
+      "loss": 0.9726,
+      "step": 6071
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.665642546172493e-06,
+      "loss": 1.0314,
+      "step": 6072
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.6651691907332e-06,
+      "loss": 1.1246,
+      "step": 6073
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.664695764282087e-06,
+      "loss": 0.9353,
+      "step": 6074
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.664222266828324e-06,
+      "loss": 1.2144,
+      "step": 6075
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.663748698381085e-06,
+      "loss": 0.9852,
+      "step": 6076
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.663275058949548e-06,
+      "loss": 0.7578,
+      "step": 6077
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.662801348542888e-06,
+      "loss": 1.115,
+      "step": 6078
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.66232756717028e-06,
+      "loss": 0.9643,
+      "step": 6079
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.661853714840909e-06,
+      "loss": 1.0674,
+      "step": 6080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.66137979156395e-06,
+      "loss": 0.9339,
+      "step": 6081
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.66090579734859e-06,
+      "loss": 0.8842,
+      "step": 6082
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.660431732204012e-06,
+      "loss": 0.8797,
+      "step": 6083
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.659957596139397e-06,
+      "loss": 0.9707,
+      "step": 6084
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.659483389163935e-06,
+      "loss": 1.019,
+      "step": 6085
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.659009111286813e-06,
+      "loss": 1.0936,
+      "step": 6086
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.658534762517218e-06,
+      "loss": 0.9915,
+      "step": 6087
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.658060342864343e-06,
+      "loss": 0.8587,
+      "step": 6088
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.657585852337378e-06,
+      "loss": 0.9216,
+      "step": 6089
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.657111290945516e-06,
+      "loss": 0.9506,
+      "step": 6090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.656636658697954e-06,
+      "loss": 0.8985,
+      "step": 6091
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.656161955603886e-06,
+      "loss": 0.9255,
+      "step": 6092
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.655687181672508e-06,
+      "loss": 0.969,
+      "step": 6093
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.65521233691302e-06,
+      "loss": 1.0635,
+      "step": 6094
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.654737421334625e-06,
+      "loss": 0.9514,
+      "step": 6095
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.654262434946518e-06,
+      "loss": 1.1836,
+      "step": 6096
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.653787377757907e-06,
+      "loss": 1.0438,
+      "step": 6097
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.653312249777995e-06,
+      "loss": 1.0991,
+      "step": 6098
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.652837051015985e-06,
+      "loss": 1.1398,
+      "step": 6099
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.652361781481088e-06,
+      "loss": 1.1856,
+      "step": 6100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.651886441182509e-06,
+      "loss": 0.9568,
+      "step": 6101
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.651411030129458e-06,
+      "loss": 0.9976,
+      "step": 6102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.650935548331147e-06,
+      "loss": 1.1033,
+      "step": 6103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.650459995796789e-06,
+      "loss": 1.0921,
+      "step": 6104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.649984372535595e-06,
+      "loss": 0.9414,
+      "step": 6105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.649508678556782e-06,
+      "loss": 0.9536,
+      "step": 6106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.649032913869567e-06,
+      "loss": 1.0919,
+      "step": 6107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.648557078483167e-06,
+      "loss": 0.7659,
+      "step": 6108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.6480811724068e-06,
+      "loss": 1.1305,
+      "step": 6109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.647605195649688e-06,
+      "loss": 1.128,
+      "step": 6110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.647129148221054e-06,
+      "loss": 0.8735,
+      "step": 6111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.646653030130118e-06,
+      "loss": 0.9996,
+      "step": 6112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.646176841386109e-06,
+      "loss": 0.9638,
+      "step": 6113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64570058199825e-06,
+      "loss": 1.0108,
+      "step": 6114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.645224251975768e-06,
+      "loss": 1.0167,
+      "step": 6115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.644747851327895e-06,
+      "loss": 0.9942,
+      "step": 6116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.644271380063858e-06,
+      "loss": 0.9499,
+      "step": 6117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64379483819289e-06,
+      "loss": 0.7494,
+      "step": 6118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.643318225724224e-06,
+      "loss": 0.8473,
+      "step": 6119
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.642841542667092e-06,
+      "loss": 1.056,
+      "step": 6120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.642364789030733e-06,
+      "loss": 0.9607,
+      "step": 6121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.641887964824383e-06,
+      "loss": 0.9365,
+      "step": 6122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64141107005728e-06,
+      "loss": 0.8357,
+      "step": 6123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.640934104738662e-06,
+      "loss": 1.0291,
+      "step": 6124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.640457068877772e-06,
+      "loss": 1.041,
+      "step": 6125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.639979962483853e-06,
+      "loss": 1.0052,
+      "step": 6126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.639502785566146e-06,
+      "loss": 1.0767,
+      "step": 6127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.639025538133899e-06,
+      "loss": 0.9772,
+      "step": 6128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.638548220196359e-06,
+      "loss": 1.1133,
+      "step": 6129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63807083176277e-06,
+      "loss": 1.0305,
+      "step": 6130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.637593372842386e-06,
+      "loss": 0.853,
+      "step": 6131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.637115843444455e-06,
+      "loss": 0.9487,
+      "step": 6132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.636638243578229e-06,
+      "loss": 0.877,
+      "step": 6133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.636160573252963e-06,
+      "loss": 1.0263,
+      "step": 6134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63568283247791e-06,
+      "loss": 0.8854,
+      "step": 6135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.635205021262328e-06,
+      "loss": 1.0947,
+      "step": 6136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.634727139615471e-06,
+      "loss": 1.0667,
+      "step": 6137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.634249187546601e-06,
+      "loss": 1.0194,
+      "step": 6138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.633771165064979e-06,
+      "loss": 1.079,
+      "step": 6139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.633293072179863e-06,
+      "loss": 0.9606,
+      "step": 6140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63281490890052e-06,
+      "loss": 1.0736,
+      "step": 6141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.632336675236211e-06,
+      "loss": 0.9155,
+      "step": 6142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.631858371196202e-06,
+      "loss": 0.9095,
+      "step": 6143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63137999678976e-06,
+      "loss": 1.0972,
+      "step": 6144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63090155202616e-06,
+      "loss": 0.9407,
+      "step": 6145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.630423036914659e-06,
+      "loss": 0.9135,
+      "step": 6146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.629944451464538e-06,
+      "loss": 0.9931,
+      "step": 6147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.629465795685068e-06,
+      "loss": 0.9539,
+      "step": 6148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.628987069585521e-06,
+      "loss": 0.8484,
+      "step": 6149
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.628508273175172e-06,
+      "loss": 0.8297,
+      "step": 6150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.628029406463299e-06,
+      "loss": 1.0377,
+      "step": 6151
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.627550469459177e-06,
+      "loss": 0.9889,
+      "step": 6152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.627071462172089e-06,
+      "loss": 1.0643,
+      "step": 6153
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.626592384611315e-06,
+      "loss": 0.9411,
+      "step": 6154
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.626113236786135e-06,
+      "loss": 0.9617,
+      "step": 6155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.625634018705836e-06,
+      "loss": 0.8746,
+      "step": 6156
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.625154730379696e-06,
+      "loss": 0.9899,
+      "step": 6157
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.624675371817008e-06,
+      "loss": 0.9712,
+      "step": 6158
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.624195943027057e-06,
+      "loss": 1.2314,
+      "step": 6159
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.623716444019132e-06,
+      "loss": 0.8073,
+      "step": 6160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.623236874802522e-06,
+      "loss": 0.9601,
+      "step": 6161
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.62275723538652e-06,
+      "loss": 0.8879,
+      "step": 6162
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.622277525780417e-06,
+      "loss": 0.9257,
+      "step": 6163
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.62179774599351e-06,
+      "loss": 0.8088,
+      "step": 6164
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.621317896035092e-06,
+      "loss": 0.9729,
+      "step": 6165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.620837975914465e-06,
+      "loss": 0.9428,
+      "step": 6166
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.62035798564092e-06,
+      "loss": 1.0092,
+      "step": 6167
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61987792522376e-06,
+      "loss": 1.0331,
+      "step": 6168
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.619397794672287e-06,
+      "loss": 0.8741,
+      "step": 6169
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.618917593995803e-06,
+      "loss": 1.0802,
+      "step": 6170
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.618437323203612e-06,
+      "loss": 0.8581,
+      "step": 6171
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.617956982305018e-06,
+      "loss": 0.881,
+      "step": 6172
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61747657130933e-06,
+      "loss": 1.0428,
+      "step": 6173
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.616996090225851e-06,
+      "loss": 0.9208,
+      "step": 6174
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.616515539063894e-06,
+      "loss": 0.9887,
+      "step": 6175
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61603491783277e-06,
+      "loss": 0.8416,
+      "step": 6176
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61555422654179e-06,
+      "loss": 1.0506,
+      "step": 6177
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.615073465200265e-06,
+      "loss": 1.1197,
+      "step": 6178
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.614592633817514e-06,
+      "loss": 0.9936,
+      "step": 6179
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61411173240285e-06,
+      "loss": 1.0352,
+      "step": 6180
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61363076096559e-06,
+      "loss": 0.9375,
+      "step": 6181
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.613149719515054e-06,
+      "loss": 1.0313,
+      "step": 6182
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.612668608060562e-06,
+      "loss": 0.9811,
+      "step": 6183
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.612187426611435e-06,
+      "loss": 0.8585,
+      "step": 6184
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.611706175176995e-06,
+      "loss": 0.8925,
+      "step": 6185
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.611224853766568e-06,
+      "loss": 0.9607,
+      "step": 6186
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61074346238948e-06,
+      "loss": 1.0711,
+      "step": 6187
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.610262001055055e-06,
+      "loss": 1.0874,
+      "step": 6188
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.609780469772623e-06,
+      "loss": 0.919,
+      "step": 6189
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.609298868551513e-06,
+      "loss": 0.9116,
+      "step": 6190
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.608817197401057e-06,
+      "loss": 0.7857,
+      "step": 6191
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.608335456330585e-06,
+      "loss": 1.1668,
+      "step": 6192
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.607853645349433e-06,
+      "loss": 1.1341,
+      "step": 6193
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.607371764466936e-06,
+      "loss": 0.9554,
+      "step": 6194
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.606889813692426e-06,
+      "loss": 0.9058,
+      "step": 6195
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.606407793035248e-06,
+      "loss": 0.9538,
+      "step": 6196
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.605925702504735e-06,
+      "loss": 1.1017,
+      "step": 6197
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60544354211023e-06,
+      "loss": 1.0139,
+      "step": 6198
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.604961311861075e-06,
+      "loss": 0.9434,
+      "step": 6199
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60447901176661e-06,
+      "loss": 0.9942,
+      "step": 6200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.603996641836183e-06,
+      "loss": 0.864,
+      "step": 6201
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60351420207914e-06,
+      "loss": 1.0187,
+      "step": 6202
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.603031692504826e-06,
+      "loss": 0.9842,
+      "step": 6203
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.602549113122589e-06,
+      "loss": 1.0814,
+      "step": 6204
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60206646394178e-06,
+      "loss": 0.9989,
+      "step": 6205
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.601583744971753e-06,
+      "loss": 1.0034,
+      "step": 6206
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.601100956221855e-06,
+      "loss": 0.8327,
+      "step": 6207
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.600618097701443e-06,
+      "loss": 0.957,
+      "step": 6208
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.600135169419874e-06,
+      "loss": 0.9335,
+      "step": 6209
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.599652171386502e-06,
+      "loss": 0.9326,
+      "step": 6210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.599169103610685e-06,
+      "loss": 1.0334,
+      "step": 6211
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.598685966101783e-06,
+      "loss": 0.8881,
+      "step": 6212
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.59820275886916e-06,
+      "loss": 0.9632,
+      "step": 6213
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.597719481922171e-06,
+      "loss": 0.9604,
+      "step": 6214
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.597236135270184e-06,
+      "loss": 1.0032,
+      "step": 6215
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.596752718922562e-06,
+      "loss": 1.1512,
+      "step": 6216
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.596269232888674e-06,
+      "loss": 1.0447,
+      "step": 6217
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.595785677177883e-06,
+      "loss": 0.8917,
+      "step": 6218
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.595302051799564e-06,
+      "loss": 1.1215,
+      "step": 6219
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.59481835676308e-06,
+      "loss": 0.774,
+      "step": 6220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.594334592077807e-06,
+      "loss": 0.9032,
+      "step": 6221
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.593850757753116e-06,
+      "loss": 1.0644,
+      "step": 6222
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.593366853798382e-06,
+      "loss": 0.9341,
+      "step": 6223
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.592882880222979e-06,
+      "loss": 0.9794,
+      "step": 6224
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.592398837036287e-06,
+      "loss": 0.918,
+      "step": 6225
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.591914724247683e-06,
+      "loss": 1.0435,
+      "step": 6226
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.591430541866544e-06,
+      "loss": 0.9839,
+      "step": 6227
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.590946289902256e-06,
+      "loss": 0.9398,
+      "step": 6228
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.590461968364195e-06,
+      "loss": 0.9764,
+      "step": 6229
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.589977577261749e-06,
+      "loss": 0.9298,
+      "step": 6230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.589493116604301e-06,
+      "loss": 1.1082,
+      "step": 6231
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.58900858640124e-06,
+      "loss": 1.0546,
+      "step": 6232
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.58852398666195e-06,
+      "loss": 0.9093,
+      "step": 6233
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.588039317395822e-06,
+      "loss": 1.1139,
+      "step": 6234
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.587554578612246e-06,
+      "loss": 1.0138,
+      "step": 6235
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.587069770320615e-06,
+      "loss": 0.9126,
+      "step": 6236
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.58658489253032e-06,
+      "loss": 1.0102,
+      "step": 6237
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.586099945250757e-06,
+      "loss": 0.8719,
+      "step": 6238
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.58561492849132e-06,
+      "loss": 0.9831,
+      "step": 6239
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.585129842261409e-06,
+      "loss": 0.9794,
+      "step": 6240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.584644686570419e-06,
+      "loss": 1.0381,
+      "step": 6241
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.58415946142775e-06,
+      "loss": 1.002,
+      "step": 6242
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.583674166842806e-06,
+      "loss": 1.092,
+      "step": 6243
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.583188802824988e-06,
+      "loss": 0.9134,
+      "step": 6244
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5827033693837e-06,
+      "loss": 0.9424,
+      "step": 6245
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.582217866528346e-06,
+      "loss": 1.0919,
+      "step": 6246
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.581732294268335e-06,
+      "loss": 0.9418,
+      "step": 6247
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.581246652613071e-06,
+      "loss": 1.0678,
+      "step": 6248
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.580760941571968e-06,
+      "loss": 0.889,
+      "step": 6249
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.580275161154432e-06,
+      "loss": 0.9633,
+      "step": 6250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.579789311369878e-06,
+      "loss": 0.8597,
+      "step": 6251
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.579303392227718e-06,
+      "loss": 1.1411,
+      "step": 6252
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.578817403737365e-06,
+      "loss": 0.9007,
+      "step": 6253
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.578331345908239e-06,
+      "loss": 0.9878,
+      "step": 6254
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.577845218749754e-06,
+      "loss": 0.7969,
+      "step": 6255
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.57735902227133e-06,
+      "loss": 1.0937,
+      "step": 6256
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.576872756482385e-06,
+      "loss": 0.8526,
+      "step": 6257
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.576386421392343e-06,
+      "loss": 0.9095,
+      "step": 6258
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.575900017010624e-06,
+      "loss": 1.0659,
+      "step": 6259
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.575413543346656e-06,
+      "loss": 1.0468,
+      "step": 6260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.574927000409858e-06,
+      "loss": 0.9645,
+      "step": 6261
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.574440388209664e-06,
+      "loss": 0.8871,
+      "step": 6262
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.573953706755495e-06,
+      "loss": 0.7868,
+      "step": 6263
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.573466956056784e-06,
+      "loss": 0.9435,
+      "step": 6264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.572980136122964e-06,
+      "loss": 1.1538,
+      "step": 6265
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.572493246963463e-06,
+      "loss": 1.0639,
+      "step": 6266
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.572006288587715e-06,
+      "loss": 0.8934,
+      "step": 6267
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.571519261005155e-06,
+      "loss": 0.8654,
+      "step": 6268
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.571032164225222e-06,
+      "loss": 0.794,
+      "step": 6269
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.570544998257348e-06,
+      "loss": 0.9439,
+      "step": 6270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.570057763110976e-06,
+      "loss": 0.782,
+      "step": 6271
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.569570458795545e-06,
+      "loss": 0.9778,
+      "step": 6272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.569083085320495e-06,
+      "loss": 0.8317,
+      "step": 6273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56859564269527e-06,
+      "loss": 0.8528,
+      "step": 6274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.568108130929315e-06,
+      "loss": 0.9655,
+      "step": 6275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.567620550032075e-06,
+      "loss": 0.9367,
+      "step": 6276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.567132900012994e-06,
+      "loss": 0.9223,
+      "step": 6277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.566645180881523e-06,
+      "loss": 0.7879,
+      "step": 6278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56615739264711e-06,
+      "loss": 0.9932,
+      "step": 6279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.565669535319207e-06,
+      "loss": 1.0948,
+      "step": 6280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.565181608907266e-06,
+      "loss": 1.1043,
+      "step": 6281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.564693613420739e-06,
+      "loss": 0.9287,
+      "step": 6282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.564205548869082e-06,
+      "loss": 1.0503,
+      "step": 6283
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.563717415261752e-06,
+      "loss": 0.9853,
+      "step": 6284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.563229212608206e-06,
+      "loss": 1.0294,
+      "step": 6285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.562740940917901e-06,
+      "loss": 0.9743,
+      "step": 6286
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.562252600200299e-06,
+      "loss": 1.0875,
+      "step": 6287
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56176419046486e-06,
+      "loss": 1.148,
+      "step": 6288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.561275711721049e-06,
+      "loss": 0.996,
+      "step": 6289
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.560787163978329e-06,
+      "loss": 0.9322,
+      "step": 6290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.560298547246164e-06,
+      "loss": 1.0902,
+      "step": 6291
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.559809861534023e-06,
+      "loss": 0.9353,
+      "step": 6292
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.559321106851374e-06,
+      "loss": 0.9972,
+      "step": 6293
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.558832283207686e-06,
+      "loss": 0.9958,
+      "step": 6294
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.558343390612428e-06,
+      "loss": 0.9736,
+      "step": 6295
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.557854429075076e-06,
+      "loss": 1.0337,
+      "step": 6296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5573653986051e-06,
+      "loss": 1.0368,
+      "step": 6297
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.556876299211977e-06,
+      "loss": 0.8533,
+      "step": 6298
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.556387130905182e-06,
+      "loss": 0.981,
+      "step": 6299
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.555897893694194e-06,
+      "loss": 1.0385,
+      "step": 6300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.55540858758849e-06,
+      "loss": 1.01,
+      "step": 6301
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.554919212597552e-06,
+      "loss": 0.9506,
+      "step": 6302
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.55442976873086e-06,
+      "loss": 0.951,
+      "step": 6303
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5539402559979e-06,
+      "loss": 0.9005,
+      "step": 6304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.55345067440815e-06,
+      "loss": 0.9766,
+      "step": 6305
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.552961023971101e-06,
+      "loss": 1.0686,
+      "step": 6306
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.552471304696237e-06,
+      "loss": 1.05,
+      "step": 6307
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.551981516593048e-06,
+      "loss": 0.8731,
+      "step": 6308
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.551491659671023e-06,
+      "loss": 0.975,
+      "step": 6309
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.551001733939654e-06,
+      "loss": 0.865,
+      "step": 6310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.550511739408428e-06,
+      "loss": 1.0732,
+      "step": 6311
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.550021676086847e-06,
+      "loss": 0.9679,
+      "step": 6312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5495315439844e-06,
+      "loss": 0.8578,
+      "step": 6313
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.549041343110583e-06,
+      "loss": 1.0378,
+      "step": 6314
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.548551073474896e-06,
+      "loss": 0.9667,
+      "step": 6315
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.548060735086838e-06,
+      "loss": 0.9428,
+      "step": 6316
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.547570327955906e-06,
+      "loss": 1.1195,
+      "step": 6317
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.547079852091605e-06,
+      "loss": 1.0981,
+      "step": 6318
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.546589307503435e-06,
+      "loss": 0.865,
+      "step": 6319
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.546098694200904e-06,
+      "loss": 0.7697,
+      "step": 6320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.545608012193514e-06,
+      "loss": 1.0006,
+      "step": 6321
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.545117261490772e-06,
+      "loss": 0.9947,
+      "step": 6322
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.544626442102188e-06,
+      "loss": 0.7845,
+      "step": 6323
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.544135554037272e-06,
+      "loss": 1.1853,
+      "step": 6324
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.543644597305531e-06,
+      "loss": 0.959,
+      "step": 6325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.543153571916481e-06,
+      "loss": 0.9726,
+      "step": 6326
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.542662477879636e-06,
+      "loss": 0.8577,
+      "step": 6327
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.542171315204505e-06,
+      "loss": 1.0013,
+      "step": 6328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.541680083900613e-06,
+      "loss": 1.0037,
+      "step": 6329
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.54118878397747e-06,
+      "loss": 0.9123,
+      "step": 6330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.540697415444598e-06,
+      "loss": 0.9991,
+      "step": 6331
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.540205978311518e-06,
+      "loss": 0.9442,
+      "step": 6332
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.539714472587747e-06,
+      "loss": 0.9398,
+      "step": 6333
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.539222898282814e-06,
+      "loss": 1.0901,
+      "step": 6334
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.538731255406241e-06,
+      "loss": 1.0426,
+      "step": 6335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53823954396755e-06,
+      "loss": 0.807,
+      "step": 6336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.537747763976273e-06,
+      "loss": 0.9941,
+      "step": 6337
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.537255915441932e-06,
+      "loss": 0.7911,
+      "step": 6338
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.536763998374062e-06,
+      "loss": 1.021,
+      "step": 6339
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.536272012782191e-06,
+      "loss": 1.0112,
+      "step": 6340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.535779958675854e-06,
+      "loss": 0.9594,
+      "step": 6341
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.535287836064581e-06,
+      "loss": 1.0116,
+      "step": 6342
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.534795644957908e-06,
+      "loss": 0.923,
+      "step": 6343
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.534303385365372e-06,
+      "loss": 0.9837,
+      "step": 6344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.533811057296508e-06,
+      "loss": 0.8724,
+      "step": 6345
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.533318660760857e-06,
+      "loss": 0.7439,
+      "step": 6346
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53282619576796e-06,
+      "loss": 1.0876,
+      "step": 6347
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.532333662327355e-06,
+      "loss": 0.8703,
+      "step": 6348
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.531841060448588e-06,
+      "loss": 0.9631,
+      "step": 6349
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5313483901412e-06,
+      "loss": 0.9543,
+      "step": 6350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53085565141474e-06,
+      "loss": 1.0794,
+      "step": 6351
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.530362844278752e-06,
+      "loss": 1.1733,
+      "step": 6352
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.529869968742784e-06,
+      "loss": 0.8347,
+      "step": 6353
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.529377024816385e-06,
+      "loss": 0.9729,
+      "step": 6354
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.528884012509108e-06,
+      "loss": 0.9649,
+      "step": 6355
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.528390931830505e-06,
+      "loss": 0.9001,
+      "step": 6356
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.527897782790127e-06,
+      "loss": 0.9439,
+      "step": 6357
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52740456539753e-06,
+      "loss": 0.8223,
+      "step": 6358
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52691127966227e-06,
+      "loss": 1.0667,
+      "step": 6359
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.526417925593901e-06,
+      "loss": 1.0281,
+      "step": 6360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.525924503201987e-06,
+      "loss": 0.9289,
+      "step": 6361
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.525431012496085e-06,
+      "loss": 1.0364,
+      "step": 6362
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.524937453485758e-06,
+      "loss": 1.0511,
+      "step": 6363
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.524443826180567e-06,
+      "loss": 0.9533,
+      "step": 6364
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.523950130590076e-06,
+      "loss": 1.0002,
+      "step": 6365
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52345636672385e-06,
+      "loss": 0.996,
+      "step": 6366
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.522962534591457e-06,
+      "loss": 1.0149,
+      "step": 6367
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.522468634202463e-06,
+      "loss": 1.0979,
+      "step": 6368
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.52197466556644e-06,
+      "loss": 0.853,
+      "step": 6369
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.521480628692956e-06,
+      "loss": 0.9056,
+      "step": 6370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.520986523591584e-06,
+      "loss": 1.1165,
+      "step": 6371
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.520492350271895e-06,
+      "loss": 0.9339,
+      "step": 6372
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.519998108743468e-06,
+      "loss": 0.9113,
+      "step": 6373
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.519503799015873e-06,
+      "loss": 0.8064,
+      "step": 6374
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.519009421098691e-06,
+      "loss": 0.9274,
+      "step": 6375
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.518514975001502e-06,
+      "loss": 1.0156,
+      "step": 6376
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.518020460733882e-06,
+      "loss": 0.8821,
+      "step": 6377
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.517525878305415e-06,
+      "loss": 0.9115,
+      "step": 6378
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51703122772568e-06,
+      "loss": 0.8889,
+      "step": 6379
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.516536509004262e-06,
+      "loss": 0.858,
+      "step": 6380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51604172215075e-06,
+      "loss": 1.0011,
+      "step": 6381
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.515546867174726e-06,
+      "loss": 0.9988,
+      "step": 6382
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.515051944085778e-06,
+      "loss": 1.0689,
+      "step": 6383
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.514556952893496e-06,
+      "loss": 1.1627,
+      "step": 6384
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.514061893607471e-06,
+      "loss": 0.8037,
+      "step": 6385
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.513566766237293e-06,
+      "loss": 0.8275,
+      "step": 6386
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.513071570792558e-06,
+      "loss": 0.9212,
+      "step": 6387
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.512576307282857e-06,
+      "loss": 1.1395,
+      "step": 6388
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.512080975717785e-06,
+      "loss": 0.8994,
+      "step": 6389
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.511585576106942e-06,
+      "loss": 1.0937,
+      "step": 6390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.511090108459924e-06,
+      "loss": 0.9917,
+      "step": 6391
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.510594572786333e-06,
+      "loss": 1.0599,
+      "step": 6392
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51009896909577e-06,
+      "loss": 0.8507,
+      "step": 6393
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.509603297397834e-06,
+      "loss": 0.836,
+      "step": 6394
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.509107557702129e-06,
+      "loss": 1.1622,
+      "step": 6395
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.508611750018263e-06,
+      "loss": 0.8873,
+      "step": 6396
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.50811587435584e-06,
+      "loss": 0.931,
+      "step": 6397
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.507619930724466e-06,
+      "loss": 0.9,
+      "step": 6398
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.507123919133754e-06,
+      "loss": 1.0999,
+      "step": 6399
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.506627839593311e-06,
+      "loss": 1.0089,
+      "step": 6400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.506131692112749e-06,
+      "loss": 0.8378,
+      "step": 6401
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.505635476701679e-06,
+      "loss": 1.0441,
+      "step": 6402
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.505139193369718e-06,
+      "loss": 1.02,
+      "step": 6403
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.50464284212648e-06,
+      "loss": 0.9873,
+      "step": 6404
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.504146422981582e-06,
+      "loss": 1.1847,
+      "step": 6405
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.503649935944644e-06,
+      "loss": 1.1005,
+      "step": 6406
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.503153381025281e-06,
+      "loss": 1.1776,
+      "step": 6407
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.502656758233117e-06,
+      "loss": 0.9232,
+      "step": 6408
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.502160067577771e-06,
+      "loss": 0.9167,
+      "step": 6409
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.50166330906887e-06,
+      "loss": 0.9035,
+      "step": 6410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.501166482716036e-06,
+      "loss": 0.8805,
+      "step": 6411
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.500669588528897e-06,
+      "loss": 0.9096,
+      "step": 6412
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.500172626517076e-06,
+      "loss": 1.2153,
+      "step": 6413
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.499675596690206e-06,
+      "loss": 1.0858,
+      "step": 6414
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.499178499057914e-06,
+      "loss": 0.8115,
+      "step": 6415
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.498681333629834e-06,
+      "loss": 0.8822,
+      "step": 6416
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.498184100415596e-06,
+      "loss": 1.0431,
+      "step": 6417
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.497686799424834e-06,
+      "loss": 1.0685,
+      "step": 6418
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.497189430667186e-06,
+      "loss": 0.8751,
+      "step": 6419
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.496691994152282e-06,
+      "loss": 0.9269,
+      "step": 6420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.496194489889766e-06,
+      "loss": 1.0501,
+      "step": 6421
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.495696917889276e-06,
+      "loss": 0.9443,
+      "step": 6422
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49519927816045e-06,
+      "loss": 0.946,
+      "step": 6423
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49470157071293e-06,
+      "loss": 0.9981,
+      "step": 6424
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.494203795556362e-06,
+      "loss": 1.0162,
+      "step": 6425
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.493705952700387e-06,
+      "loss": 0.9337,
+      "step": 6426
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49320804215465e-06,
+      "loss": 0.9597,
+      "step": 6427
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.492710063928802e-06,
+      "loss": 0.8559,
+      "step": 6428
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.492212018032489e-06,
+      "loss": 1.0801,
+      "step": 6429
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.491713904475358e-06,
+      "loss": 0.9365,
+      "step": 6430
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.491215723267063e-06,
+      "loss": 0.8753,
+      "step": 6431
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.490717474417256e-06,
+      "loss": 0.8718,
+      "step": 6432
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.490219157935589e-06,
+      "loss": 0.9589,
+      "step": 6433
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.489720773831717e-06,
+      "loss": 0.8343,
+      "step": 6434
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.489222322115298e-06,
+      "loss": 0.9856,
+      "step": 6435
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.488723802795988e-06,
+      "loss": 0.9547,
+      "step": 6436
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.488225215883446e-06,
+      "loss": 0.9994,
+      "step": 6437
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48772656138733e-06,
+      "loss": 0.9187,
+      "step": 6438
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.487227839317305e-06,
+      "loss": 1.1138,
+      "step": 6439
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48672904968303e-06,
+      "loss": 0.945,
+      "step": 6440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.486230192494172e-06,
+      "loss": 0.8028,
+      "step": 6441
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.485731267760393e-06,
+      "loss": 0.9431,
+      "step": 6442
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.485232275491363e-06,
+      "loss": 1.0025,
+      "step": 6443
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.484733215696747e-06,
+      "loss": 0.9879,
+      "step": 6444
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.484234088386217e-06,
+      "loss": 0.9685,
+      "step": 6445
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.483734893569442e-06,
+      "loss": 0.9837,
+      "step": 6446
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48323563125609e-06,
+      "loss": 0.9771,
+      "step": 6447
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.482736301455842e-06,
+      "loss": 0.9124,
+      "step": 6448
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.482236904178367e-06,
+      "loss": 1.0712,
+      "step": 6449
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48173743943334e-06,
+      "loss": 0.9654,
+      "step": 6450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.481237907230441e-06,
+      "loss": 1.0988,
+      "step": 6451
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.480738307579348e-06,
+      "loss": 0.8927,
+      "step": 6452
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48023864048974e-06,
+      "loss": 0.8817,
+      "step": 6453
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.479738905971297e-06,
+      "loss": 0.9278,
+      "step": 6454
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.479239104033703e-06,
+      "loss": 0.8718,
+      "step": 6455
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47873923468664e-06,
+      "loss": 0.9638,
+      "step": 6456
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.478239297939793e-06,
+      "loss": 0.8886,
+      "step": 6457
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47773929380285e-06,
+      "loss": 0.9854,
+      "step": 6458
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.477239222285497e-06,
+      "loss": 0.9149,
+      "step": 6459
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47673908339742e-06,
+      "loss": 0.9881,
+      "step": 6460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.476238877148314e-06,
+      "loss": 1.0147,
+      "step": 6461
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.475738603547868e-06,
+      "loss": 0.9847,
+      "step": 6462
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.475238262605774e-06,
+      "loss": 1.0924,
+      "step": 6463
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47473785433173e-06,
+      "loss": 0.9383,
+      "step": 6464
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.474237378735424e-06,
+      "loss": 1.0026,
+      "step": 6465
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47373683582656e-06,
+      "loss": 0.9221,
+      "step": 6466
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47323622561483e-06,
+      "loss": 0.7906,
+      "step": 6467
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47273554810994e-06,
+      "loss": 1.1264,
+      "step": 6468
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.472234803321585e-06,
+      "loss": 1.0788,
+      "step": 6469
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.471733991259467e-06,
+      "loss": 1.1244,
+      "step": 6470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.471233111933291e-06,
+      "loss": 0.9972,
+      "step": 6471
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.470732165352762e-06,
+      "loss": 1.0192,
+      "step": 6472
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.470231151527583e-06,
+      "loss": 1.0064,
+      "step": 6473
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.469730070467463e-06,
+      "loss": 0.9335,
+      "step": 6474
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.469228922182112e-06,
+      "loss": 0.8987,
+      "step": 6475
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.468727706681234e-06,
+      "loss": 0.9721,
+      "step": 6476
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.468226423974546e-06,
+      "loss": 0.8254,
+      "step": 6477
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.467725074071758e-06,
+      "loss": 1.2029,
+      "step": 6478
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.467223656982581e-06,
+      "loss": 1.0079,
+      "step": 6479
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.466722172716733e-06,
+      "loss": 1.0482,
+      "step": 6480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.466220621283929e-06,
+      "loss": 0.9159,
+      "step": 6481
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.465719002693888e-06,
+      "loss": 1.0998,
+      "step": 6482
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.465217316956329e-06,
+      "loss": 0.9036,
+      "step": 6483
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.464715564080968e-06,
+      "loss": 0.8908,
+      "step": 6484
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.464213744077529e-06,
+      "loss": 1.0433,
+      "step": 6485
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.463711856955735e-06,
+      "loss": 0.8463,
+      "step": 6486
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.46320990272531e-06,
+      "loss": 0.9969,
+      "step": 6487
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.462707881395978e-06,
+      "loss": 1.0532,
+      "step": 6488
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.462205792977465e-06,
+      "loss": 0.8346,
+      "step": 6489
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.461703637479502e-06,
+      "loss": 0.7174,
+      "step": 6490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.461201414911816e-06,
+      "loss": 1.0668,
+      "step": 6491
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.460699125284137e-06,
+      "loss": 1.0183,
+      "step": 6492
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.460196768606199e-06,
+      "loss": 0.8575,
+      "step": 6493
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.459694344887732e-06,
+      "loss": 1.0606,
+      "step": 6494
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.459191854138474e-06,
+      "loss": 0.8473,
+      "step": 6495
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.458689296368154e-06,
+      "loss": 1.079,
+      "step": 6496
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.458186671586516e-06,
+      "loss": 0.9851,
+      "step": 6497
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.457683979803297e-06,
+      "loss": 0.8504,
+      "step": 6498
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.457181221028234e-06,
+      "loss": 0.8838,
+      "step": 6499
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45667839527107e-06,
+      "loss": 0.7181,
+      "step": 6500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.456175502541546e-06,
+      "loss": 1.0227,
+      "step": 6501
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.455672542849405e-06,
+      "loss": 1.0612,
+      "step": 6502
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45516951620439e-06,
+      "loss": 1.0761,
+      "step": 6503
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.454666422616253e-06,
+      "loss": 1.0826,
+      "step": 6504
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.454163262094737e-06,
+      "loss": 0.9645,
+      "step": 6505
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.453660034649591e-06,
+      "loss": 0.8887,
+      "step": 6506
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.453156740290565e-06,
+      "loss": 0.8185,
+      "step": 6507
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45265337902741e-06,
+      "loss": 0.9398,
+      "step": 6508
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45214995086988e-06,
+      "loss": 1.0129,
+      "step": 6509
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.451646455827728e-06,
+      "loss": 0.9164,
+      "step": 6510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45114289391071e-06,
+      "loss": 0.9317,
+      "step": 6511
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.450639265128577e-06,
+      "loss": 1.0788,
+      "step": 6512
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.450135569491094e-06,
+      "loss": 0.8884,
+      "step": 6513
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.449631807008016e-06,
+      "loss": 0.9811,
+      "step": 6514
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.449127977689105e-06,
+      "loss": 0.999,
+      "step": 6515
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.44862408154412e-06,
+      "loss": 1.0335,
+      "step": 6516
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.448120118582828e-06,
+      "loss": 1.0409,
+      "step": 6517
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.447616088814992e-06,
+      "loss": 0.795,
+      "step": 6518
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.447111992250372e-06,
+      "loss": 1.0363,
+      "step": 6519
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.446607828898743e-06,
+      "loss": 1.0645,
+      "step": 6520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.446103598769867e-06,
+      "loss": 0.9929,
+      "step": 6521
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.445599301873518e-06,
+      "loss": 0.9791,
+      "step": 6522
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.445094938219461e-06,
+      "loss": 1.1202,
+      "step": 6523
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.444590507817471e-06,
+      "loss": 0.8108,
+      "step": 6524
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.444086010677324e-06,
+      "loss": 0.8614,
+      "step": 6525
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.44358144680879e-06,
+      "loss": 1.0908,
+      "step": 6526
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.443076816221648e-06,
+      "loss": 1.0455,
+      "step": 6527
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.442572118925672e-06,
+      "loss": 1.0298,
+      "step": 6528
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.442067354930644e-06,
+      "loss": 1.0482,
+      "step": 6529
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.44156252424634e-06,
+      "loss": 1.0715,
+      "step": 6530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.441057626882542e-06,
+      "loss": 0.9792,
+      "step": 6531
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.440552662849033e-06,
+      "loss": 0.884,
+      "step": 6532
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.440047632155599e-06,
+      "loss": 1.0362,
+      "step": 6533
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.439542534812021e-06,
+      "loss": 0.9037,
+      "step": 6534
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.439037370828088e-06,
+      "loss": 0.9248,
+      "step": 6535
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.438532140213585e-06,
+      "loss": 1.0137,
+      "step": 6536
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.438026842978301e-06,
+      "loss": 0.9387,
+      "step": 6537
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43752147913203e-06,
+      "loss": 1.0476,
+      "step": 6538
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.437016048684556e-06,
+      "loss": 1.0556,
+      "step": 6539
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.436510551645677e-06,
+      "loss": 1.0454,
+      "step": 6540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.436004988025187e-06,
+      "loss": 0.8482,
+      "step": 6541
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.435499357832878e-06,
+      "loss": 0.8846,
+      "step": 6542
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43499366107855e-06,
+      "loss": 0.9737,
+      "step": 6543
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.434487897771998e-06,
+      "loss": 0.8482,
+      "step": 6544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.433982067923021e-06,
+      "loss": 0.715,
+      "step": 6545
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.433476171541423e-06,
+      "loss": 1.0317,
+      "step": 6546
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.432970208637002e-06,
+      "loss": 0.8588,
+      "step": 6547
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43246417921956e-06,
+      "loss": 1.0327,
+      "step": 6548
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.431958083298906e-06,
+      "loss": 0.9634,
+      "step": 6549
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.431451920884842e-06,
+      "loss": 1.0969,
+      "step": 6550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.430945691987174e-06,
+      "loss": 0.9255,
+      "step": 6551
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.430439396615712e-06,
+      "loss": 1.0957,
+      "step": 6552
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.429933034780264e-06,
+      "loss": 1.046,
+      "step": 6553
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.429426606490642e-06,
+      "loss": 0.9823,
+      "step": 6554
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.428920111756658e-06,
+      "loss": 0.9046,
+      "step": 6555
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.428413550588125e-06,
+      "loss": 0.916,
+      "step": 6556
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.427906922994858e-06,
+      "loss": 0.8094,
+      "step": 6557
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42740022898667e-06,
+      "loss": 1.0556,
+      "step": 6558
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.426893468573381e-06,
+      "loss": 0.9348,
+      "step": 6559
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.426386641764807e-06,
+      "loss": 0.877,
+      "step": 6560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42587974857077e-06,
+      "loss": 0.8948,
+      "step": 6561
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42537278900109e-06,
+      "loss": 0.9936,
+      "step": 6562
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42486576306559e-06,
+      "loss": 0.9599,
+      "step": 6563
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.424358670774092e-06,
+      "loss": 1.2001,
+      "step": 6564
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.423851512136422e-06,
+      "loss": 0.944,
+      "step": 6565
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.423344287162406e-06,
+      "loss": 0.9752,
+      "step": 6566
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42283699586187e-06,
+      "loss": 0.9841,
+      "step": 6567
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.422329638244643e-06,
+      "loss": 1.009,
+      "step": 6568
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.421822214320557e-06,
+      "loss": 1.0944,
+      "step": 6569
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42131472409944e-06,
+      "loss": 0.9745,
+      "step": 6570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.420807167591127e-06,
+      "loss": 1.1732,
+      "step": 6571
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.420299544805453e-06,
+      "loss": 0.9559,
+      "step": 6572
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41979185575225e-06,
+      "loss": 0.9593,
+      "step": 6573
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.419284100441354e-06,
+      "loss": 0.785,
+      "step": 6574
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.418776278882605e-06,
+      "loss": 1.0976,
+      "step": 6575
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.418268391085841e-06,
+      "loss": 1.0201,
+      "step": 6576
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.417760437060905e-06,
+      "loss": 0.9128,
+      "step": 6577
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.417252416817632e-06,
+      "loss": 0.9798,
+      "step": 6578
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41674433036587e-06,
+      "loss": 0.9701,
+      "step": 6579
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.416236177715462e-06,
+      "loss": 0.9664,
+      "step": 6580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.415727958876254e-06,
+      "loss": 0.9379,
+      "step": 6581
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41521967385809e-06,
+      "loss": 0.9813,
+      "step": 6582
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41471132267082e-06,
+      "loss": 1.0701,
+      "step": 6583
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.414202905324293e-06,
+      "loss": 0.9107,
+      "step": 6584
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41369442182836e-06,
+      "loss": 0.9432,
+      "step": 6585
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41318587219287e-06,
+      "loss": 0.9277,
+      "step": 6586
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41267725642768e-06,
+      "loss": 0.9237,
+      "step": 6587
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.412168574542641e-06,
+      "loss": 1.1025,
+      "step": 6588
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.411659826547609e-06,
+      "loss": 1.028,
+      "step": 6589
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.411151012452442e-06,
+      "loss": 1.0552,
+      "step": 6590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.410642132267e-06,
+      "loss": 0.9134,
+      "step": 6591
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.410133186001138e-06,
+      "loss": 0.9002,
+      "step": 6592
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.409624173664721e-06,
+      "loss": 0.8203,
+      "step": 6593
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.409115095267608e-06,
+      "loss": 0.9541,
+      "step": 6594
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.408605950819664e-06,
+      "loss": 1.0777,
+      "step": 6595
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40809674033075e-06,
+      "loss": 1.089,
+      "step": 6596
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.407587463810739e-06,
+      "loss": 0.7948,
+      "step": 6597
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.407078121269491e-06,
+      "loss": 1.0781,
+      "step": 6598
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.406568712716879e-06,
+      "loss": 0.9625,
+      "step": 6599
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40605923816277e-06,
+      "loss": 0.8863,
+      "step": 6600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.405549697617035e-06,
+      "loss": 0.8339,
+      "step": 6601
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40504009108955e-06,
+      "loss": 1.0504,
+      "step": 6602
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.404530418590184e-06,
+      "loss": 1.1264,
+      "step": 6603
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.404020680128813e-06,
+      "loss": 0.9634,
+      "step": 6604
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.403510875715313e-06,
+      "loss": 1.1116,
+      "step": 6605
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.403001005359562e-06,
+      "loss": 1.0984,
+      "step": 6606
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.402491069071439e-06,
+      "loss": 0.8866,
+      "step": 6607
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.401981066860823e-06,
+      "loss": 1.0726,
+      "step": 6608
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.401470998737597e-06,
+      "loss": 0.8412,
+      "step": 6609
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40096086471164e-06,
+      "loss": 0.9353,
+      "step": 6610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.400450664792838e-06,
+      "loss": 0.9323,
+      "step": 6611
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.399940398991077e-06,
+      "loss": 0.9411,
+      "step": 6612
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.399430067316241e-06,
+      "loss": 0.8293,
+      "step": 6613
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39891966977822e-06,
+      "loss": 1.0353,
+      "step": 6614
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.3984092063869e-06,
+      "loss": 1.003,
+      "step": 6615
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.397898677152173e-06,
+      "loss": 0.933,
+      "step": 6616
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39738808208393e-06,
+      "loss": 0.8922,
+      "step": 6617
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.396877421192062e-06,
+      "loss": 1.0121,
+      "step": 6618
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.396366694486466e-06,
+      "loss": 0.9839,
+      "step": 6619
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.395855901977036e-06,
+      "loss": 0.9066,
+      "step": 6620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.395345043673668e-06,
+      "loss": 0.8095,
+      "step": 6621
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.394834119586259e-06,
+      "loss": 0.8994,
+      "step": 6622
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39432312972471e-06,
+      "loss": 0.7523,
+      "step": 6623
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.393812074098919e-06,
+      "loss": 0.9148,
+      "step": 6624
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.393300952718791e-06,
+      "loss": 0.9531,
+      "step": 6625
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.392789765594224e-06,
+      "loss": 1.0926,
+      "step": 6626
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.392278512735127e-06,
+      "loss": 1.1918,
+      "step": 6627
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.391767194151403e-06,
+      "loss": 1.0229,
+      "step": 6628
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39125580985296e-06,
+      "loss": 1.0408,
+      "step": 6629
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.390744359849702e-06,
+      "loss": 1.113,
+      "step": 6630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.390232844151545e-06,
+      "loss": 0.898,
+      "step": 6631
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.389721262768393e-06,
+      "loss": 0.9738,
+      "step": 6632
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.38920961571016e-06,
+      "loss": 0.9949,
+      "step": 6633
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.388697902986763e-06,
+      "loss": 0.7139,
+      "step": 6634
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.388186124608108e-06,
+      "loss": 0.886,
+      "step": 6635
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.387674280584119e-06,
+      "loss": 1.0014,
+      "step": 6636
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.387162370924709e-06,
+      "loss": 0.8243,
+      "step": 6637
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.386650395639795e-06,
+      "loss": 0.8143,
+      "step": 6638
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.386138354739299e-06,
+      "loss": 0.8221,
+      "step": 6639
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.385626248233138e-06,
+      "loss": 0.8885,
+      "step": 6640
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.385114076131239e-06,
+      "loss": 0.9678,
+      "step": 6641
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.384601838443522e-06,
+      "loss": 0.9241,
+      "step": 6642
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.384089535179912e-06,
+      "loss": 0.9112,
+      "step": 6643
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.383577166350335e-06,
+      "loss": 0.9774,
+      "step": 6644
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.383064731964718e-06,
+      "loss": 1.0992,
+      "step": 6645
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.38255223203299e-06,
+      "loss": 1.0416,
+      "step": 6646
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.382039666565077e-06,
+      "loss": 0.9023,
+      "step": 6647
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.381527035570916e-06,
+      "loss": 1.0783,
+      "step": 6648
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.381014339060434e-06,
+      "loss": 1.0509,
+      "step": 6649
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.380501577043565e-06,
+      "loss": 0.8803,
+      "step": 6650
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.379988749530246e-06,
+      "loss": 0.8956,
+      "step": 6651
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.379475856530413e-06,
+      "loss": 0.9315,
+      "step": 6652
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.378962898053999e-06,
+      "loss": 1.1014,
+      "step": 6653
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.378449874110947e-06,
+      "loss": 1.1156,
+      "step": 6654
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.377936784711195e-06,
+      "loss": 0.9895,
+      "step": 6655
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.377423629864686e-06,
+      "loss": 0.9576,
+      "step": 6656
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.376910409581359e-06,
+      "loss": 0.9747,
+      "step": 6657
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.376397123871157e-06,
+      "loss": 1.0278,
+      "step": 6658
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.375883772744028e-06,
+      "loss": 0.9673,
+      "step": 6659
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.375370356209917e-06,
+      "loss": 1.0033,
+      "step": 6660
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.37485687427877e-06,
+      "loss": 0.9189,
+      "step": 6661
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.374343326960536e-06,
+      "loss": 1.0999,
+      "step": 6662
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.373829714265168e-06,
+      "loss": 0.8733,
+      "step": 6663
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.373316036202614e-06,
+      "loss": 0.7553,
+      "step": 6664
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.372802292782824e-06,
+      "loss": 0.9212,
+      "step": 6665
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.372288484015758e-06,
+      "loss": 0.9762,
+      "step": 6666
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.371774609911367e-06,
+      "loss": 0.9258,
+      "step": 6667
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.371260670479607e-06,
+      "loss": 0.9387,
+      "step": 6668
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.370746665730436e-06,
+      "loss": 0.8906,
+      "step": 6669
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.370232595673814e-06,
+      "loss": 0.8788,
+      "step": 6670
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3697184603197e-06,
+      "loss": 0.7098,
+      "step": 6671
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.369204259678054e-06,
+      "loss": 0.9272,
+      "step": 6672
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.36868999375884e-06,
+      "loss": 0.9755,
+      "step": 6673
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.368175662572021e-06,
+      "loss": 0.9326,
+      "step": 6674
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.367661266127565e-06,
+      "loss": 1.0565,
+      "step": 6675
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.367146804435432e-06,
+      "loss": 0.9737,
+      "step": 6676
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.366632277505598e-06,
+      "loss": 0.9865,
+      "step": 6677
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.366117685348025e-06,
+      "loss": 1.0514,
+      "step": 6678
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.365603027972684e-06,
+      "loss": 0.9985,
+      "step": 6679
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.36508830538955e-06,
+      "loss": 0.9289,
+      "step": 6680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.364573517608592e-06,
+      "loss": 1.0252,
+      "step": 6681
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.364058664639788e-06,
+      "loss": 0.8122,
+      "step": 6682
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.363543746493108e-06,
+      "loss": 0.8264,
+      "step": 6683
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.36302876317853e-06,
+      "loss": 0.8912,
+      "step": 6684
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.362513714706034e-06,
+      "loss": 0.9787,
+      "step": 6685
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.361998601085598e-06,
+      "loss": 0.9489,
+      "step": 6686
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.361483422327202e-06,
+      "loss": 1.2201,
+      "step": 6687
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.360968178440825e-06,
+      "loss": 1.0072,
+      "step": 6688
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.360452869436453e-06,
+      "loss": 1.0013,
+      "step": 6689
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35993749532407e-06,
+      "loss": 0.9493,
+      "step": 6690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35942205611366e-06,
+      "loss": 1.2233,
+      "step": 6691
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35890655181521e-06,
+      "loss": 0.9989,
+      "step": 6692
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.358390982438706e-06,
+      "loss": 1.0744,
+      "step": 6693
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.357875347994139e-06,
+      "loss": 0.8326,
+      "step": 6694
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3573596484915e-06,
+      "loss": 1.004,
+      "step": 6695
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.356843883940777e-06,
+      "loss": 0.6724,
+      "step": 6696
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.356328054351966e-06,
+      "loss": 1.076,
+      "step": 6697
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.355812159735062e-06,
+      "loss": 0.949,
+      "step": 6698
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.355296200100057e-06,
+      "loss": 0.9334,
+      "step": 6699
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35478017545695e-06,
+      "loss": 0.8425,
+      "step": 6700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.354264085815736e-06,
+      "loss": 0.7989,
+      "step": 6701
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.353747931186419e-06,
+      "loss": 0.9072,
+      "step": 6702
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.353231711578996e-06,
+      "loss": 0.8456,
+      "step": 6703
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.352715427003468e-06,
+      "loss": 0.7678,
+      "step": 6704
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35219907746984e-06,
+      "loss": 0.9554,
+      "step": 6705
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.351682662988115e-06,
+      "loss": 0.957,
+      "step": 6706
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3511661835683e-06,
+      "loss": 0.9145,
+      "step": 6707
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.350649639220398e-06,
+      "loss": 1.0435,
+      "step": 6708
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35013302995442e-06,
+      "loss": 0.901,
+      "step": 6709
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.349616355780376e-06,
+      "loss": 0.8616,
+      "step": 6710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.349099616708273e-06,
+      "loss": 0.9057,
+      "step": 6711
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.348582812748127e-06,
+      "loss": 0.9413,
+      "step": 6712
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.348065943909947e-06,
+      "loss": 1.007,
+      "step": 6713
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.347549010203749e-06,
+      "loss": 0.9184,
+      "step": 6714
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.347032011639548e-06,
+      "loss": 1.0732,
+      "step": 6715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.346514948227362e-06,
+      "loss": 0.9923,
+      "step": 6716
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.345997819977207e-06,
+      "loss": 0.9113,
+      "step": 6717
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.345480626899103e-06,
+      "loss": 0.8509,
+      "step": 6718
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.344963369003072e-06,
+      "loss": 0.8801,
+      "step": 6719
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.344446046299134e-06,
+      "loss": 1.0316,
+      "step": 6720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.343928658797312e-06,
+      "loss": 1.0158,
+      "step": 6721
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.343411206507633e-06,
+      "loss": 1.0004,
+      "step": 6722
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.342893689440116e-06,
+      "loss": 0.9777,
+      "step": 6723
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.342376107604795e-06,
+      "loss": 0.9586,
+      "step": 6724
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.341858461011694e-06,
+      "loss": 0.963,
+      "step": 6725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.341340749670844e-06,
+      "loss": 0.8376,
+      "step": 6726
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.340822973592275e-06,
+      "loss": 1.0107,
+      "step": 6727
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.340305132786017e-06,
+      "loss": 0.9988,
+      "step": 6728
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.339787227262106e-06,
+      "loss": 1.0802,
+      "step": 6729
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.339269257030576e-06,
+      "loss": 0.9674,
+      "step": 6730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.338751222101463e-06,
+      "loss": 0.8995,
+      "step": 6731
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3382331224848e-06,
+      "loss": 1.0425,
+      "step": 6732
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.337714958190627e-06,
+      "loss": 0.9696,
+      "step": 6733
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.337196729228985e-06,
+      "loss": 0.8311,
+      "step": 6734
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.336678435609913e-06,
+      "loss": 0.7302,
+      "step": 6735
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.336160077343453e-06,
+      "loss": 1.0043,
+      "step": 6736
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.335641654439648e-06,
+      "loss": 0.9538,
+      "step": 6737
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.335123166908544e-06,
+      "loss": 0.8151,
+      "step": 6738
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.334604614760185e-06,
+      "loss": 0.9912,
+      "step": 6739
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.334085998004614e-06,
+      "loss": 1.018,
+      "step": 6740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.333567316651889e-06,
+      "loss": 0.9331,
+      "step": 6741
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.333048570712051e-06,
+      "loss": 0.7364,
+      "step": 6742
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.332529760195151e-06,
+      "loss": 1.0449,
+      "step": 6743
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.332010885111246e-06,
+      "loss": 0.9454,
+      "step": 6744
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.331491945470385e-06,
+      "loss": 1.043,
+      "step": 6745
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.330972941282622e-06,
+      "loss": 1.1143,
+      "step": 6746
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.330453872558015e-06,
+      "loss": 1.0019,
+      "step": 6747
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.329934739306619e-06,
+      "loss": 0.9589,
+      "step": 6748
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.329415541538493e-06,
+      "loss": 0.9506,
+      "step": 6749
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.328896279263696e-06,
+      "loss": 1.0397,
+      "step": 6750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32837695249229e-06,
+      "loss": 1.0045,
+      "step": 6751
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.327857561234333e-06,
+      "loss": 0.9217,
+      "step": 6752
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.327338105499893e-06,
+      "loss": 1.0585,
+      "step": 6753
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.326818585299032e-06,
+      "loss": 1.0132,
+      "step": 6754
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.326299000641813e-06,
+      "loss": 1.0029,
+      "step": 6755
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.325779351538307e-06,
+      "loss": 0.9841,
+      "step": 6756
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.325259637998581e-06,
+      "loss": 1.092,
+      "step": 6757
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.324739860032704e-06,
+      "loss": 0.882,
+      "step": 6758
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.324220017650746e-06,
+      "loss": 0.8562,
+      "step": 6759
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.323700110862779e-06,
+      "loss": 0.934,
+      "step": 6760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.323180139678876e-06,
+      "loss": 0.9823,
+      "step": 6761
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.322660104109112e-06,
+      "loss": 0.9673,
+      "step": 6762
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32214000416356e-06,
+      "loss": 0.8188,
+      "step": 6763
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.321619839852302e-06,
+      "loss": 1.0286,
+      "step": 6764
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.321099611185413e-06,
+      "loss": 1.0579,
+      "step": 6765
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.320579318172972e-06,
+      "loss": 0.8931,
+      "step": 6766
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32005896082506e-06,
+      "loss": 1.0543,
+      "step": 6767
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.319538539151758e-06,
+      "loss": 0.893,
+      "step": 6768
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.319018053163147e-06,
+      "loss": 0.8892,
+      "step": 6769
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.318497502869319e-06,
+      "loss": 0.9002,
+      "step": 6770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.317976888280352e-06,
+      "loss": 1.1362,
+      "step": 6771
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.317456209406337e-06,
+      "loss": 1.0452,
+      "step": 6772
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.31693546625736e-06,
+      "loss": 0.9997,
+      "step": 6773
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.31641465884351e-06,
+      "loss": 0.976,
+      "step": 6774
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.315893787174877e-06,
+      "loss": 0.886,
+      "step": 6775
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.315372851261558e-06,
+      "loss": 1.0002,
+      "step": 6776
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.314851851113639e-06,
+      "loss": 0.9084,
+      "step": 6777
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.314330786741218e-06,
+      "loss": 0.9855,
+      "step": 6778
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.31380965815439e-06,
+      "loss": 0.9783,
+      "step": 6779
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.313288465363251e-06,
+      "loss": 0.8179,
+      "step": 6780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3127672083779e-06,
+      "loss": 0.8631,
+      "step": 6781
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.312245887208436e-06,
+      "loss": 1.0002,
+      "step": 6782
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.311724501864959e-06,
+      "loss": 1.0017,
+      "step": 6783
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.311203052357572e-06,
+      "loss": 1.0295,
+      "step": 6784
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.310681538696378e-06,
+      "loss": 1.0218,
+      "step": 6785
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.310159960891477e-06,
+      "loss": 0.8819,
+      "step": 6786
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.309638318952977e-06,
+      "loss": 0.8876,
+      "step": 6787
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.30911661289099e-06,
+      "loss": 0.9402,
+      "step": 6788
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.308594842715614e-06,
+      "loss": 1.0873,
+      "step": 6789
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.308073008436967e-06,
+      "loss": 1.0041,
+      "step": 6790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.307551110065154e-06,
+      "loss": 0.9942,
+      "step": 6791
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.30702914761029e-06,
+      "loss": 1.0302,
+      "step": 6792
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.306507121082486e-06,
+      "loss": 0.743,
+      "step": 6793
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.305985030491855e-06,
+      "loss": 1.0037,
+      "step": 6794
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.305462875848515e-06,
+      "loss": 0.7626,
+      "step": 6795
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.304940657162584e-06,
+      "loss": 1.0593,
+      "step": 6796
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.304418374444174e-06,
+      "loss": 0.9373,
+      "step": 6797
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.303896027703408e-06,
+      "loss": 0.9202,
+      "step": 6798
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.303373616950408e-06,
+      "loss": 0.9848,
+      "step": 6799
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.302851142195293e-06,
+      "loss": 0.9681,
+      "step": 6800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.302328603448184e-06,
+      "loss": 1.0198,
+      "step": 6801
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.301806000719211e-06,
+      "loss": 0.9994,
+      "step": 6802
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.301283334018493e-06,
+      "loss": 0.8849,
+      "step": 6803
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.30076060335616e-06,
+      "loss": 0.9488,
+      "step": 6804
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.300237808742339e-06,
+      "loss": 0.796,
+      "step": 6805
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.299714950187158e-06,
+      "loss": 0.9285,
+      "step": 6806
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29919202770075e-06,
+      "loss": 1.0682,
+      "step": 6807
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.298669041293245e-06,
+      "loss": 1.0704,
+      "step": 6808
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.298145990974776e-06,
+      "loss": 0.8915,
+      "step": 6809
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.297622876755474e-06,
+      "loss": 1.0131,
+      "step": 6810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.297099698645478e-06,
+      "loss": 0.9344,
+      "step": 6811
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.296576456654924e-06,
+      "loss": 1.0762,
+      "step": 6812
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.296053150793948e-06,
+      "loss": 1.0047,
+      "step": 6813
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29552978107269e-06,
+      "loss": 1.1452,
+      "step": 6814
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.295006347501292e-06,
+      "loss": 0.9713,
+      "step": 6815
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29448285008989e-06,
+      "loss": 0.9583,
+      "step": 6816
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.293959288848632e-06,
+      "loss": 0.9715,
+      "step": 6817
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29343566378766e-06,
+      "loss": 0.9141,
+      "step": 6818
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29291197491712e-06,
+      "loss": 0.9102,
+      "step": 6819
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.292388222247155e-06,
+      "loss": 0.8511,
+      "step": 6820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.291864405787916e-06,
+      "loss": 1.0114,
+      "step": 6821
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.291340525549552e-06,
+      "loss": 1.0608,
+      "step": 6822
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.290816581542212e-06,
+      "loss": 1.059,
+      "step": 6823
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.290292573776045e-06,
+      "loss": 0.9764,
+      "step": 6824
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.289768502261207e-06,
+      "loss": 1.0702,
+      "step": 6825
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.289244367007852e-06,
+      "loss": 0.9457,
+      "step": 6826
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.288720168026131e-06,
+      "loss": 0.9835,
+      "step": 6827
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.288195905326205e-06,
+      "loss": 1.0771,
+      "step": 6828
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.287671578918228e-06,
+      "loss": 0.9652,
+      "step": 6829
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28714718881236e-06,
+      "loss": 0.9088,
+      "step": 6830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28662273501876e-06,
+      "loss": 0.8407,
+      "step": 6831
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.286098217547592e-06,
+      "loss": 1.0006,
+      "step": 6832
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.285573636409015e-06,
+      "loss": 0.9223,
+      "step": 6833
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.285048991613193e-06,
+      "loss": 1.074,
+      "step": 6834
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.284524283170294e-06,
+      "loss": 0.9831,
+      "step": 6835
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.283999511090483e-06,
+      "loss": 0.9477,
+      "step": 6836
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.283474675383925e-06,
+      "loss": 0.9842,
+      "step": 6837
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.282949776060789e-06,
+      "loss": 1.0776,
+      "step": 6838
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.282424813131248e-06,
+      "loss": 1.0174,
+      "step": 6839
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.281899786605468e-06,
+      "loss": 1.0852,
+      "step": 6840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.281374696493628e-06,
+      "loss": 0.956,
+      "step": 6841
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.280849542805893e-06,
+      "loss": 0.9748,
+      "step": 6842
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.280324325552444e-06,
+      "loss": 0.9435,
+      "step": 6843
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.279799044743457e-06,
+      "loss": 0.9701,
+      "step": 6844
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.279273700389106e-06,
+      "loss": 0.9222,
+      "step": 6845
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.278748292499573e-06,
+      "loss": 1.0201,
+      "step": 6846
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.278222821085033e-06,
+      "loss": 0.7586,
+      "step": 6847
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.27769728615567e-06,
+      "loss": 1.1068,
+      "step": 6848
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.277171687721666e-06,
+      "loss": 1.0906,
+      "step": 6849
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.276646025793202e-06,
+      "loss": 0.9302,
+      "step": 6850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.276120300380468e-06,
+      "loss": 0.8537,
+      "step": 6851
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.275594511493645e-06,
+      "loss": 0.7375,
+      "step": 6852
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.27506865914292e-06,
+      "loss": 0.932,
+      "step": 6853
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.274542743338484e-06,
+      "loss": 0.8846,
+      "step": 6854
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.274016764090525e-06,
+      "loss": 1.0839,
+      "step": 6855
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.273490721409234e-06,
+      "loss": 1.1784,
+      "step": 6856
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.272964615304801e-06,
+      "loss": 0.9402,
+      "step": 6857
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.272438445787424e-06,
+      "loss": 0.9517,
+      "step": 6858
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.271912212867292e-06,
+      "loss": 0.7146,
+      "step": 6859
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.271385916554605e-06,
+      "loss": 1.0669,
+      "step": 6860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.270859556859558e-06,
+      "loss": 0.924,
+      "step": 6861
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.270333133792347e-06,
+      "loss": 0.9659,
+      "step": 6862
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.269806647363175e-06,
+      "loss": 1.0978,
+      "step": 6863
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.26928009758224e-06,
+      "loss": 0.9262,
+      "step": 6864
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.268753484459745e-06,
+      "loss": 1.1585,
+      "step": 6865
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.268226808005892e-06,
+      "loss": 0.9586,
+      "step": 6866
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.267700068230888e-06,
+      "loss": 0.8806,
+      "step": 6867
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.267173265144933e-06,
+      "loss": 0.9455,
+      "step": 6868
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.26664639875824e-06,
+      "loss": 0.9586,
+      "step": 6869
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.266119469081012e-06,
+      "loss": 0.9674,
+      "step": 6870
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.265592476123463e-06,
+      "loss": 1.161,
+      "step": 6871
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2650654198958e-06,
+      "loss": 1.1311,
+      "step": 6872
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.264538300408233e-06,
+      "loss": 0.9608,
+      "step": 6873
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.264011117670978e-06,
+      "loss": 1.0592,
+      "step": 6874
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.263483871694247e-06,
+      "loss": 0.8248,
+      "step": 6875
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.262956562488258e-06,
+      "loss": 0.9044,
+      "step": 6876
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.262429190063226e-06,
+      "loss": 0.9627,
+      "step": 6877
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.261901754429367e-06,
+      "loss": 0.98,
+      "step": 6878
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2613742555969e-06,
+      "loss": 1.0351,
+      "step": 6879
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.26084669357605e-06,
+      "loss": 1.1871,
+      "step": 6880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.260319068377034e-06,
+      "loss": 0.9791,
+      "step": 6881
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.259791380010074e-06,
+      "loss": 1.0248,
+      "step": 6882
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.259263628485397e-06,
+      "loss": 1.0396,
+      "step": 6883
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.258735813813227e-06,
+      "loss": 1.0217,
+      "step": 6884
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.258207936003788e-06,
+      "loss": 1.0908,
+      "step": 6885
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25767999506731e-06,
+      "loss": 1.0732,
+      "step": 6886
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.257151991014022e-06,
+      "loss": 0.9782,
+      "step": 6887
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.256623923854153e-06,
+      "loss": 0.9841,
+      "step": 6888
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.256095793597933e-06,
+      "loss": 1.0168,
+      "step": 6889
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.255567600255596e-06,
+      "loss": 1.1564,
+      "step": 6890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.255039343837375e-06,
+      "loss": 0.9523,
+      "step": 6891
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.254511024353505e-06,
+      "loss": 0.931,
+      "step": 6892
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.253982641814223e-06,
+      "loss": 0.8784,
+      "step": 6893
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.253454196229765e-06,
+      "loss": 0.8217,
+      "step": 6894
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.252925687610371e-06,
+      "loss": 0.9057,
+      "step": 6895
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.252397115966279e-06,
+      "loss": 0.9733,
+      "step": 6896
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25186848130773e-06,
+      "loss": 1.0028,
+      "step": 6897
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.251339783644967e-06,
+      "loss": 0.9159,
+      "step": 6898
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.250811022988233e-06,
+      "loss": 0.9371,
+      "step": 6899
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.250282199347773e-06,
+      "loss": 0.9441,
+      "step": 6900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.249753312733832e-06,
+      "loss": 0.8642,
+      "step": 6901
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.24922436315666e-06,
+      "loss": 1.1337,
+      "step": 6902
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.248695350626501e-06,
+      "loss": 1.026,
+      "step": 6903
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.248166275153608e-06,
+      "loss": 0.8965,
+      "step": 6904
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.24763713674823e-06,
+      "loss": 0.9911,
+      "step": 6905
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.247107935420617e-06,
+      "loss": 1.0565,
+      "step": 6906
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.246578671181026e-06,
+      "loss": 0.9588,
+      "step": 6907
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.24604934403971e-06,
+      "loss": 0.9588,
+      "step": 6908
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.245519954006925e-06,
+      "loss": 1.163,
+      "step": 6909
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.244990501092924e-06,
+      "loss": 1.0348,
+      "step": 6910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.244460985307972e-06,
+      "loss": 1.0562,
+      "step": 6911
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.243931406662323e-06,
+      "loss": 0.8937,
+      "step": 6912
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.243401765166236e-06,
+      "loss": 0.898,
+      "step": 6913
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.242872060829978e-06,
+      "loss": 1.1356,
+      "step": 6914
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.24234229366381e-06,
+      "loss": 1.037,
+      "step": 6915
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.241812463677994e-06,
+      "loss": 0.8854,
+      "step": 6916
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.241282570882797e-06,
+      "loss": 1.0447,
+      "step": 6917
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.240752615288488e-06,
+      "loss": 1.0325,
+      "step": 6918
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.240222596905329e-06,
+      "loss": 1.0174,
+      "step": 6919
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.239692515743594e-06,
+      "loss": 1.0611,
+      "step": 6920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.239162371813552e-06,
+      "loss": 0.9106,
+      "step": 6921
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.23863216512547e-06,
+      "loss": 0.8948,
+      "step": 6922
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.23810189568963e-06,
+      "loss": 0.8425,
+      "step": 6923
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.237571563516296e-06,
+      "loss": 1.0678,
+      "step": 6924
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.237041168615748e-06,
+      "loss": 1.1723,
+      "step": 6925
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.236510710998262e-06,
+      "loss": 0.9532,
+      "step": 6926
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.235980190674117e-06,
+      "loss": 0.8451,
+      "step": 6927
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.235449607653588e-06,
+      "loss": 0.9648,
+      "step": 6928
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.234918961946958e-06,
+      "loss": 0.8536,
+      "step": 6929
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.234388253564506e-06,
+      "loss": 1.0175,
+      "step": 6930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.233857482516515e-06,
+      "loss": 1.0888,
+      "step": 6931
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.233326648813269e-06,
+      "loss": 0.9876,
+      "step": 6932
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.232795752465052e-06,
+      "loss": 1.08,
+      "step": 6933
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.232264793482153e-06,
+      "loss": 0.8859,
+      "step": 6934
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.231733771874855e-06,
+      "loss": 1.0747,
+      "step": 6935
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.23120268765345e-06,
+      "loss": 1.1665,
+      "step": 6936
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.230671540828224e-06,
+      "loss": 1.1667,
+      "step": 6937
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.23014033140947e-06,
+      "loss": 1.0946,
+      "step": 6938
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22960905940748e-06,
+      "loss": 0.8622,
+      "step": 6939
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.229077724832547e-06,
+      "loss": 1.3221,
+      "step": 6940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.228546327694965e-06,
+      "loss": 1.0159,
+      "step": 6941
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22801486800503e-06,
+      "loss": 1.0875,
+      "step": 6942
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22748334577304e-06,
+      "loss": 0.8388,
+      "step": 6943
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.226951761009291e-06,
+      "loss": 0.8782,
+      "step": 6944
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.226420113724088e-06,
+      "loss": 0.6969,
+      "step": 6945
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.225888403927723e-06,
+      "loss": 0.8353,
+      "step": 6946
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.225356631630504e-06,
+      "loss": 0.7947,
+      "step": 6947
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22482479684273e-06,
+      "loss": 1.0152,
+      "step": 6948
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22429289957471e-06,
+      "loss": 0.9361,
+      "step": 6949
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.223760939836743e-06,
+      "loss": 1.0404,
+      "step": 6950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22322891763914e-06,
+      "loss": 1.0323,
+      "step": 6951
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.222696832992208e-06,
+      "loss": 1.0777,
+      "step": 6952
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.222164685906258e-06,
+      "loss": 1.0649,
+      "step": 6953
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.221632476391596e-06,
+      "loss": 0.9817,
+      "step": 6954
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.221100204458534e-06,
+      "loss": 1.0851,
+      "step": 6955
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.220567870117389e-06,
+      "loss": 1.0338,
+      "step": 6956
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22003547337847e-06,
+      "loss": 0.9955,
+      "step": 6957
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.219503014252096e-06,
+      "loss": 0.8464,
+      "step": 6958
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.218970492748581e-06,
+      "loss": 1.081,
+      "step": 6959
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.218437908878243e-06,
+      "loss": 1.0732,
+      "step": 6960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2179052626514e-06,
+      "loss": 0.9848,
+      "step": 6961
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.217372554078373e-06,
+      "loss": 0.9504,
+      "step": 6962
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.216839783169483e-06,
+      "loss": 0.9406,
+      "step": 6963
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.21630694993505e-06,
+      "loss": 0.8292,
+      "step": 6964
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.215774054385401e-06,
+      "loss": 0.8759,
+      "step": 6965
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.215241096530859e-06,
+      "loss": 0.8312,
+      "step": 6966
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.21470807638175e-06,
+      "loss": 1.023,
+      "step": 6967
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2141749939484e-06,
+      "loss": 0.9115,
+      "step": 6968
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.213641849241141e-06,
+      "loss": 0.8562,
+      "step": 6969
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2131086422703e-06,
+      "loss": 1.009,
+      "step": 6970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.212575373046205e-06,
+      "loss": 1.0186,
+      "step": 6971
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.212042041579192e-06,
+      "loss": 1.0163,
+      "step": 6972
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.211508647879594e-06,
+      "loss": 0.8516,
+      "step": 6973
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.210975191957744e-06,
+      "loss": 0.9135,
+      "step": 6974
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.210441673823978e-06,
+      "loss": 0.9704,
+      "step": 6975
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.209908093488635e-06,
+      "loss": 1.1019,
+      "step": 6976
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.209374450962047e-06,
+      "loss": 1.0179,
+      "step": 6977
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.208840746254559e-06,
+      "loss": 1.025,
+      "step": 6978
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.208306979376511e-06,
+      "loss": 1.0015,
+      "step": 6979
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.20777315033824e-06,
+      "loss": 1.0493,
+      "step": 6980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.207239259150093e-06,
+      "loss": 1.1224,
+      "step": 6981
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.206705305822414e-06,
+      "loss": 0.8624,
+      "step": 6982
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.206171290365546e-06,
+      "loss": 0.9714,
+      "step": 6983
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.205637212789836e-06,
+      "loss": 0.7966,
+      "step": 6984
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.205103073105634e-06,
+      "loss": 1.0132,
+      "step": 6985
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.204568871323284e-06,
+      "loss": 0.9965,
+      "step": 6986
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.204034607453142e-06,
+      "loss": 1.0411,
+      "step": 6987
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.203500281505554e-06,
+      "loss": 1.0369,
+      "step": 6988
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.202965893490877e-06,
+      "loss": 1.0852,
+      "step": 6989
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.20243144341946e-06,
+      "loss": 1.0873,
+      "step": 6990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.201896931301662e-06,
+      "loss": 1.0017,
+      "step": 6991
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.201362357147837e-06,
+      "loss": 0.9085,
+      "step": 6992
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.200827720968343e-06,
+      "loss": 0.9311,
+      "step": 6993
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.200293022773537e-06,
+      "loss": 0.8623,
+      "step": 6994
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.199758262573779e-06,
+      "loss": 0.8721,
+      "step": 6995
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.199223440379431e-06,
+      "loss": 1.0063,
+      "step": 6996
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.198688556200855e-06,
+      "loss": 0.8874,
+      "step": 6997
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.198153610048414e-06,
+      "loss": 1.1102,
+      "step": 6998
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.197618601932472e-06,
+      "loss": 0.9099,
+      "step": 6999
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.197083531863394e-06,
+      "loss": 1.0101,
+      "step": 7000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.196548399851549e-06,
+      "loss": 0.841,
+      "step": 7001
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.196013205907304e-06,
+      "loss": 0.754,
+      "step": 7002
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.195477950041027e-06,
+      "loss": 0.9283,
+      "step": 7003
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.194942632263092e-06,
+      "loss": 0.9474,
+      "step": 7004
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.194407252583864e-06,
+      "loss": 0.8987,
+      "step": 7005
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.193871811013724e-06,
+      "loss": 1.0019,
+      "step": 7006
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19333630756304e-06,
+      "loss": 0.9576,
+      "step": 7007
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19280074224219e-06,
+      "loss": 0.9567,
+      "step": 7008
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19226511506155e-06,
+      "loss": 0.8528,
+      "step": 7009
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.191729426031498e-06,
+      "loss": 0.9599,
+      "step": 7010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.191193675162411e-06,
+      "loss": 0.8931,
+      "step": 7011
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.190657862464673e-06,
+      "loss": 0.9004,
+      "step": 7012
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19012198794866e-06,
+      "loss": 1.0877,
+      "step": 7013
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.189586051624758e-06,
+      "loss": 0.9928,
+      "step": 7014
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18905005350335e-06,
+      "loss": 0.9573,
+      "step": 7015
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18851399359482e-06,
+      "loss": 0.6726,
+      "step": 7016
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.187977871909557e-06,
+      "loss": 0.7598,
+      "step": 7017
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.187441688457943e-06,
+      "loss": 1.001,
+      "step": 7018
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.186905443250371e-06,
+      "loss": 0.8085,
+      "step": 7019
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.186369136297227e-06,
+      "loss": 1.2547,
+      "step": 7020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.185832767608905e-06,
+      "loss": 0.9865,
+      "step": 7021
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.185296337195797e-06,
+      "loss": 1.0003,
+      "step": 7022
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.184759845068293e-06,
+      "loss": 1.0545,
+      "step": 7023
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18422329123679e-06,
+      "loss": 0.9847,
+      "step": 7024
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.183686675711683e-06,
+      "loss": 0.9051,
+      "step": 7025
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18314999850337e-06,
+      "loss": 1.0923,
+      "step": 7026
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.182613259622247e-06,
+      "loss": 1.0783,
+      "step": 7027
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.182076459078713e-06,
+      "loss": 1.0252,
+      "step": 7028
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.181539596883172e-06,
+      "loss": 1.0144,
+      "step": 7029
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.181002673046021e-06,
+      "loss": 0.7348,
+      "step": 7030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.180465687577665e-06,
+      "loss": 1.0541,
+      "step": 7031
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.179928640488508e-06,
+      "loss": 0.9829,
+      "step": 7032
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.179391531788955e-06,
+      "loss": 0.8801,
+      "step": 7033
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.178854361489413e-06,
+      "loss": 0.8538,
+      "step": 7034
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.178317129600287e-06,
+      "loss": 0.9182,
+      "step": 7035
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.177779836131988e-06,
+      "loss": 0.874,
+      "step": 7036
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.177242481094927e-06,
+      "loss": 0.7653,
+      "step": 7037
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.176705064499512e-06,
+      "loss": 0.9884,
+      "step": 7038
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.176167586356158e-06,
+      "loss": 0.9635,
+      "step": 7039
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.175630046675278e-06,
+      "loss": 0.8214,
+      "step": 7040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.175092445467286e-06,
+      "loss": 0.984,
+      "step": 7041
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.174554782742596e-06,
+      "loss": 0.9123,
+      "step": 7042
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.17401705851163e-06,
+      "loss": 1.085,
+      "step": 7043
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.1734792727848e-06,
+      "loss": 0.8444,
+      "step": 7044
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.172941425572532e-06,
+      "loss": 0.9607,
+      "step": 7045
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.172403516885244e-06,
+      "loss": 0.9646,
+      "step": 7046
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.171865546733355e-06,
+      "loss": 0.9463,
+      "step": 7047
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.171327515127291e-06,
+      "loss": 0.9326,
+      "step": 7048
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.170789422077478e-06,
+      "loss": 0.8177,
+      "step": 7049
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.170251267594338e-06,
+      "loss": 0.9789,
+      "step": 7050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.169713051688298e-06,
+      "loss": 1.1378,
+      "step": 7051
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.169174774369787e-06,
+      "loss": 1.1343,
+      "step": 7052
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.168636435649235e-06,
+      "loss": 0.8875,
+      "step": 7053
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16809803553707e-06,
+      "loss": 0.8413,
+      "step": 7054
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.167559574043723e-06,
+      "loss": 1.1187,
+      "step": 7055
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.167021051179627e-06,
+      "loss": 0.9345,
+      "step": 7056
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16648246695522e-06,
+      "loss": 0.9032,
+      "step": 7057
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16594382138093e-06,
+      "loss": 0.8972,
+      "step": 7058
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.165405114467199e-06,
+      "loss": 0.9907,
+      "step": 7059
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.164866346224462e-06,
+      "loss": 1.1002,
+      "step": 7060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.164327516663157e-06,
+      "loss": 0.8715,
+      "step": 7061
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.163788625793723e-06,
+      "loss": 0.9508,
+      "step": 7062
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.163249673626603e-06,
+      "loss": 0.9723,
+      "step": 7063
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.162710660172239e-06,
+      "loss": 0.7859,
+      "step": 7064
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16217158544107e-06,
+      "loss": 1.0253,
+      "step": 7065
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.161632449443546e-06,
+      "loss": 0.9538,
+      "step": 7066
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.161093252190112e-06,
+      "loss": 0.9142,
+      "step": 7067
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16055399369121e-06,
+      "loss": 0.9666,
+      "step": 7068
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.160014673957295e-06,
+      "loss": 0.9537,
+      "step": 7069
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.15947529299881e-06,
+      "loss": 0.9661,
+      "step": 7070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.15893585082621e-06,
+      "loss": 0.8757,
+      "step": 7071
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.158396347449942e-06,
+      "loss": 1.0124,
+      "step": 7072
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.157856782880464e-06,
+      "loss": 0.9661,
+      "step": 7073
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.157317157128226e-06,
+      "loss": 1.1664,
+      "step": 7074
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.156777470203683e-06,
+      "loss": 0.9037,
+      "step": 7075
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.156237722117293e-06,
+      "loss": 0.9847,
+      "step": 7076
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.155697912879514e-06,
+      "loss": 1.0793,
+      "step": 7077
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.155158042500803e-06,
+      "loss": 0.9711,
+      "step": 7078
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.154618110991622e-06,
+      "loss": 1.1366,
+      "step": 7079
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.154078118362428e-06,
+      "loss": 0.906,
+      "step": 7080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.153538064623687e-06,
+      "loss": 0.9505,
+      "step": 7081
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.152997949785862e-06,
+      "loss": 1.096,
+      "step": 7082
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.152457773859417e-06,
+      "loss": 1.0245,
+      "step": 7083
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.151917536854817e-06,
+      "loss": 1.0448,
+      "step": 7084
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.15137723878253e-06,
+      "loss": 1.0024,
+      "step": 7085
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.150836879653025e-06,
+      "loss": 0.969,
+      "step": 7086
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.150296459476769e-06,
+      "loss": 0.8296,
+      "step": 7087
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.149755978264232e-06,
+      "loss": 0.8855,
+      "step": 7088
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.14921543602589e-06,
+      "loss": 0.9919,
+      "step": 7089
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.148674832772213e-06,
+      "loss": 1.0643,
+      "step": 7090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.148134168513675e-06,
+      "loss": 0.8729,
+      "step": 7091
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.147593443260753e-06,
+      "loss": 0.9677,
+      "step": 7092
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.14705265702392e-06,
+      "loss": 0.8875,
+      "step": 7093
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.146511809813656e-06,
+      "loss": 0.8865,
+      "step": 7094
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.145970901640439e-06,
+      "loss": 0.9246,
+      "step": 7095
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.145429932514751e-06,
+      "loss": 1.0628,
+      "step": 7096
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.14488890244707e-06,
+      "loss": 0.9442,
+      "step": 7097
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.144347811447881e-06,
+      "loss": 0.7951,
+      "step": 7098
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.143806659527667e-06,
+      "loss": 1.0191,
+      "step": 7099
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.143265446696909e-06,
+      "loss": 0.8466,
+      "step": 7100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.1427241729661e-06,
+      "loss": 0.9477,
+      "step": 7101
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.14218283834572e-06,
+      "loss": 0.931,
+      "step": 7102
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.141641442846262e-06,
+      "loss": 0.9471,
+      "step": 7103
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.141099986478212e-06,
+      "loss": 0.8427,
+      "step": 7104
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.140558469252065e-06,
+      "loss": 0.9357,
+      "step": 7105
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.140016891178306e-06,
+      "loss": 1.0395,
+      "step": 7106
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.139475252267434e-06,
+      "loss": 0.8081,
+      "step": 7107
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.138933552529943e-06,
+      "loss": 0.9156,
+      "step": 7108
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.138391791976326e-06,
+      "loss": 0.9045,
+      "step": 7109
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.137849970617078e-06,
+      "loss": 1.0619,
+      "step": 7110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.137308088462699e-06,
+      "loss": 0.9663,
+      "step": 7111
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.136766145523688e-06,
+      "loss": 0.91,
+      "step": 7112
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.136224141810545e-06,
+      "loss": 1.0572,
+      "step": 7113
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.135682077333769e-06,
+      "loss": 0.8306,
+      "step": 7114
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.135139952103865e-06,
+      "loss": 1.0406,
+      "step": 7115
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.134597766131336e-06,
+      "loss": 0.9675,
+      "step": 7116
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.134055519426688e-06,
+      "loss": 0.9795,
+      "step": 7117
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.133513212000422e-06,
+      "loss": 1.0419,
+      "step": 7118
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.13297084386305e-06,
+      "loss": 0.9167,
+      "step": 7119
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.132428415025077e-06,
+      "loss": 1.0001,
+      "step": 7120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.131885925497016e-06,
+      "loss": 1.0112,
+      "step": 7121
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.131343375289376e-06,
+      "loss": 0.9721,
+      "step": 7122
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.130800764412666e-06,
+      "loss": 0.9603,
+      "step": 7123
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.130258092877402e-06,
+      "loss": 1.0035,
+      "step": 7124
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.129715360694099e-06,
+      "loss": 0.9976,
+      "step": 7125
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.12917256787327e-06,
+      "loss": 0.8002,
+      "step": 7126
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.128629714425431e-06,
+      "loss": 0.9225,
+      "step": 7127
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.128086800361104e-06,
+      "loss": 0.9082,
+      "step": 7128
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.1275438256908e-06,
+      "loss": 0.9378,
+      "step": 7129
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.127000790425048e-06,
+      "loss": 1.022,
+      "step": 7130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.126457694574364e-06,
+      "loss": 0.9282,
+      "step": 7131
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.125914538149272e-06,
+      "loss": 1.0224,
+      "step": 7132
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.125371321160293e-06,
+      "loss": 0.9808,
+      "step": 7133
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.124828043617955e-06,
+      "loss": 1.1807,
+      "step": 7134
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.124284705532784e-06,
+      "loss": 0.7685,
+      "step": 7135
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.123741306915302e-06,
+      "loss": 0.8347,
+      "step": 7136
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.123197847776043e-06,
+      "loss": 0.9663,
+      "step": 7137
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.122654328125534e-06,
+      "loss": 1.0822,
+      "step": 7138
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.122110747974306e-06,
+      "loss": 0.8954,
+      "step": 7139
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.12156710733289e-06,
+      "loss": 1.0158,
+      "step": 7140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.12102340621182e-06,
+      "loss": 0.8092,
+      "step": 7141
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.120479644621628e-06,
+      "loss": 0.9739,
+      "step": 7142
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.11993582257285e-06,
+      "loss": 1.0044,
+      "step": 7143
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.119391940076024e-06,
+      "loss": 0.9697,
+      "step": 7144
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.11884799714169e-06,
+      "loss": 0.9191,
+      "step": 7145
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.11830399378038e-06,
+      "loss": 0.8243,
+      "step": 7146
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.117759930002638e-06,
+      "loss": 0.9977,
+      "step": 7147
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.117215805819006e-06,
+      "loss": 1.0901,
+      "step": 7148
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.116671621240024e-06,
+      "loss": 1.0454,
+      "step": 7149
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.116127376276236e-06,
+      "loss": 0.9765,
+      "step": 7150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.115583070938187e-06,
+      "loss": 1.0754,
+      "step": 7151
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.115038705236422e-06,
+      "loss": 0.8917,
+      "step": 7152
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.114494279181489e-06,
+      "loss": 1.0856,
+      "step": 7153
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.113949792783937e-06,
+      "loss": 0.898,
+      "step": 7154
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.113405246054314e-06,
+      "loss": 0.9406,
+      "step": 7155
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.112860639003169e-06,
+      "loss": 0.9145,
+      "step": 7156
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.112315971641056e-06,
+      "loss": 0.9335,
+      "step": 7157
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.111771243978528e-06,
+      "loss": 0.9529,
+      "step": 7158
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.111226456026138e-06,
+      "loss": 0.8512,
+      "step": 7159
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.110681607794439e-06,
+      "loss": 0.8781,
+      "step": 7160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.110136699293991e-06,
+      "loss": 1.0872,
+      "step": 7161
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.10959173053535e-06,
+      "loss": 0.9715,
+      "step": 7162
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.109046701529076e-06,
+      "loss": 1.1074,
+      "step": 7163
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.108501612285724e-06,
+      "loss": 0.9833,
+      "step": 7164
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.107956462815862e-06,
+      "loss": 1.017,
+      "step": 7165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.107411253130046e-06,
+      "loss": 1.044,
+      "step": 7166
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.106865983238843e-06,
+      "loss": 0.9646,
+      "step": 7167
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.106320653152817e-06,
+      "loss": 1.0223,
+      "step": 7168
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.105775262882531e-06,
+      "loss": 1.0352,
+      "step": 7169
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.105229812438554e-06,
+      "loss": 1.1322,
+      "step": 7170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.104684301831456e-06,
+      "loss": 0.9618,
+      "step": 7171
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.104138731071802e-06,
+      "loss": 0.9759,
+      "step": 7172
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.103593100170164e-06,
+      "loss": 0.8764,
+      "step": 7173
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.103047409137114e-06,
+      "loss": 0.8377,
+      "step": 7174
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.102501657983227e-06,
+      "loss": 1.0355,
+      "step": 7175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.101955846719069e-06,
+      "loss": 0.9191,
+      "step": 7176
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.101409975355223e-06,
+      "loss": 1.1262,
+      "step": 7177
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.100864043902264e-06,
+      "loss": 0.8287,
+      "step": 7178
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.100318052370767e-06,
+      "loss": 0.8863,
+      "step": 7179
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.099772000771308e-06,
+      "loss": 0.9331,
+      "step": 7180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.099225889114471e-06,
+      "loss": 0.9967,
+      "step": 7181
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.098679717410835e-06,
+      "loss": 0.7886,
+      "step": 7182
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.098133485670985e-06,
+      "loss": 0.9189,
+      "step": 7183
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.0975871939055e-06,
+      "loss": 1.0199,
+      "step": 7184
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.097040842124964e-06,
+      "loss": 0.8389,
+      "step": 7185
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.096494430339967e-06,
+      "loss": 0.9575,
+      "step": 7186
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.095947958561091e-06,
+      "loss": 0.9865,
+      "step": 7187
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.095401426798926e-06,
+      "loss": 0.6921,
+      "step": 7188
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.09485483506406e-06,
+      "loss": 0.8917,
+      "step": 7189
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.094308183367085e-06,
+      "loss": 0.9257,
+      "step": 7190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.09376147171859e-06,
+      "loss": 1.0828,
+      "step": 7191
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.093214700129168e-06,
+      "loss": 0.9597,
+      "step": 7192
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.092667868609413e-06,
+      "loss": 0.8434,
+      "step": 7193
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.09212097716992e-06,
+      "loss": 0.9432,
+      "step": 7194
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.091574025821286e-06,
+      "loss": 0.8875,
+      "step": 7195
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.091027014574103e-06,
+      "loss": 1.0809,
+      "step": 7196
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.090479943438975e-06,
+      "loss": 0.9951,
+      "step": 7197
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.089932812426497e-06,
+      "loss": 0.806,
+      "step": 7198
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.089385621547274e-06,
+      "loss": 0.9419,
+      "step": 7199
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.088838370811905e-06,
+      "loss": 0.934,
+      "step": 7200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.088291060230992e-06,
+      "loss": 0.9693,
+      "step": 7201
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.087743689815141e-06,
+      "loss": 0.9445,
+      "step": 7202
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.087196259574958e-06,
+      "loss": 0.9798,
+      "step": 7203
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.086648769521046e-06,
+      "loss": 0.9261,
+      "step": 7204
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.086101219664014e-06,
+      "loss": 0.8428,
+      "step": 7205
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.085553610014471e-06,
+      "loss": 1.1228,
+      "step": 7206
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.085005940583028e-06,
+      "loss": 0.9276,
+      "step": 7207
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.084458211380294e-06,
+      "loss": 0.8883,
+      "step": 7208
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.083910422416882e-06,
+      "loss": 1.0251,
+      "step": 7209
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.083362573703405e-06,
+      "loss": 0.8598,
+      "step": 7210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.082814665250476e-06,
+      "loss": 1.0094,
+      "step": 7211
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.082266697068714e-06,
+      "loss": 0.9012,
+      "step": 7212
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.081718669168734e-06,
+      "loss": 0.8896,
+      "step": 7213
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.081170581561152e-06,
+      "loss": 0.8763,
+      "step": 7214
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.080622434256593e-06,
+      "loss": 0.9153,
+      "step": 7215
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.08007422726567e-06,
+      "loss": 0.8804,
+      "step": 7216
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.079525960599006e-06,
+      "loss": 0.7649,
+      "step": 7217
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.078977634267228e-06,
+      "loss": 1.1036,
+      "step": 7218
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.078429248280956e-06,
+      "loss": 0.9882,
+      "step": 7219
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.077880802650814e-06,
+      "loss": 0.9626,
+      "step": 7220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.077332297387432e-06,
+      "loss": 1.1058,
+      "step": 7221
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.076783732501434e-06,
+      "loss": 1.0209,
+      "step": 7222
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.076235108003448e-06,
+      "loss": 1.0779,
+      "step": 7223
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.075686423904106e-06,
+      "loss": 0.8842,
+      "step": 7224
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.075137680214036e-06,
+      "loss": 0.8859,
+      "step": 7225
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.074588876943872e-06,
+      "loss": 0.8748,
+      "step": 7226
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.074040014104246e-06,
+      "loss": 1.1172,
+      "step": 7227
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.073491091705793e-06,
+      "loss": 0.9475,
+      "step": 7228
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.072942109759146e-06,
+      "loss": 0.919,
+      "step": 7229
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.072393068274942e-06,
+      "loss": 1.0981,
+      "step": 7230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.07184396726382e-06,
+      "loss": 0.8539,
+      "step": 7231
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.071294806736419e-06,
+      "loss": 1.0186,
+      "step": 7232
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.070745586703375e-06,
+      "loss": 0.8983,
+      "step": 7233
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.070196307175333e-06,
+      "loss": 0.8021,
+      "step": 7234
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.069646968162933e-06,
+      "loss": 0.9854,
+      "step": 7235
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.069097569676822e-06,
+      "loss": 0.8723,
+      "step": 7236
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.06854811172764e-06,
+      "loss": 0.9748,
+      "step": 7237
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.067998594326037e-06,
+      "loss": 0.939,
+      "step": 7238
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.067449017482654e-06,
+      "loss": 0.9876,
+      "step": 7239
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.066899381208145e-06,
+      "loss": 0.9183,
+      "step": 7240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.066349685513153e-06,
+      "loss": 0.9317,
+      "step": 7241
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.065799930408334e-06,
+      "loss": 1.0056,
+      "step": 7242
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.065250115904335e-06,
+      "loss": 1.0481,
+      "step": 7243
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.064700242011812e-06,
+      "loss": 1.0222,
+      "step": 7244
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.064150308741413e-06,
+      "loss": 0.7394,
+      "step": 7245
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.0636003161038e-06,
+      "loss": 0.9817,
+      "step": 7246
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.063050264109627e-06,
+      "loss": 0.8263,
+      "step": 7247
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.062500152769547e-06,
+      "loss": 0.9767,
+      "step": 7248
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.061949982094223e-06,
+      "loss": 0.9933,
+      "step": 7249
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.061399752094312e-06,
+      "loss": 0.8858,
+      "step": 7250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.060849462780476e-06,
+      "loss": 0.8109,
+      "step": 7251
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.060299114163375e-06,
+      "loss": 0.943,
+      "step": 7252
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.059748706253674e-06,
+      "loss": 0.8052,
+      "step": 7253
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.059198239062036e-06,
+      "loss": 0.9744,
+      "step": 7254
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.058647712599126e-06,
+      "loss": 0.9359,
+      "step": 7255
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05809712687561e-06,
+      "loss": 1.0744,
+      "step": 7256
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.057546481902154e-06,
+      "loss": 0.9059,
+      "step": 7257
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.056995777689432e-06,
+      "loss": 0.9891,
+      "step": 7258
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.056445014248108e-06,
+      "loss": 0.815,
+      "step": 7259
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.055894191588855e-06,
+      "loss": 1.0062,
+      "step": 7260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.055343309722346e-06,
+      "loss": 0.8453,
+      "step": 7261
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.054792368659253e-06,
+      "loss": 0.9965,
+      "step": 7262
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.054241368410252e-06,
+      "loss": 0.9349,
+      "step": 7263
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.053690308986015e-06,
+      "loss": 0.7803,
+      "step": 7264
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.053139190397224e-06,
+      "loss": 1.1575,
+      "step": 7265
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05258801265455e-06,
+      "loss": 1.0269,
+      "step": 7266
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05203677576868e-06,
+      "loss": 0.9632,
+      "step": 7267
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.051485479750286e-06,
+      "loss": 0.9093,
+      "step": 7268
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.050934124610055e-06,
+      "loss": 1.0806,
+      "step": 7269
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.050382710358666e-06,
+      "loss": 0.8866,
+      "step": 7270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.049831237006805e-06,
+      "loss": 0.9288,
+      "step": 7271
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.049279704565157e-06,
+      "loss": 0.9076,
+      "step": 7272
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.048728113044404e-06,
+      "loss": 0.9549,
+      "step": 7273
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.048176462455237e-06,
+      "loss": 0.8917,
+      "step": 7274
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.047624752808341e-06,
+      "loss": 0.9472,
+      "step": 7275
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.047072984114408e-06,
+      "loss": 1.0366,
+      "step": 7276
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.046521156384126e-06,
+      "loss": 0.8726,
+      "step": 7277
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04596926962819e-06,
+      "loss": 0.9394,
+      "step": 7278
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.045417323857287e-06,
+      "loss": 1.0211,
+      "step": 7279
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.044865319082116e-06,
+      "loss": 1.0307,
+      "step": 7280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04431325531337e-06,
+      "loss": 0.6715,
+      "step": 7281
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.043761132561745e-06,
+      "loss": 0.9446,
+      "step": 7282
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04320895083794e-06,
+      "loss": 1.0205,
+      "step": 7283
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.042656710152648e-06,
+      "loss": 0.8672,
+      "step": 7284
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.042104410516576e-06,
+      "loss": 0.9576,
+      "step": 7285
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04155205194042e-06,
+      "loss": 0.9202,
+      "step": 7286
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.040999634434883e-06,
+      "loss": 1.0373,
+      "step": 7287
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.040447158010668e-06,
+      "loss": 1.0179,
+      "step": 7288
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.03989462267848e-06,
+      "loss": 1.0155,
+      "step": 7289
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.039342028449022e-06,
+      "loss": 1.2118,
+      "step": 7290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.038789375333001e-06,
+      "loss": 0.7951,
+      "step": 7291
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.038236663341126e-06,
+      "loss": 0.9787,
+      "step": 7292
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.037683892484104e-06,
+      "loss": 0.9173,
+      "step": 7293
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.037131062772647e-06,
+      "loss": 1.1292,
+      "step": 7294
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.036578174217464e-06,
+      "loss": 1.0867,
+      "step": 7295
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.036025226829268e-06,
+      "loss": 0.9551,
+      "step": 7296
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.035472220618772e-06,
+      "loss": 1.0688,
+      "step": 7297
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.034919155596688e-06,
+      "loss": 1.0475,
+      "step": 7298
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.034366031773736e-06,
+      "loss": 1.014,
+      "step": 7299
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.033812849160632e-06,
+      "loss": 0.9125,
+      "step": 7300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.033259607768089e-06,
+      "loss": 0.9776,
+      "step": 7301
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.032706307606832e-06,
+      "loss": 0.8112,
+      "step": 7302
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.032152948687576e-06,
+      "loss": 0.969,
+      "step": 7303
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.031599531021045e-06,
+      "loss": 0.9981,
+      "step": 7304
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.03104605461796e-06,
+      "loss": 0.7787,
+      "step": 7305
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.030492519489049e-06,
+      "loss": 1.0154,
+      "step": 7306
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02993892564503e-06,
+      "loss": 1.0856,
+      "step": 7307
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.029385273096631e-06,
+      "loss": 1.0011,
+      "step": 7308
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02883156185458e-06,
+      "loss": 0.8501,
+      "step": 7309
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.028277791929605e-06,
+      "loss": 1.1438,
+      "step": 7310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.027723963332433e-06,
+      "loss": 0.9667,
+      "step": 7311
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.027170076073799e-06,
+      "loss": 1.0014,
+      "step": 7312
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.026616130164428e-06,
+      "loss": 1.0821,
+      "step": 7313
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02606212561506e-06,
+      "loss": 0.9722,
+      "step": 7314
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.025508062436423e-06,
+      "loss": 0.855,
+      "step": 7315
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.024953940639253e-06,
+      "loss": 0.9286,
+      "step": 7316
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.024399760234286e-06,
+      "loss": 1.0147,
+      "step": 7317
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02384552123226e-06,
+      "loss": 1.0368,
+      "step": 7318
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.023291223643914e-06,
+      "loss": 0.9949,
+      "step": 7319
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.022736867479986e-06,
+      "loss": 0.8213,
+      "step": 7320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.022182452751218e-06,
+      "loss": 1.0005,
+      "step": 7321
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.021627979468348e-06,
+      "loss": 0.8746,
+      "step": 7322
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.021073447642123e-06,
+      "loss": 1.0279,
+      "step": 7323
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.020518857283284e-06,
+      "loss": 1.023,
+      "step": 7324
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.019964208402578e-06,
+      "loss": 0.8684,
+      "step": 7325
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.019409501010752e-06,
+      "loss": 0.8466,
+      "step": 7326
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.01885473511855e-06,
+      "loss": 0.909,
+      "step": 7327
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.018299910736723e-06,
+      "loss": 0.9663,
+      "step": 7328
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.017745027876018e-06,
+      "loss": 0.9619,
+      "step": 7329
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.01719008654719e-06,
+      "loss": 0.9633,
+      "step": 7330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.016635086760988e-06,
+      "loss": 1.0554,
+      "step": 7331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.016080028528166e-06,
+      "loss": 0.7514,
+      "step": 7332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.015524911859476e-06,
+      "loss": 0.9791,
+      "step": 7333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.014969736765677e-06,
+      "loss": 1.1883,
+      "step": 7334
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.014414503257521e-06,
+      "loss": 0.9215,
+      "step": 7335
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.01385921134577e-06,
+      "loss": 0.8877,
+      "step": 7336
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.013303861041178e-06,
+      "loss": 0.9839,
+      "step": 7337
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.012748452354508e-06,
+      "loss": 0.9747,
+      "step": 7338
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.012192985296521e-06,
+      "loss": 0.9929,
+      "step": 7339
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.011637459877976e-06,
+      "loss": 0.9836,
+      "step": 7340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.01108187610964e-06,
+      "loss": 0.8334,
+      "step": 7341
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.010526234002278e-06,
+      "loss": 0.9415,
+      "step": 7342
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.00997053356665e-06,
+      "loss": 1.1565,
+      "step": 7343
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.009414774813527e-06,
+      "loss": 0.8074,
+      "step": 7344
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.008858957753674e-06,
+      "loss": 1.0694,
+      "step": 7345
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.008303082397863e-06,
+      "loss": 0.9287,
+      "step": 7346
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.007747148756861e-06,
+      "loss": 0.853,
+      "step": 7347
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.00719115684144e-06,
+      "loss": 0.9795,
+      "step": 7348
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.006635106662374e-06,
+      "loss": 0.8786,
+      "step": 7349
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.006078998230435e-06,
+      "loss": 1.0312,
+      "step": 7350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.005522831556397e-06,
+      "loss": 0.9854,
+      "step": 7351
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.004966606651035e-06,
+      "loss": 1.1008,
+      "step": 7352
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.004410323525125e-06,
+      "loss": 1.0014,
+      "step": 7353
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.003853982189449e-06,
+      "loss": 1.0718,
+      "step": 7354
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.003297582654783e-06,
+      "loss": 0.8705,
+      "step": 7355
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.002741124931906e-06,
+      "loss": 0.9568,
+      "step": 7356
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.002184609031603e-06,
+      "loss": 1.0456,
+      "step": 7357
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.001628034964654e-06,
+      "loss": 0.9246,
+      "step": 7358
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.001071402741843e-06,
+      "loss": 1.203,
+      "step": 7359
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.000514712373952e-06,
+      "loss": 0.8611,
+      "step": 7360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.999957963871771e-06,
+      "loss": 1.031,
+      "step": 7361
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.999401157246086e-06,
+      "loss": 1.0507,
+      "step": 7362
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.99884429250768e-06,
+      "loss": 1.0382,
+      "step": 7363
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.99828736966735e-06,
+      "loss": 0.8572,
+      "step": 7364
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.99773038873588e-06,
+      "loss": 1.0917,
+      "step": 7365
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.997173349724066e-06,
+      "loss": 0.9352,
+      "step": 7366
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.996616252642695e-06,
+      "loss": 0.9972,
+      "step": 7367
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.996059097502566e-06,
+      "loss": 1.1365,
+      "step": 7368
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.995501884314472e-06,
+      "loss": 0.9234,
+      "step": 7369
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.994944613089207e-06,
+      "loss": 0.8892,
+      "step": 7370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.99438728383757e-06,
+      "loss": 0.9183,
+      "step": 7371
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.99382989657036e-06,
+      "loss": 0.9842,
+      "step": 7372
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.993272451298374e-06,
+      "loss": 0.8663,
+      "step": 7373
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.992714948032411e-06,
+      "loss": 0.9829,
+      "step": 7374
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.992157386783277e-06,
+      "loss": 0.9931,
+      "step": 7375
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.991599767561772e-06,
+      "loss": 0.9827,
+      "step": 7376
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.991042090378702e-06,
+      "loss": 0.8833,
+      "step": 7377
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.990484355244868e-06,
+      "loss": 0.9203,
+      "step": 7378
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.989926562171077e-06,
+      "loss": 0.9848,
+      "step": 7379
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.98936871116814e-06,
+      "loss": 0.9328,
+      "step": 7380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.988810802246859e-06,
+      "loss": 0.734,
+      "step": 7381
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.988252835418048e-06,
+      "loss": 0.9506,
+      "step": 7382
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.987694810692517e-06,
+      "loss": 1.0822,
+      "step": 7383
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.987136728081077e-06,
+      "loss": 0.9875,
+      "step": 7384
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.986578587594539e-06,
+      "loss": 0.9858,
+      "step": 7385
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.986020389243719e-06,
+      "loss": 0.8618,
+      "step": 7386
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.985462133039431e-06,
+      "loss": 1.0127,
+      "step": 7387
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.984903818992491e-06,
+      "loss": 0.9305,
+      "step": 7388
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.984345447113718e-06,
+      "loss": 1.2169,
+      "step": 7389
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.98378701741393e-06,
+      "loss": 1.0162,
+      "step": 7390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.98322852990394e-06,
+      "loss": 0.9172,
+      "step": 7391
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.982669984594578e-06,
+      "loss": 0.8797,
+      "step": 7392
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.982111381496663e-06,
+      "loss": 0.9291,
+      "step": 7393
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.981552720621014e-06,
+      "loss": 0.87,
+      "step": 7394
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.980994001978458e-06,
+      "loss": 0.8502,
+      "step": 7395
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.980435225579819e-06,
+      "loss": 0.9078,
+      "step": 7396
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.979876391435923e-06,
+      "loss": 0.9871,
+      "step": 7397
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.9793174995576e-06,
+      "loss": 0.7585,
+      "step": 7398
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.978758549955676e-06,
+      "loss": 0.8699,
+      "step": 7399
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.978199542640978e-06,
+      "loss": 1.0105,
+      "step": 7400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.977640477624343e-06,
+      "loss": 1.1887,
+      "step": 7401
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.977081354916597e-06,
+      "loss": 1.0177,
+      "step": 7402
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.976522174528576e-06,
+      "loss": 0.9879,
+      "step": 7403
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.975962936471111e-06,
+      "loss": 0.7602,
+      "step": 7404
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.975403640755042e-06,
+      "loss": 1.0213,
+      "step": 7405
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.974844287391202e-06,
+      "loss": 1.0038,
+      "step": 7406
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.974284876390427e-06,
+      "loss": 1.0001,
+      "step": 7407
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.973725407763558e-06,
+      "loss": 0.9708,
+      "step": 7408
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.973165881521435e-06,
+      "loss": 0.95,
+      "step": 7409
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.972606297674895e-06,
+      "loss": 0.8612,
+      "step": 7410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.972046656234785e-06,
+      "loss": 0.9756,
+      "step": 7411
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.971486957211942e-06,
+      "loss": 0.93,
+      "step": 7412
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.970927200617215e-06,
+      "loss": 1.001,
+      "step": 7413
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.97036738646145e-06,
+      "loss": 0.8883,
+      "step": 7414
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.969807514755487e-06,
+      "loss": 1.0151,
+      "step": 7415
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.969247585510178e-06,
+      "loss": 0.8911,
+      "step": 7416
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.96868759873637e-06,
+      "loss": 0.9631,
+      "step": 7417
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.968127554444916e-06,
+      "loss": 1.0247,
+      "step": 7418
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.967567452646661e-06,
+      "loss": 1.0302,
+      "step": 7419
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.96700729335246e-06,
+      "loss": 0.9282,
+      "step": 7420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.966447076573167e-06,
+      "loss": 0.9583,
+      "step": 7421
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.965886802319633e-06,
+      "loss": 0.9881,
+      "step": 7422
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.965326470602716e-06,
+      "loss": 0.7823,
+      "step": 7423
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.964766081433272e-06,
+      "loss": 0.9394,
+      "step": 7424
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.964205634822157e-06,
+      "loss": 0.9288,
+      "step": 7425
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.963645130780231e-06,
+      "loss": 1.104,
+      "step": 7426
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.963084569318352e-06,
+      "loss": 0.883,
+      "step": 7427
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.962523950447383e-06,
+      "loss": 0.8962,
+      "step": 7428
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.961963274178183e-06,
+      "loss": 0.9094,
+      "step": 7429
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.961402540521617e-06,
+      "loss": 0.9654,
+      "step": 7430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.960841749488549e-06,
+      "loss": 1.0932,
+      "step": 7431
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.96028090108984e-06,
+      "loss": 0.7853,
+      "step": 7432
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.959719995336364e-06,
+      "loss": 0.9106,
+      "step": 7433
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.959159032238983e-06,
+      "loss": 0.967,
+      "step": 7434
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.95859801180857e-06,
+      "loss": 0.862,
+      "step": 7435
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.958036934055988e-06,
+      "loss": 0.9877,
+      "step": 7436
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.957475798992111e-06,
+      "loss": 1.0561,
+      "step": 7437
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.956914606627814e-06,
+      "loss": 1.1641,
+      "step": 7438
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.956353356973965e-06,
+      "loss": 0.8374,
+      "step": 7439
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.955792050041442e-06,
+      "loss": 1.247,
+      "step": 7440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.955230685841119e-06,
+      "loss": 0.9462,
+      "step": 7441
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.95466926438387e-06,
+      "loss": 0.9575,
+      "step": 7442
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.954107785680575e-06,
+      "loss": 0.8795,
+      "step": 7443
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.953546249742112e-06,
+      "loss": 0.9234,
+      "step": 7444
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.95298465657936e-06,
+      "loss": 1.0746,
+      "step": 7445
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.9524230062032e-06,
+      "loss": 1.0417,
+      "step": 7446
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.951861298624514e-06,
+      "loss": 0.865,
+      "step": 7447
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.951299533854187e-06,
+      "loss": 0.9196,
+      "step": 7448
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.950737711903099e-06,
+      "loss": 1.067,
+      "step": 7449
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.950175832782137e-06,
+      "loss": 0.8824,
+      "step": 7450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.949613896502187e-06,
+      "loss": 0.9326,
+      "step": 7451
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.949051903074139e-06,
+      "loss": 0.9987,
+      "step": 7452
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.948489852508876e-06,
+      "loss": 0.9827,
+      "step": 7453
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.947927744817295e-06,
+      "loss": 0.9148,
+      "step": 7454
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.94736558001028e-06,
+      "loss": 0.9325,
+      "step": 7455
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.946803358098725e-06,
+      "loss": 0.9774,
+      "step": 7456
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.946241079093523e-06,
+      "loss": 0.9313,
+      "step": 7457
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.945678743005572e-06,
+      "loss": 0.9129,
+      "step": 7458
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.945116349845761e-06,
+      "loss": 1.0474,
+      "step": 7459
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.944553899624989e-06,
+      "loss": 0.9511,
+      "step": 7460
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.943991392354153e-06,
+      "loss": 0.8602,
+      "step": 7461
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.943428828044152e-06,
+      "loss": 0.9498,
+      "step": 7462
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.942866206705885e-06,
+      "loss": 0.7528,
+      "step": 7463
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.942303528350252e-06,
+      "loss": 1.1074,
+      "step": 7464
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.941740792988157e-06,
+      "loss": 0.9891,
+      "step": 7465
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.9411780006305e-06,
+      "loss": 1.0419,
+      "step": 7466
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.940615151288188e-06,
+      "loss": 0.9814,
+      "step": 7467
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.940052244972122e-06,
+      "loss": 0.9463,
+      "step": 7468
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.939489281693213e-06,
+      "loss": 0.9996,
+      "step": 7469
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.938926261462366e-06,
+      "loss": 0.982,
+      "step": 7470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93836318429049e-06,
+      "loss": 0.9116,
+      "step": 7471
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.937800050188492e-06,
+      "loss": 1.0741,
+      "step": 7472
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.937236859167285e-06,
+      "loss": 1.0131,
+      "step": 7473
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93667361123778e-06,
+      "loss": 0.8127,
+      "step": 7474
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93611030641089e-06,
+      "loss": 0.8624,
+      "step": 7475
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.935546944697528e-06,
+      "loss": 0.827,
+      "step": 7476
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93498352610861e-06,
+      "loss": 1.0586,
+      "step": 7477
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.934420050655053e-06,
+      "loss": 0.7798,
+      "step": 7478
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.933856518347774e-06,
+      "loss": 0.7981,
+      "step": 7479
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.933292929197688e-06,
+      "loss": 0.9377,
+      "step": 7480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.932729283215719e-06,
+      "loss": 1.0402,
+      "step": 7481
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.932165580412784e-06,
+      "loss": 1.0132,
+      "step": 7482
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.931601820799806e-06,
+      "loss": 1.0654,
+      "step": 7483
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.931038004387708e-06,
+      "loss": 1.1149,
+      "step": 7484
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.930474131187413e-06,
+      "loss": 0.8853,
+      "step": 7485
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.929910201209848e-06,
+      "loss": 1.0667,
+      "step": 7486
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.929346214465936e-06,
+      "loss": 0.8011,
+      "step": 7487
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.928782170966607e-06,
+      "loss": 1.0445,
+      "step": 7488
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.928218070722788e-06,
+      "loss": 0.8168,
+      "step": 7489
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.927653913745406e-06,
+      "loss": 0.728,
+      "step": 7490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.927089700045396e-06,
+      "loss": 0.7729,
+      "step": 7491
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.926525429633684e-06,
+      "loss": 1.0973,
+      "step": 7492
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.925961102521209e-06,
+      "loss": 1.0729,
+      "step": 7493
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.9253967187189e-06,
+      "loss": 1.1839,
+      "step": 7494
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.924832278237692e-06,
+      "loss": 0.9788,
+      "step": 7495
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.924267781088523e-06,
+      "loss": 0.9808,
+      "step": 7496
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.92370322728233e-06,
+      "loss": 0.9327,
+      "step": 7497
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.923138616830049e-06,
+      "loss": 1.0922,
+      "step": 7498
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.92257394974262e-06,
+      "loss": 1.0526,
+      "step": 7499
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.922009226030985e-06,
+      "loss": 0.9734,
+      "step": 7500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.921444445706083e-06,
+      "loss": 0.9238,
+      "step": 7501
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.920879608778857e-06,
+      "loss": 0.9553,
+      "step": 7502
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.920314715260253e-06,
+      "loss": 0.8734,
+      "step": 7503
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.919749765161213e-06,
+      "loss": 1.0221,
+      "step": 7504
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.919184758492685e-06,
+      "loss": 1.0726,
+      "step": 7505
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.918619695265612e-06,
+      "loss": 1.0391,
+      "step": 7506
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.918054575490943e-06,
+      "loss": 0.9753,
+      "step": 7507
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.917489399179631e-06,
+      "loss": 0.9754,
+      "step": 7508
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.916924166342625e-06,
+      "loss": 1.0169,
+      "step": 7509
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.916358876990872e-06,
+      "loss": 0.9839,
+      "step": 7510
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91579353113533e-06,
+      "loss": 0.9703,
+      "step": 7511
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91522812878695e-06,
+      "loss": 1.014,
+      "step": 7512
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.914662669956683e-06,
+      "loss": 1.0231,
+      "step": 7513
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91409715465549e-06,
+      "loss": 1.0899,
+      "step": 7514
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.913531582894326e-06,
+      "loss": 1.1067,
+      "step": 7515
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.912965954684148e-06,
+      "loss": 1.0599,
+      "step": 7516
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.912400270035915e-06,
+      "loss": 0.9687,
+      "step": 7517
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.911834528960589e-06,
+      "loss": 0.948,
+      "step": 7518
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.911268731469128e-06,
+      "loss": 0.8664,
+      "step": 7519
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.910702877572496e-06,
+      "loss": 1.0114,
+      "step": 7520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.910136967281657e-06,
+      "loss": 1.0654,
+      "step": 7521
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.909571000607572e-06,
+      "loss": 0.9338,
+      "step": 7522
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90900497756121e-06,
+      "loss": 0.8845,
+      "step": 7523
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.908438898153538e-06,
+      "loss": 0.9301,
+      "step": 7524
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.907872762395519e-06,
+      "loss": 0.9743,
+      "step": 7525
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.907306570298128e-06,
+      "loss": 0.8532,
+      "step": 7526
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90674032187233e-06,
+      "loss": 0.9136,
+      "step": 7527
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.906174017129098e-06,
+      "loss": 0.9554,
+      "step": 7528
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.905607656079403e-06,
+      "loss": 0.8692,
+      "step": 7529
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.905041238734219e-06,
+      "loss": 1.0137,
+      "step": 7530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90447476510452e-06,
+      "loss": 0.9447,
+      "step": 7531
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.903908235201282e-06,
+      "loss": 0.9074,
+      "step": 7532
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90334164903548e-06,
+      "loss": 0.9132,
+      "step": 7533
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.902775006618094e-06,
+      "loss": 1.0585,
+      "step": 7534
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.9022083079601e-06,
+      "loss": 0.8469,
+      "step": 7535
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.901641553072475e-06,
+      "loss": 0.9293,
+      "step": 7536
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.901074741966206e-06,
+      "loss": 0.9196,
+      "step": 7537
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.900507874652272e-06,
+      "loss": 0.872,
+      "step": 7538
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.899940951141656e-06,
+      "loss": 1.0687,
+      "step": 7539
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.899373971445342e-06,
+      "loss": 1.0744,
+      "step": 7540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.898806935574314e-06,
+      "loss": 0.9571,
+      "step": 7541
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.898239843539558e-06,
+      "loss": 1.1586,
+      "step": 7542
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.897672695352065e-06,
+      "loss": 0.9813,
+      "step": 7543
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.897105491022819e-06,
+      "loss": 0.9749,
+      "step": 7544
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.896538230562812e-06,
+      "loss": 0.9751,
+      "step": 7545
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.895970913983034e-06,
+      "loss": 1.0756,
+      "step": 7546
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.895403541294478e-06,
+      "loss": 0.9278,
+      "step": 7547
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.894836112508132e-06,
+      "loss": 0.9799,
+      "step": 7548
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.894268627634994e-06,
+      "loss": 0.8665,
+      "step": 7549
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.893701086686058e-06,
+      "loss": 0.8255,
+      "step": 7550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.893133489672321e-06,
+      "loss": 1.0493,
+      "step": 7551
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.892565836604778e-06,
+      "loss": 0.8403,
+      "step": 7552
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.89199812749443e-06,
+      "loss": 0.9805,
+      "step": 7553
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.891430362352271e-06,
+      "loss": 0.9875,
+      "step": 7554
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.890862541189306e-06,
+      "loss": 0.9922,
+      "step": 7555
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.890294664016536e-06,
+      "loss": 1.0447,
+      "step": 7556
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.889726730844961e-06,
+      "loss": 0.989,
+      "step": 7557
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.889158741685587e-06,
+      "loss": 1.0334,
+      "step": 7558
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.888590696549416e-06,
+      "loss": 1.1022,
+      "step": 7559
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.888022595447458e-06,
+      "loss": 1.047,
+      "step": 7560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.887454438390716e-06,
+      "loss": 0.8497,
+      "step": 7561
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.8868862253902e-06,
+      "loss": 0.8709,
+      "step": 7562
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.886317956456916e-06,
+      "loss": 0.7656,
+      "step": 7563
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.88574963160188e-06,
+      "loss": 0.9522,
+      "step": 7564
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.885181250836097e-06,
+      "loss": 0.9064,
+      "step": 7565
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.884612814170583e-06,
+      "loss": 0.9792,
+      "step": 7566
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.884044321616348e-06,
+      "loss": 0.9669,
+      "step": 7567
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.88347577318441e-06,
+      "loss": 1.1455,
+      "step": 7568
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.882907168885782e-06,
+      "loss": 0.9832,
+      "step": 7569
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.882338508731484e-06,
+      "loss": 0.9451,
+      "step": 7570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.88176979273253e-06,
+      "loss": 0.7478,
+      "step": 7571
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.881201020899938e-06,
+      "loss": 1.0839,
+      "step": 7572
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.880632193244732e-06,
+      "loss": 0.8469,
+      "step": 7573
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.88006330977793e-06,
+      "loss": 0.7956,
+      "step": 7574
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.879494370510556e-06,
+      "loss": 1.0305,
+      "step": 7575
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.87892537545363e-06,
+      "loss": 0.7969,
+      "step": 7576
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.878356324618178e-06,
+      "loss": 1.1809,
+      "step": 7577
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.877787218015226e-06,
+      "loss": 1.1156,
+      "step": 7578
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.877218055655798e-06,
+      "loss": 1.0613,
+      "step": 7579
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.876648837550925e-06,
+      "loss": 0.8635,
+      "step": 7580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.876079563711631e-06,
+      "loss": 1.137,
+      "step": 7581
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.875510234148948e-06,
+      "loss": 0.9806,
+      "step": 7582
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.874940848873908e-06,
+      "loss": 1.0224,
+      "step": 7583
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.874371407897541e-06,
+      "loss": 1.1338,
+      "step": 7584
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.873801911230879e-06,
+      "loss": 1.1146,
+      "step": 7585
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.873232358884958e-06,
+      "loss": 0.9087,
+      "step": 7586
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.872662750870812e-06,
+      "loss": 0.6839,
+      "step": 7587
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.872093087199475e-06,
+      "loss": 0.9105,
+      "step": 7588
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.871523367881986e-06,
+      "loss": 0.9272,
+      "step": 7589
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.870953592929385e-06,
+      "loss": 0.865,
+      "step": 7590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.870383762352709e-06,
+      "loss": 0.8052,
+      "step": 7591
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.869813876162997e-06,
+      "loss": 0.9786,
+      "step": 7592
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.869243934371296e-06,
+      "loss": 0.9697,
+      "step": 7593
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86867393698864e-06,
+      "loss": 0.9859,
+      "step": 7594
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86810388402608e-06,
+      "loss": 0.8973,
+      "step": 7595
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.867533775494656e-06,
+      "loss": 0.9793,
+      "step": 7596
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.866963611405418e-06,
+      "loss": 0.9611,
+      "step": 7597
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.866393391769408e-06,
+      "loss": 0.8678,
+      "step": 7598
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.865823116597678e-06,
+      "loss": 1.0588,
+      "step": 7599
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.865252785901274e-06,
+      "loss": 0.9248,
+      "step": 7600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.864682399691248e-06,
+      "loss": 0.9983,
+      "step": 7601
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.864111957978649e-06,
+      "loss": 0.985,
+      "step": 7602
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.863541460774533e-06,
+      "loss": 1.0214,
+      "step": 7603
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.862970908089949e-06,
+      "loss": 0.9708,
+      "step": 7604
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.862400299935953e-06,
+      "loss": 0.8679,
+      "step": 7605
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.8618296363236e-06,
+      "loss": 0.9956,
+      "step": 7606
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.861258917263948e-06,
+      "loss": 0.9947,
+      "step": 7607
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.860688142768053e-06,
+      "loss": 0.8706,
+      "step": 7608
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.860117312846975e-06,
+      "loss": 0.9865,
+      "step": 7609
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.859546427511773e-06,
+      "loss": 0.9743,
+      "step": 7610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.858975486773506e-06,
+      "loss": 0.9741,
+      "step": 7611
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.858404490643239e-06,
+      "loss": 0.9793,
+      "step": 7612
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.857833439132032e-06,
+      "loss": 1.0241,
+      "step": 7613
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.857262332250952e-06,
+      "loss": 0.9745,
+      "step": 7614
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.856691170011062e-06,
+      "loss": 0.997,
+      "step": 7615
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.856119952423427e-06,
+      "loss": 1.0032,
+      "step": 7616
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.855548679499118e-06,
+      "loss": 0.9749,
+      "step": 7617
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.854977351249199e-06,
+      "loss": 1.0007,
+      "step": 7618
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.854405967684741e-06,
+      "loss": 0.9246,
+      "step": 7619
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.853834528816817e-06,
+      "loss": 0.9437,
+      "step": 7620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.853263034656496e-06,
+      "loss": 0.9147,
+      "step": 7621
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.852691485214852e-06,
+      "loss": 1.0237,
+      "step": 7622
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.852119880502955e-06,
+      "loss": 1.0238,
+      "step": 7623
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.851548220531884e-06,
+      "loss": 1.0814,
+      "step": 7624
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.850976505312713e-06,
+      "loss": 1.1019,
+      "step": 7625
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.850404734856518e-06,
+      "loss": 0.8632,
+      "step": 7626
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.849832909174379e-06,
+      "loss": 0.921,
+      "step": 7627
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.849261028277373e-06,
+      "loss": 0.9335,
+      "step": 7628
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.84868909217658e-06,
+      "loss": 1.1281,
+      "step": 7629
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.848117100883084e-06,
+      "loss": 0.8665,
+      "step": 7630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.847545054407964e-06,
+      "loss": 1.0703,
+      "step": 7631
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.846972952762305e-06,
+      "loss": 0.9481,
+      "step": 7632
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.84640079595719e-06,
+      "loss": 0.9733,
+      "step": 7633
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.845828584003708e-06,
+      "loss": 1.0111,
+      "step": 7634
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.84525631691294e-06,
+      "loss": 1.0709,
+      "step": 7635
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.844683994695977e-06,
+      "loss": 0.9305,
+      "step": 7636
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.844111617363906e-06,
+      "loss": 0.9084,
+      "step": 7637
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.843539184927818e-06,
+      "loss": 1.0112,
+      "step": 7638
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.842966697398803e-06,
+      "loss": 1.0253,
+      "step": 7639
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.842394154787954e-06,
+      "loss": 0.8358,
+      "step": 7640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.841821557106364e-06,
+      "loss": 0.9979,
+      "step": 7641
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.841248904365124e-06,
+      "loss": 0.9668,
+      "step": 7642
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.840676196575333e-06,
+      "loss": 1.0133,
+      "step": 7643
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.840103433748082e-06,
+      "loss": 0.9814,
+      "step": 7644
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.839530615894473e-06,
+      "loss": 0.9939,
+      "step": 7645
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.838957743025604e-06,
+      "loss": 0.9208,
+      "step": 7646
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.838384815152569e-06,
+      "loss": 0.9656,
+      "step": 7647
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.837811832286474e-06,
+      "loss": 1.0067,
+      "step": 7648
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83723879443842e-06,
+      "loss": 1.0242,
+      "step": 7649
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.836665701619506e-06,
+      "loss": 0.8948,
+      "step": 7650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83609255384084e-06,
+      "loss": 0.8764,
+      "step": 7651
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83551935111352e-06,
+      "loss": 1.1152,
+      "step": 7652
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.834946093448658e-06,
+      "loss": 0.8954,
+      "step": 7653
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83437278085736e-06,
+      "loss": 1.1255,
+      "step": 7654
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.833799413350732e-06,
+      "loss": 1.2698,
+      "step": 7655
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83322599093988e-06,
+      "loss": 0.8828,
+      "step": 7656
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83265251363592e-06,
+      "loss": 1.0221,
+      "step": 7657
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.832078981449958e-06,
+      "loss": 0.9247,
+      "step": 7658
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83150539439311e-06,
+      "loss": 0.9083,
+      "step": 7659
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.830931752476487e-06,
+      "loss": 1.0046,
+      "step": 7660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.830358055711204e-06,
+      "loss": 0.9842,
+      "step": 7661
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.829784304108374e-06,
+      "loss": 0.9392,
+      "step": 7662
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.829210497679116e-06,
+      "loss": 0.8797,
+      "step": 7663
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.828636636434546e-06,
+      "loss": 0.8682,
+      "step": 7664
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.828062720385784e-06,
+      "loss": 1.049,
+      "step": 7665
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.827488749543947e-06,
+      "loss": 0.9634,
+      "step": 7666
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.826914723920157e-06,
+      "loss": 0.633,
+      "step": 7667
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.826340643525536e-06,
+      "loss": 0.902,
+      "step": 7668
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.825766508371207e-06,
+      "loss": 0.8839,
+      "step": 7669
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.82519231846829e-06,
+      "loss": 1.0196,
+      "step": 7670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.824618073827914e-06,
+      "loss": 1.0447,
+      "step": 7671
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.824043774461205e-06,
+      "loss": 1.0144,
+      "step": 7672
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.823469420379288e-06,
+      "loss": 1.0963,
+      "step": 7673
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.82289501159329e-06,
+      "loss": 0.8418,
+      "step": 7674
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.822320548114342e-06,
+      "loss": 1.0154,
+      "step": 7675
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.821746029953574e-06,
+      "loss": 1.0318,
+      "step": 7676
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.821171457122116e-06,
+      "loss": 0.9533,
+      "step": 7677
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.820596829631103e-06,
+      "loss": 0.8155,
+      "step": 7678
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.820022147491665e-06,
+      "loss": 0.9107,
+      "step": 7679
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.819447410714938e-06,
+      "loss": 1.0881,
+      "step": 7680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.818872619312055e-06,
+      "loss": 1.0332,
+      "step": 7681
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.818297773294156e-06,
+      "loss": 0.9473,
+      "step": 7682
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.817722872672377e-06,
+      "loss": 0.8569,
+      "step": 7683
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.817147917457857e-06,
+      "loss": 0.9407,
+      "step": 7684
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.816572907661735e-06,
+      "loss": 0.9493,
+      "step": 7685
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.815997843295151e-06,
+      "loss": 0.7599,
+      "step": 7686
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.815422724369248e-06,
+      "loss": 0.8494,
+      "step": 7687
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.814847550895169e-06,
+      "loss": 0.9547,
+      "step": 7688
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.814272322884056e-06,
+      "loss": 1.0345,
+      "step": 7689
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.813697040347055e-06,
+      "loss": 0.8746,
+      "step": 7690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.813121703295312e-06,
+      "loss": 0.894,
+      "step": 7691
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.812546311739976e-06,
+      "loss": 1.098,
+      "step": 7692
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.81197086569219e-06,
+      "loss": 0.9494,
+      "step": 7693
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.81139536516311e-06,
+      "loss": 0.9831,
+      "step": 7694
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.810819810163882e-06,
+      "loss": 0.9175,
+      "step": 7695
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.810244200705655e-06,
+      "loss": 0.9601,
+      "step": 7696
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.809668536799587e-06,
+      "loss": 0.9463,
+      "step": 7697
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.809092818456826e-06,
+      "loss": 0.9103,
+      "step": 7698
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.80851704568853e-06,
+      "loss": 1.2175,
+      "step": 7699
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.807941218505853e-06,
+      "loss": 0.8727,
+      "step": 7700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.807365336919952e-06,
+      "loss": 0.9158,
+      "step": 7701
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.806789400941985e-06,
+      "loss": 0.8659,
+      "step": 7702
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.80621341058311e-06,
+      "loss": 0.7602,
+      "step": 7703
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.805637365854485e-06,
+      "loss": 0.829,
+      "step": 7704
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.805061266767275e-06,
+      "loss": 0.9318,
+      "step": 7705
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.804485113332638e-06,
+      "loss": 0.91,
+      "step": 7706
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.803908905561738e-06,
+      "loss": 0.9177,
+      "step": 7707
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.80333264346574e-06,
+      "loss": 0.9582,
+      "step": 7708
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.802756327055806e-06,
+      "loss": 0.8859,
+      "step": 7709
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.802179956343105e-06,
+      "loss": 1.007,
+      "step": 7710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.801603531338804e-06,
+      "loss": 0.9512,
+      "step": 7711
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.801027052054068e-06,
+      "loss": 0.9805,
+      "step": 7712
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.800450518500071e-06,
+      "loss": 0.9486,
+      "step": 7713
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.799873930687979e-06,
+      "loss": 0.8808,
+      "step": 7714
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.799297288628964e-06,
+      "loss": 0.9045,
+      "step": 7715
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.798720592334199e-06,
+      "loss": 0.9782,
+      "step": 7716
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.798143841814858e-06,
+      "loss": 1.0328,
+      "step": 7717
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.797567037082114e-06,
+      "loss": 0.9116,
+      "step": 7718
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.796990178147144e-06,
+      "loss": 0.9645,
+      "step": 7719
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.796413265021125e-06,
+      "loss": 0.9734,
+      "step": 7720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.795836297715231e-06,
+      "loss": 0.7244,
+      "step": 7721
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.795259276240644e-06,
+      "loss": 0.93,
+      "step": 7722
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.794682200608543e-06,
+      "loss": 1.0914,
+      "step": 7723
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.794105070830107e-06,
+      "loss": 1.0179,
+      "step": 7724
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.793527886916522e-06,
+      "loss": 0.9417,
+      "step": 7725
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.792950648878964e-06,
+      "loss": 0.8967,
+      "step": 7726
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.792373356728623e-06,
+      "loss": 0.9724,
+      "step": 7727
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79179601047668e-06,
+      "loss": 0.8757,
+      "step": 7728
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.791218610134324e-06,
+      "loss": 0.9909,
+      "step": 7729
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.790641155712742e-06,
+      "loss": 0.9769,
+      "step": 7730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.790063647223119e-06,
+      "loss": 0.8656,
+      "step": 7731
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.789486084676646e-06,
+      "loss": 1.0011,
+      "step": 7732
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.788908468084511e-06,
+      "loss": 1.0564,
+      "step": 7733
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.788330797457908e-06,
+      "loss": 0.9209,
+      "step": 7734
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.78775307280803e-06,
+      "loss": 0.9456,
+      "step": 7735
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.787175294146068e-06,
+      "loss": 0.9806,
+      "step": 7736
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.786597461483215e-06,
+      "loss": 1.1022,
+      "step": 7737
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.78601957483067e-06,
+      "loss": 0.9565,
+      "step": 7738
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.785441634199628e-06,
+      "loss": 0.9752,
+      "step": 7739
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.784863639601286e-06,
+      "loss": 0.9701,
+      "step": 7740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.784285591046841e-06,
+      "loss": 1.1037,
+      "step": 7741
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.783707488547496e-06,
+      "loss": 0.9166,
+      "step": 7742
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.78312933211445e-06,
+      "loss": 0.8783,
+      "step": 7743
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.782551121758904e-06,
+      "loss": 0.9453,
+      "step": 7744
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.781972857492059e-06,
+      "loss": 0.998,
+      "step": 7745
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.781394539325124e-06,
+      "loss": 0.9212,
+      "step": 7746
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.7808161672693e-06,
+      "loss": 1.1206,
+      "step": 7747
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.780237741335794e-06,
+      "loss": 0.941,
+      "step": 7748
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.77965926153581e-06,
+      "loss": 0.7077,
+      "step": 7749
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.779080727880562e-06,
+      "loss": 1.1428,
+      "step": 7750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.778502140381253e-06,
+      "loss": 1.1485,
+      "step": 7751
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.777923499049095e-06,
+      "loss": 0.8566,
+      "step": 7752
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.7773448038953e-06,
+      "loss": 0.9836,
+      "step": 7753
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.77676605493108e-06,
+      "loss": 1.0295,
+      "step": 7754
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.776187252167647e-06,
+      "loss": 0.9703,
+      "step": 7755
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.775608395616217e-06,
+      "loss": 0.9311,
+      "step": 7756
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.775029485288002e-06,
+      "loss": 0.8918,
+      "step": 7757
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.77445052119422e-06,
+      "loss": 1.1459,
+      "step": 7758
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.773871503346089e-06,
+      "loss": 0.8666,
+      "step": 7759
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.773292431754828e-06,
+      "loss": 0.8663,
+      "step": 7760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.772713306431655e-06,
+      "loss": 0.818,
+      "step": 7761
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.77213412738779e-06,
+      "loss": 0.8597,
+      "step": 7762
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.771554894634455e-06,
+      "loss": 0.7465,
+      "step": 7763
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.770975608182872e-06,
+      "loss": 1.0337,
+      "step": 7764
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.770396268044267e-06,
+      "loss": 0.8423,
+      "step": 7765
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.769816874229862e-06,
+      "loss": 0.8845,
+      "step": 7766
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.769237426750884e-06,
+      "loss": 0.9714,
+      "step": 7767
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.768657925618556e-06,
+      "loss": 0.7474,
+      "step": 7768
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.76807837084411e-06,
+      "loss": 0.8786,
+      "step": 7769
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.767498762438775e-06,
+      "loss": 0.8632,
+      "step": 7770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.76691910041378e-06,
+      "loss": 0.8699,
+      "step": 7771
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.766339384780353e-06,
+      "loss": 1.0509,
+      "step": 7772
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.765759615549728e-06,
+      "loss": 0.9735,
+      "step": 7773
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.765179792733139e-06,
+      "loss": 0.8452,
+      "step": 7774
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.764599916341817e-06,
+      "loss": 1.0234,
+      "step": 7775
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.764019986387002e-06,
+      "loss": 0.9125,
+      "step": 7776
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.763440002879923e-06,
+      "loss": 0.937,
+      "step": 7777
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.762859965831823e-06,
+      "loss": 0.9619,
+      "step": 7778
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.762279875253938e-06,
+      "loss": 0.8832,
+      "step": 7779
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.761699731157506e-06,
+      "loss": 0.8351,
+      "step": 7780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.761119533553768e-06,
+      "loss": 0.8253,
+      "step": 7781
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.760539282453966e-06,
+      "loss": 1.1197,
+      "step": 7782
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.759958977869341e-06,
+      "loss": 1.0426,
+      "step": 7783
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.759378619811137e-06,
+      "loss": 0.9198,
+      "step": 7784
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.7587982082906e-06,
+      "loss": 0.9214,
+      "step": 7785
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.758217743318972e-06,
+      "loss": 0.9382,
+      "step": 7786
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.7576372249075e-06,
+      "loss": 1.0133,
+      "step": 7787
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.757056653067434e-06,
+      "loss": 1.0077,
+      "step": 7788
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.75647602781002e-06,
+      "loss": 0.9999,
+      "step": 7789
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.755895349146508e-06,
+      "loss": 0.8705,
+      "step": 7790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.755314617088149e-06,
+      "loss": 0.9865,
+      "step": 7791
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.754733831646195e-06,
+      "loss": 0.898,
+      "step": 7792
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.754152992831896e-06,
+      "loss": 1.0574,
+      "step": 7793
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.753572100656508e-06,
+      "loss": 0.9242,
+      "step": 7794
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.752991155131286e-06,
+      "loss": 1.0509,
+      "step": 7795
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.752410156267485e-06,
+      "loss": 0.7131,
+      "step": 7796
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.751829104076361e-06,
+      "loss": 0.8348,
+      "step": 7797
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.751247998569174e-06,
+      "loss": 1.0014,
+      "step": 7798
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.750666839757179e-06,
+      "loss": 0.9767,
+      "step": 7799
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.750085627651637e-06,
+      "loss": 1.1326,
+      "step": 7800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.749504362263813e-06,
+      "loss": 0.8824,
+      "step": 7801
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.748923043604964e-06,
+      "loss": 1.1383,
+      "step": 7802
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.748341671686355e-06,
+      "loss": 0.9093,
+      "step": 7803
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.747760246519249e-06,
+      "loss": 1.1164,
+      "step": 7804
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.747178768114912e-06,
+      "loss": 0.9532,
+      "step": 7805
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.746597236484609e-06,
+      "loss": 0.9483,
+      "step": 7806
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.746015651639608e-06,
+      "loss": 0.9339,
+      "step": 7807
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.745434013591177e-06,
+      "loss": 0.8585,
+      "step": 7808
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.744852322350583e-06,
+      "loss": 0.9499,
+      "step": 7809
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.744270577929099e-06,
+      "loss": 1.0665,
+      "step": 7810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.743688780337995e-06,
+      "loss": 0.8559,
+      "step": 7811
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.743106929588544e-06,
+      "loss": 0.8397,
+      "step": 7812
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.742525025692018e-06,
+      "loss": 1.0214,
+      "step": 7813
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.741943068659692e-06,
+      "loss": 1.011,
+      "step": 7814
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.741361058502841e-06,
+      "loss": 0.9752,
+      "step": 7815
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.74077899523274e-06,
+      "loss": 1.0018,
+      "step": 7816
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.74019687886067e-06,
+      "loss": 1.0779,
+      "step": 7817
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.739614709397904e-06,
+      "loss": 1.0588,
+      "step": 7818
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.739032486855726e-06,
+      "loss": 0.7973,
+      "step": 7819
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.738450211245416e-06,
+      "loss": 0.8995,
+      "step": 7820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.737867882578254e-06,
+      "loss": 1.1497,
+      "step": 7821
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.737285500865522e-06,
+      "loss": 0.9629,
+      "step": 7822
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.736703066118506e-06,
+      "loss": 0.9605,
+      "step": 7823
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.736120578348486e-06,
+      "loss": 0.887,
+      "step": 7824
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.735538037566753e-06,
+      "loss": 0.8828,
+      "step": 7825
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.734955443784588e-06,
+      "loss": 0.8122,
+      "step": 7826
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.734372797013284e-06,
+      "loss": 0.9343,
+      "step": 7827
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.733790097264127e-06,
+      "loss": 0.9064,
+      "step": 7828
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.733207344548408e-06,
+      "loss": 0.8761,
+      "step": 7829
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.732624538877416e-06,
+      "loss": 0.9484,
+      "step": 7830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.732041680262443e-06,
+      "loss": 0.9225,
+      "step": 7831
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.73145876871478e-06,
+      "loss": 0.9584,
+      "step": 7832
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.730875804245726e-06,
+      "loss": 1.0542,
+      "step": 7833
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.730292786866572e-06,
+      "loss": 1.015,
+      "step": 7834
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.729709716588616e-06,
+      "loss": 0.9578,
+      "step": 7835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.729126593423151e-06,
+      "loss": 0.9624,
+      "step": 7836
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.728543417381478e-06,
+      "loss": 0.7832,
+      "step": 7837
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.727960188474897e-06,
+      "loss": 1.065,
+      "step": 7838
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.727376906714703e-06,
+      "loss": 0.9361,
+      "step": 7839
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.726793572112203e-06,
+      "loss": 1.1522,
+      "step": 7840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.726210184678695e-06,
+      "loss": 0.912,
+      "step": 7841
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.725626744425482e-06,
+      "loss": 1.0321,
+      "step": 7842
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.72504325136387e-06,
+      "loss": 0.9784,
+      "step": 7843
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.724459705505163e-06,
+      "loss": 1.079,
+      "step": 7844
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.723876106860669e-06,
+      "loss": 0.97,
+      "step": 7845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.723292455441692e-06,
+      "loss": 1.0039,
+      "step": 7846
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.72270875125954e-06,
+      "loss": 0.933,
+      "step": 7847
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.722124994325525e-06,
+      "loss": 0.9409,
+      "step": 7848
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.721541184650956e-06,
+      "loss": 0.8975,
+      "step": 7849
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.720957322247142e-06,
+      "loss": 1.0285,
+      "step": 7850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.7203734071254e-06,
+      "loss": 0.9272,
+      "step": 7851
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.719789439297038e-06,
+      "loss": 1.0355,
+      "step": 7852
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.719205418773373e-06,
+      "loss": 1.108,
+      "step": 7853
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71862134556572e-06,
+      "loss": 0.7888,
+      "step": 7854
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.718037219685395e-06,
+      "loss": 1.0077,
+      "step": 7855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.717453041143716e-06,
+      "loss": 1.0277,
+      "step": 7856
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.716868809952001e-06,
+      "loss": 1.0028,
+      "step": 7857
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.716284526121568e-06,
+      "loss": 0.9219,
+      "step": 7858
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71570018966374e-06,
+      "loss": 1.0078,
+      "step": 7859
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.715115800589835e-06,
+      "loss": 0.92,
+      "step": 7860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.714531358911179e-06,
+      "loss": 1.0219,
+      "step": 7861
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.713946864639091e-06,
+      "loss": 1.031,
+      "step": 7862
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.7133623177849e-06,
+      "loss": 0.8481,
+      "step": 7863
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.712777718359929e-06,
+      "loss": 0.9733,
+      "step": 7864
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.712193066375505e-06,
+      "loss": 0.9409,
+      "step": 7865
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.711608361842957e-06,
+      "loss": 0.8374,
+      "step": 7866
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71102360477361e-06,
+      "loss": 0.883,
+      "step": 7867
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.710438795178799e-06,
+      "loss": 1.0635,
+      "step": 7868
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.709853933069848e-06,
+      "loss": 0.9327,
+      "step": 7869
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.709269018458092e-06,
+      "loss": 0.8475,
+      "step": 7870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.708684051354865e-06,
+      "loss": 0.8351,
+      "step": 7871
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.708099031771498e-06,
+      "loss": 1.0179,
+      "step": 7872
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.707513959719326e-06,
+      "loss": 0.9085,
+      "step": 7873
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.706928835209686e-06,
+      "loss": 0.994,
+      "step": 7874
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.706343658253913e-06,
+      "loss": 0.8147,
+      "step": 7875
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.705758428863347e-06,
+      "loss": 0.8672,
+      "step": 7876
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.705173147049326e-06,
+      "loss": 0.9178,
+      "step": 7877
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.704587812823186e-06,
+      "loss": 0.8798,
+      "step": 7878
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.70400242619627e-06,
+      "loss": 1.1386,
+      "step": 7879
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.703416987179924e-06,
+      "loss": 1.0638,
+      "step": 7880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.702831495785486e-06,
+      "loss": 1.1156,
+      "step": 7881
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.7022459520243e-06,
+      "loss": 1.117,
+      "step": 7882
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.701660355907712e-06,
+      "loss": 0.9399,
+      "step": 7883
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.701074707447068e-06,
+      "loss": 1.0767,
+      "step": 7884
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.700489006653713e-06,
+      "loss": 1.0153,
+      "step": 7885
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.699903253538996e-06,
+      "loss": 0.8195,
+      "step": 7886
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.699317448114267e-06,
+      "loss": 1.0687,
+      "step": 7887
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.698731590390874e-06,
+      "loss": 0.9716,
+      "step": 7888
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.698145680380168e-06,
+      "loss": 1.0391,
+      "step": 7889
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.697559718093504e-06,
+      "loss": 1.0092,
+      "step": 7890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69697370354223e-06,
+      "loss": 0.9773,
+      "step": 7891
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.696387636737703e-06,
+      "loss": 0.9852,
+      "step": 7892
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.695801517691277e-06,
+      "loss": 1.0584,
+      "step": 7893
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.695215346414308e-06,
+      "loss": 1.0408,
+      "step": 7894
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.694629122918155e-06,
+      "loss": 0.9447,
+      "step": 7895
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.694042847214171e-06,
+      "loss": 0.8147,
+      "step": 7896
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69345651931372e-06,
+      "loss": 1.0928,
+      "step": 7897
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69287013922816e-06,
+      "loss": 1.1894,
+      "step": 7898
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69228370696885e-06,
+      "loss": 0.8546,
+      "step": 7899
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.691697222547156e-06,
+      "loss": 0.6762,
+      "step": 7900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.691110685974437e-06,
+      "loss": 0.801,
+      "step": 7901
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69052409726206e-06,
+      "loss": 1.0055,
+      "step": 7902
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.689937456421388e-06,
+      "loss": 0.9785,
+      "step": 7903
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.689350763463788e-06,
+      "loss": 0.8561,
+      "step": 7904
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.68876401840063e-06,
+      "loss": 0.9527,
+      "step": 7905
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.688177221243275e-06,
+      "loss": 0.9743,
+      "step": 7906
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6875903720031e-06,
+      "loss": 0.9293,
+      "step": 7907
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.687003470691468e-06,
+      "loss": 1.0508,
+      "step": 7908
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.686416517319755e-06,
+      "loss": 0.8888,
+      "step": 7909
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.68582951189933e-06,
+      "loss": 0.7932,
+      "step": 7910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.685242454441569e-06,
+      "loss": 0.8745,
+      "step": 7911
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.684655344957844e-06,
+      "loss": 0.8609,
+      "step": 7912
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.684068183459529e-06,
+      "loss": 0.9432,
+      "step": 7913
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.683480969958005e-06,
+      "loss": 0.9398,
+      "step": 7914
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.682893704464642e-06,
+      "loss": 0.9987,
+      "step": 7915
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.682306386990825e-06,
+      "loss": 0.8858,
+      "step": 7916
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.681719017547929e-06,
+      "loss": 1.052,
+      "step": 7917
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.681131596147336e-06,
+      "loss": 0.9911,
+      "step": 7918
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.680544122800427e-06,
+      "loss": 0.9585,
+      "step": 7919
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.67995659751858e-06,
+      "loss": 0.9571,
+      "step": 7920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.679369020313185e-06,
+      "loss": 0.9205,
+      "step": 7921
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.678781391195624e-06,
+      "loss": 1.0298,
+      "step": 7922
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.67819371017728e-06,
+      "loss": 1.0274,
+      "step": 7923
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.67760597726954e-06,
+      "loss": 0.9749,
+      "step": 7924
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.677018192483792e-06,
+      "loss": 1.0042,
+      "step": 7925
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.676430355831423e-06,
+      "loss": 1.2355,
+      "step": 7926
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.675842467323824e-06,
+      "loss": 0.9574,
+      "step": 7927
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.675254526972385e-06,
+      "loss": 1.0337,
+      "step": 7928
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.674666534788495e-06,
+      "loss": 0.9654,
+      "step": 7929
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.674078490783547e-06,
+      "loss": 1.2454,
+      "step": 7930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.673490394968935e-06,
+      "loss": 0.8896,
+      "step": 7931
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.672902247356053e-06,
+      "loss": 1.0315,
+      "step": 7932
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.672314047956298e-06,
+      "loss": 0.9753,
+      "step": 7933
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.671725796781064e-06,
+      "loss": 0.9086,
+      "step": 7934
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.671137493841748e-06,
+      "loss": 0.9215,
+      "step": 7935
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.67054913914975e-06,
+      "loss": 1.0352,
+      "step": 7936
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.669960732716466e-06,
+      "loss": 0.9196,
+      "step": 7937
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.669372274553299e-06,
+      "loss": 0.9921,
+      "step": 7938
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.668783764671652e-06,
+      "loss": 1.1352,
+      "step": 7939
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.668195203082924e-06,
+      "loss": 0.9337,
+      "step": 7940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.667606589798517e-06,
+      "loss": 0.8378,
+      "step": 7941
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66701792482984e-06,
+      "loss": 0.9778,
+      "step": 7942
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.666429208188295e-06,
+      "loss": 0.9531,
+      "step": 7943
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.665840439885289e-06,
+      "loss": 0.9631,
+      "step": 7944
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.665251619932228e-06,
+      "loss": 1.0092,
+      "step": 7945
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.664662748340523e-06,
+      "loss": 0.9426,
+      "step": 7946
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66407382512158e-06,
+      "loss": 0.9435,
+      "step": 7947
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.663484850286812e-06,
+      "loss": 0.9903,
+      "step": 7948
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66289582384763e-06,
+      "loss": 0.7314,
+      "step": 7949
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.662306745815445e-06,
+      "loss": 0.9908,
+      "step": 7950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66171761620167e-06,
+      "loss": 0.9136,
+      "step": 7951
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.661128435017721e-06,
+      "loss": 0.9684,
+      "step": 7952
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66053920227501e-06,
+      "loss": 0.9086,
+      "step": 7953
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.659949917984959e-06,
+      "loss": 0.8334,
+      "step": 7954
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.659360582158979e-06,
+      "loss": 0.8363,
+      "step": 7955
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.65877119480849e-06,
+      "loss": 1.063,
+      "step": 7956
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.658181755944915e-06,
+      "loss": 0.831,
+      "step": 7957
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.65759226557967e-06,
+      "loss": 0.9673,
+      "step": 7958
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.657002723724178e-06,
+      "loss": 0.7963,
+      "step": 7959
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.656413130389862e-06,
+      "loss": 0.9194,
+      "step": 7960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.655823485588142e-06,
+      "loss": 0.8736,
+      "step": 7961
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.655233789330445e-06,
+      "loss": 0.9872,
+      "step": 7962
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.654644041628196e-06,
+      "loss": 1.0383,
+      "step": 7963
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.654054242492821e-06,
+      "loss": 1.101,
+      "step": 7964
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.653464391935748e-06,
+      "loss": 1.0746,
+      "step": 7965
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.652874489968402e-06,
+      "loss": 0.9769,
+      "step": 7966
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.652284536602217e-06,
+      "loss": 1.0256,
+      "step": 7967
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.65169453184862e-06,
+      "loss": 1.0892,
+      "step": 7968
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.651104475719043e-06,
+      "loss": 1.044,
+      "step": 7969
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.650514368224918e-06,
+      "loss": 0.8734,
+      "step": 7970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.64992420937768e-06,
+      "loss": 0.9786,
+      "step": 7971
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.649333999188759e-06,
+      "loss": 1.0764,
+      "step": 7972
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.648743737669592e-06,
+      "loss": 0.8589,
+      "step": 7973
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.64815342483162e-06,
+      "loss": 0.8807,
+      "step": 7974
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.647563060686272e-06,
+      "loss": 0.9806,
+      "step": 7975
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.646972645244993e-06,
+      "loss": 0.9798,
+      "step": 7976
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.646382178519218e-06,
+      "loss": 0.7131,
+      "step": 7977
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.645791660520388e-06,
+      "loss": 1.1755,
+      "step": 7978
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.645201091259945e-06,
+      "loss": 0.9086,
+      "step": 7979
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.644610470749332e-06,
+      "loss": 1.0516,
+      "step": 7980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.644019798999991e-06,
+      "loss": 1.0139,
+      "step": 7981
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.643429076023365e-06,
+      "loss": 1.1403,
+      "step": 7982
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.642838301830902e-06,
+      "loss": 1.1854,
+      "step": 7983
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.642247476434044e-06,
+      "loss": 0.9386,
+      "step": 7984
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.641656599844242e-06,
+      "loss": 1.0565,
+      "step": 7985
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.641065672072942e-06,
+      "loss": 1.073,
+      "step": 7986
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.640474693131594e-06,
+      "loss": 0.934,
+      "step": 7987
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.63988366303165e-06,
+      "loss": 1.0634,
+      "step": 7988
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.639292581784554e-06,
+      "loss": 0.8635,
+      "step": 7989
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.638701449401766e-06,
+      "loss": 0.9324,
+      "step": 7990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.638110265894736e-06,
+      "loss": 0.8415,
+      "step": 7991
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.637519031274918e-06,
+      "loss": 0.9221,
+      "step": 7992
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.636927745553765e-06,
+      "loss": 1.0164,
+      "step": 7993
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.636336408742739e-06,
+      "loss": 0.9279,
+      "step": 7994
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.635745020853289e-06,
+      "loss": 0.8956,
+      "step": 7995
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.635153581896879e-06,
+      "loss": 1.1281,
+      "step": 7996
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.634562091884968e-06,
+      "loss": 0.8565,
+      "step": 7997
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.63397055082901e-06,
+      "loss": 0.9341,
+      "step": 7998
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.633378958740474e-06,
+      "loss": 0.9174,
+      "step": 7999
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.632787315630816e-06,
+      "loss": 0.9846,
+      "step": 8000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.632195621511502e-06,
+      "loss": 1.0049,
+      "step": 8001
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.631603876393994e-06,
+      "loss": 0.7999,
+      "step": 8002
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.631012080289756e-06,
+      "loss": 1.0925,
+      "step": 8003
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.63042023321026e-06,
+      "loss": 0.9153,
+      "step": 8004
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.629828335166967e-06,
+      "loss": 0.9323,
+      "step": 8005
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.629236386171346e-06,
+      "loss": 0.9396,
+      "step": 8006
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.628644386234864e-06,
+      "loss": 1.0711,
+      "step": 8007
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.628052335368996e-06,
+      "loss": 0.9911,
+      "step": 8008
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.627460233585211e-06,
+      "loss": 0.7016,
+      "step": 8009
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.626868080894978e-06,
+      "loss": 0.8727,
+      "step": 8010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.626275877309773e-06,
+      "loss": 0.8281,
+      "step": 8011
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.625683622841069e-06,
+      "loss": 0.9548,
+      "step": 8012
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.62509131750034e-06,
+      "loss": 0.8663,
+      "step": 8013
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.624498961299061e-06,
+      "loss": 0.9321,
+      "step": 8014
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.62390655424871e-06,
+      "loss": 0.9597,
+      "step": 8015
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6233140963607655e-06,
+      "loss": 1.043,
+      "step": 8016
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.622721587646707e-06,
+      "loss": 1.0673,
+      "step": 8017
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6221290281180104e-06,
+      "loss": 0.9472,
+      "step": 8018
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.621536417786159e-06,
+      "loss": 0.8826,
+      "step": 8019
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.620943756662636e-06,
+      "loss": 1.0206,
+      "step": 8020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.62035104475892e-06,
+      "loss": 1.0944,
+      "step": 8021
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.619758282086498e-06,
+      "loss": 0.9272,
+      "step": 8022
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.619165468656853e-06,
+      "loss": 1.1227,
+      "step": 8023
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.618572604481471e-06,
+      "loss": 0.9578,
+      "step": 8024
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.61797968957184e-06,
+      "loss": 1.0174,
+      "step": 8025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.617386723939444e-06,
+      "loss": 0.8687,
+      "step": 8026
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6167937075957765e-06,
+      "loss": 1.067,
+      "step": 8027
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.616200640552324e-06,
+      "loss": 1.0252,
+      "step": 8028
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.615607522820576e-06,
+      "loss": 0.9727,
+      "step": 8029
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.615014354412027e-06,
+      "loss": 1.0932,
+      "step": 8030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.614421135338168e-06,
+      "loss": 0.9623,
+      "step": 8031
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.613827865610493e-06,
+      "loss": 0.9669,
+      "step": 8032
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.613234545240494e-06,
+      "loss": 1.0004,
+      "step": 8033
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.61264117423967e-06,
+      "loss": 0.9256,
+      "step": 8034
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.612047752619515e-06,
+      "loss": 0.9026,
+      "step": 8035
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.611454280391529e-06,
+      "loss": 1.132,
+      "step": 8036
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.610860757567209e-06,
+      "loss": 0.8579,
+      "step": 8037
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.6102671841580515e-06,
+      "loss": 1.0721,
+      "step": 8038
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.609673560175561e-06,
+      "loss": 1.0407,
+      "step": 8039
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.609079885631236e-06,
+      "loss": 0.8861,
+      "step": 8040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.6084861605365825e-06,
+      "loss": 1.0679,
+      "step": 8041
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.6078923849031e-06,
+      "loss": 0.8641,
+      "step": 8042
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.607298558742293e-06,
+      "loss": 0.8091,
+      "step": 8043
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.606704682065669e-06,
+      "loss": 0.9533,
+      "step": 8044
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.6061107548847324e-06,
+      "loss": 1.0823,
+      "step": 8045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.605516777210993e-06,
+      "loss": 0.895,
+      "step": 8046
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.604922749055954e-06,
+      "loss": 1.0311,
+      "step": 8047
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.604328670431129e-06,
+      "loss": 0.8236,
+      "step": 8048
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.603734541348027e-06,
+      "loss": 0.9456,
+      "step": 8049
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.60314036181816e-06,
+      "loss": 0.986,
+      "step": 8050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.602546131853037e-06,
+      "loss": 0.9451,
+      "step": 8051
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.601951851464174e-06,
+      "loss": 0.9342,
+      "step": 8052
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.601357520663084e-06,
+      "loss": 0.863,
+      "step": 8053
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.600763139461283e-06,
+      "loss": 1.0405,
+      "step": 8054
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.600168707870285e-06,
+      "loss": 0.8949,
+      "step": 8055
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.599574225901608e-06,
+      "loss": 0.8624,
+      "step": 8056
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5989796935667704e-06,
+      "loss": 1.0112,
+      "step": 8057
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.598385110877291e-06,
+      "loss": 1.0218,
+      "step": 8058
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.597790477844689e-06,
+      "loss": 0.7674,
+      "step": 8059
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.597195794480485e-06,
+      "loss": 0.9917,
+      "step": 8060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.596601060796204e-06,
+      "loss": 0.8661,
+      "step": 8061
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.596006276803365e-06,
+      "loss": 1.0067,
+      "step": 8062
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.595411442513494e-06,
+      "loss": 1.0853,
+      "step": 8063
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.594816557938114e-06,
+      "loss": 1.1315,
+      "step": 8064
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.594221623088752e-06,
+      "loss": 0.8837,
+      "step": 8065
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.593626637976935e-06,
+      "loss": 0.9517,
+      "step": 8066
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.59303160261419e-06,
+      "loss": 0.8324,
+      "step": 8067
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.592436517012046e-06,
+      "loss": 1.1125,
+      "step": 8068
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5918413811820324e-06,
+      "loss": 0.9905,
+      "step": 8069
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5912461951356805e-06,
+      "loss": 0.9613,
+      "step": 8070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.59065095888452e-06,
+      "loss": 1.0659,
+      "step": 8071
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.590055672440087e-06,
+      "loss": 1.0284,
+      "step": 8072
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.589460335813911e-06,
+      "loss": 0.9615,
+      "step": 8073
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5888649490175295e-06,
+      "loss": 0.8826,
+      "step": 8074
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5882695120624765e-06,
+      "loss": 0.963,
+      "step": 8075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.587674024960287e-06,
+      "loss": 0.917,
+      "step": 8076
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.587078487722502e-06,
+      "loss": 1.0632,
+      "step": 8077
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5864829003606585e-06,
+      "loss": 1.0359,
+      "step": 8078
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.585887262886296e-06,
+      "loss": 0.9764,
+      "step": 8079
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.585291575310952e-06,
+      "loss": 0.8382,
+      "step": 8080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.584695837646172e-06,
+      "loss": 1.1739,
+      "step": 8081
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.584100049903496e-06,
+      "loss": 1.1767,
+      "step": 8082
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.583504212094468e-06,
+      "loss": 0.9779,
+      "step": 8083
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.582908324230631e-06,
+      "loss": 0.9044,
+      "step": 8084
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.582312386323532e-06,
+      "loss": 0.9361,
+      "step": 8085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.581716398384717e-06,
+      "loss": 1.1315,
+      "step": 8086
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.581120360425731e-06,
+      "loss": 0.8747,
+      "step": 8087
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.580524272458123e-06,
+      "loss": 1.0194,
+      "step": 8088
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.579928134493444e-06,
+      "loss": 0.9526,
+      "step": 8089
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.579331946543242e-06,
+      "loss": 0.7967,
+      "step": 8090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.57873570861907e-06,
+      "loss": 0.8774,
+      "step": 8091
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.578139420732477e-06,
+      "loss": 0.8789,
+      "step": 8092
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.577543082895018e-06,
+      "loss": 0.8606,
+      "step": 8093
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.576946695118246e-06,
+      "loss": 1.0155,
+      "step": 8094
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5763502574137174e-06,
+      "loss": 1.036,
+      "step": 8095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.575753769792986e-06,
+      "loss": 0.7455,
+      "step": 8096
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.575157232267609e-06,
+      "loss": 0.9312,
+      "step": 8097
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.574560644849147e-06,
+      "loss": 0.914,
+      "step": 8098
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5739640075491546e-06,
+      "loss": 0.8181,
+      "step": 8099
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.573367320379193e-06,
+      "loss": 1.0272,
+      "step": 8100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.572770583350826e-06,
+      "loss": 1.0013,
+      "step": 8101
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.57217379647561e-06,
+      "loss": 0.9869,
+      "step": 8102
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.571576959765111e-06,
+      "loss": 0.9554,
+      "step": 8103
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.570980073230893e-06,
+      "loss": 1.1092,
+      "step": 8104
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.570383136884519e-06,
+      "loss": 1.2089,
+      "step": 8105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.569786150737551e-06,
+      "loss": 1.0343,
+      "step": 8106
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.569189114801563e-06,
+      "loss": 0.9957,
+      "step": 8107
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.568592029088117e-06,
+      "loss": 0.9471,
+      "step": 8108
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.567994893608784e-06,
+      "loss": 0.8802,
+      "step": 8109
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.567397708375133e-06,
+      "loss": 0.962,
+      "step": 8110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.566800473398732e-06,
+      "loss": 0.849,
+      "step": 8111
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.566203188691156e-06,
+      "loss": 0.9019,
+      "step": 8112
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.565605854263975e-06,
+      "loss": 0.9354,
+      "step": 8113
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.565008470128761e-06,
+      "loss": 1.0028,
+      "step": 8114
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5644110362970915e-06,
+      "loss": 0.8561,
+      "step": 8115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.563813552780539e-06,
+      "loss": 0.955,
+      "step": 8116
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.56321601959068e-06,
+      "loss": 1.0492,
+      "step": 8117
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.562618436739093e-06,
+      "loss": 0.9268,
+      "step": 8118
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.562020804237356e-06,
+      "loss": 1.1382,
+      "step": 8119
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5614231220970445e-06,
+      "loss": 0.9575,
+      "step": 8120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.560825390329742e-06,
+      "loss": 1.0464,
+      "step": 8121
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5602276089470305e-06,
+      "loss": 1.0381,
+      "step": 8122
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5596297779604874e-06,
+      "loss": 0.9891,
+      "step": 8123
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.559031897381699e-06,
+      "loss": 0.8419,
+      "step": 8124
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.558433967222248e-06,
+      "loss": 1.1118,
+      "step": 8125
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.55783598749372e-06,
+      "loss": 0.9446,
+      "step": 8126
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.557237958207699e-06,
+      "loss": 0.9703,
+      "step": 8127
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.556639879375772e-06,
+      "loss": 0.9636,
+      "step": 8128
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.556041751009527e-06,
+      "loss": 0.9792,
+      "step": 8129
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.555443573120554e-06,
+      "loss": 0.8202,
+      "step": 8130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.554845345720442e-06,
+      "loss": 0.8377,
+      "step": 8131
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5542470688207805e-06,
+      "loss": 1.0458,
+      "step": 8132
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.553648742433162e-06,
+      "loss": 0.8471,
+      "step": 8133
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.553050366569177e-06,
+      "loss": 0.8306,
+      "step": 8134
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5524519412404195e-06,
+      "loss": 0.9048,
+      "step": 8135
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.551853466458486e-06,
+      "loss": 0.9526,
+      "step": 8136
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5512549422349714e-06,
+      "loss": 1.0152,
+      "step": 8137
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.550656368581471e-06,
+      "loss": 1.0611,
+      "step": 8138
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.55005774550958e-06,
+      "loss": 0.9899,
+      "step": 8139
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5494590730309e-06,
+      "loss": 1.0794,
+      "step": 8140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.548860351157028e-06,
+      "loss": 0.871,
+      "step": 8141
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.548261579899565e-06,
+      "loss": 0.9811,
+      "step": 8142
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.547662759270113e-06,
+      "loss": 0.9654,
+      "step": 8143
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.547063889280272e-06,
+      "loss": 1.0487,
+      "step": 8144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.546464969941646e-06,
+      "loss": 0.906,
+      "step": 8145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.545866001265838e-06,
+      "loss": 0.9615,
+      "step": 8146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5452669832644545e-06,
+      "loss": 1.1211,
+      "step": 8147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.544667915949102e-06,
+      "loss": 0.9671,
+      "step": 8148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.544068799331384e-06,
+      "loss": 1.059,
+      "step": 8149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.543469633422911e-06,
+      "loss": 0.9169,
+      "step": 8150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.542870418235291e-06,
+      "loss": 1.0738,
+      "step": 8151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.542271153780132e-06,
+      "loss": 0.8926,
+      "step": 8152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.541671840069047e-06,
+      "loss": 1.1047,
+      "step": 8153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.541072477113647e-06,
+      "loss": 0.9626,
+      "step": 8154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.540473064925544e-06,
+      "loss": 1.0159,
+      "step": 8155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.539873603516353e-06,
+      "loss": 0.9409,
+      "step": 8156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.539274092897683e-06,
+      "loss": 0.9865,
+      "step": 8157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.538674533081159e-06,
+      "loss": 0.9619,
+      "step": 8158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.538074924078388e-06,
+      "loss": 0.778,
+      "step": 8159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.537475265900992e-06,
+      "loss": 1.0025,
+      "step": 8160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.536875558560588e-06,
+      "loss": 0.9892,
+      "step": 8161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5362758020687954e-06,
+      "loss": 0.9682,
+      "step": 8162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.535675996437235e-06,
+      "loss": 1.1179,
+      "step": 8163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.535076141677526e-06,
+      "loss": 0.9213,
+      "step": 8164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.534476237801294e-06,
+      "loss": 0.9902,
+      "step": 8165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.533876284820158e-06,
+      "loss": 0.9528,
+      "step": 8166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.533276282745745e-06,
+      "loss": 0.9655,
+      "step": 8167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5326762315896795e-06,
+      "loss": 1.0634,
+      "step": 8168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.532076131363585e-06,
+      "loss": 1.0155,
+      "step": 8169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.531475982079089e-06,
+      "loss": 0.9343,
+      "step": 8170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.530875783747822e-06,
+      "loss": 0.8535,
+      "step": 8171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5302755363814095e-06,
+      "loss": 1.0238,
+      "step": 8172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.529675239991483e-06,
+      "loss": 0.8476,
+      "step": 8173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5290748945896715e-06,
+      "loss": 1.0308,
+      "step": 8174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.52847450018761e-06,
+      "loss": 0.9053,
+      "step": 8175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.527874056796927e-06,
+      "loss": 0.8972,
+      "step": 8176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.527273564429257e-06,
+      "loss": 0.841,
+      "step": 8177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.526673023096236e-06,
+      "loss": 0.8353,
+      "step": 8178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.526072432809498e-06,
+      "loss": 0.9616,
+      "step": 8179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.52547179358068e-06,
+      "loss": 1.0093,
+      "step": 8180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.524871105421418e-06,
+      "loss": 0.8986,
+      "step": 8181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.524270368343352e-06,
+      "loss": 0.9261,
+      "step": 8182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.523669582358119e-06,
+      "loss": 0.9673,
+      "step": 8183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.523068747477361e-06,
+      "loss": 0.7527,
+      "step": 8184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.522467863712719e-06,
+      "loss": 0.8507,
+      "step": 8185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.521866931075834e-06,
+      "loss": 0.9166,
+      "step": 8186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.521265949578349e-06,
+      "loss": 1.1198,
+      "step": 8187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5206649192319085e-06,
+      "loss": 0.956,
+      "step": 8188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5200638400481575e-06,
+      "loss": 1.0432,
+      "step": 8189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.51946271203874e-06,
+      "loss": 0.717,
+      "step": 8190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.518861535215305e-06,
+      "loss": 0.9947,
+      "step": 8191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.518260309589498e-06,
+      "loss": 0.9621,
+      "step": 8192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.517659035172968e-06,
+      "loss": 0.9708,
+      "step": 8193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.517057711977367e-06,
+      "loss": 1.0353,
+      "step": 8194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.516456340014344e-06,
+      "loss": 1.0153,
+      "step": 8195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.515854919295548e-06,
+      "loss": 0.9451,
+      "step": 8196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.515253449832636e-06,
+      "loss": 0.9281,
+      "step": 8197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.514651931637258e-06,
+      "loss": 0.9845,
+      "step": 8198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5140503647210695e-06,
+      "loss": 0.8367,
+      "step": 8199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.513448749095725e-06,
+      "loss": 1.0159,
+      "step": 8200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5128470847728816e-06,
+      "loss": 1.0316,
+      "step": 8201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.512245371764197e-06,
+      "loss": 0.9483,
+      "step": 8202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.511643610081328e-06,
+      "loss": 0.8711,
+      "step": 8203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.51104179973593e-06,
+      "loss": 0.9107,
+      "step": 8204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5104399407396696e-06,
+      "loss": 0.9103,
+      "step": 8205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.509838033104205e-06,
+      "loss": 0.9077,
+      "step": 8206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.509236076841198e-06,
+      "loss": 0.9976,
+      "step": 8207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.508634071962311e-06,
+      "loss": 0.9102,
+      "step": 8208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.508032018479208e-06,
+      "loss": 1.0244,
+      "step": 8209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.507429916403553e-06,
+      "loss": 0.7976,
+      "step": 8210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.506827765747012e-06,
+      "loss": 1.1079,
+      "step": 8211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.506225566521252e-06,
+      "loss": 0.9257,
+      "step": 8212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.50562331873794e-06,
+      "loss": 0.9522,
+      "step": 8213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.505021022408745e-06,
+      "loss": 0.9015,
+      "step": 8214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.504418677545336e-06,
+      "loss": 0.9148,
+      "step": 8215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.503816284159384e-06,
+      "loss": 1.0352,
+      "step": 8216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.503213842262558e-06,
+      "loss": 0.8591,
+      "step": 8217
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.502611351866532e-06,
+      "loss": 1.1495,
+      "step": 8218
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.50200881298298e-06,
+      "loss": 1.0001,
+      "step": 8219
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.501406225623573e-06,
+      "loss": 0.9244,
+      "step": 8220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.500803589799988e-06,
+      "loss": 0.9656,
+      "step": 8221
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.500200905523901e-06,
+      "loss": 1.0123,
+      "step": 8222
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4995981728069885e-06,
+      "loss": 0.9835,
+      "step": 8223
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.498995391660929e-06,
+      "loss": 1.0133,
+      "step": 8224
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.498392562097401e-06,
+      "loss": 1.0762,
+      "step": 8225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.497789684128083e-06,
+      "loss": 1.036,
+      "step": 8226
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4971867577646566e-06,
+      "loss": 1.0306,
+      "step": 8227
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.496583783018803e-06,
+      "loss": 0.8881,
+      "step": 8228
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.495980759902204e-06,
+      "loss": 0.9505,
+      "step": 8229
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.495377688426546e-06,
+      "loss": 1.0586,
+      "step": 8230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.49477456860351e-06,
+      "loss": 1.0605,
+      "step": 8231
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.494171400444783e-06,
+      "loss": 0.8984,
+      "step": 8232
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.49356818396205e-06,
+      "loss": 1.0863,
+      "step": 8233
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.492964919167001e-06,
+      "loss": 0.9493,
+      "step": 8234
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.492361606071321e-06,
+      "loss": 1.072,
+      "step": 8235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.491758244686701e-06,
+      "loss": 0.9163,
+      "step": 8236
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.49115483502483e-06,
+      "loss": 0.9946,
+      "step": 8237
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.490551377097398e-06,
+      "loss": 0.8931,
+      "step": 8238
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.489947870916099e-06,
+      "loss": 1.0154,
+      "step": 8239
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.489344316492624e-06,
+      "loss": 0.9428,
+      "step": 8240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.488740713838669e-06,
+      "loss": 1.0651,
+      "step": 8241
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.488137062965925e-06,
+      "loss": 0.9544,
+      "step": 8242
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.487533363886093e-06,
+      "loss": 1.1064,
+      "step": 8243
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.486929616610864e-06,
+      "loss": 0.9242,
+      "step": 8244
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.486325821151938e-06,
+      "loss": 0.9293,
+      "step": 8245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.485721977521015e-06,
+      "loss": 1.0256,
+      "step": 8246
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.48511808572979e-06,
+      "loss": 0.9581,
+      "step": 8247
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4845141457899675e-06,
+      "loss": 1.0794,
+      "step": 8248
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.483910157713246e-06,
+      "loss": 0.8191,
+      "step": 8249
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.483306121511329e-06,
+      "loss": 1.0881,
+      "step": 8250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4827020371959204e-06,
+      "loss": 1.0089,
+      "step": 8251
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.482097904778721e-06,
+      "loss": 0.8526,
+      "step": 8252
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.481493724271439e-06,
+      "loss": 1.0289,
+      "step": 8253
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.480889495685779e-06,
+      "loss": 1.0169,
+      "step": 8254
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.480285219033448e-06,
+      "loss": 1.0459,
+      "step": 8255
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.479680894326153e-06,
+      "loss": 0.8917,
+      "step": 8256
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.479076521575602e-06,
+      "loss": 0.9919,
+      "step": 8257
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.478472100793507e-06,
+      "loss": 0.9501,
+      "step": 8258
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4778676319915774e-06,
+      "loss": 1.0568,
+      "step": 8259
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.477263115181523e-06,
+      "loss": 1.0044,
+      "step": 8260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.47665855037506e-06,
+      "loss": 0.9635,
+      "step": 8261
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.476053937583898e-06,
+      "loss": 1.0912,
+      "step": 8262
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.475449276819753e-06,
+      "loss": 0.9636,
+      "step": 8263
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.474844568094339e-06,
+      "loss": 0.9714,
+      "step": 8264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.474239811419374e-06,
+      "loss": 0.8272,
+      "step": 8265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.473635006806574e-06,
+      "loss": 0.8872,
+      "step": 8266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.473030154267655e-06,
+      "loss": 1.0111,
+      "step": 8267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.472425253814339e-06,
+      "loss": 0.8963,
+      "step": 8268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.471820305458345e-06,
+      "loss": 0.9346,
+      "step": 8269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.471215309211392e-06,
+      "loss": 0.9743,
+      "step": 8270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.470610265085205e-06,
+      "loss": 0.9294,
+      "step": 8271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.470005173091504e-06,
+      "loss": 0.9773,
+      "step": 8272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.469400033242011e-06,
+      "loss": 1.059,
+      "step": 8273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.468794845548454e-06,
+      "loss": 0.9042,
+      "step": 8274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4681896100225575e-06,
+      "loss": 0.9414,
+      "step": 8275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.467584326676046e-06,
+      "loss": 0.9006,
+      "step": 8276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.466978995520648e-06,
+      "loss": 1.0981,
+      "step": 8277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4663736165680925e-06,
+      "loss": 0.9196,
+      "step": 8278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.465768189830107e-06,
+      "loss": 0.8914,
+      "step": 8279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.465162715318422e-06,
+      "loss": 0.9516,
+      "step": 8280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4645571930447705e-06,
+      "loss": 0.947,
+      "step": 8281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.46395162302088e-06,
+      "loss": 0.9039,
+      "step": 8282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4633460052584874e-06,
+      "loss": 0.9203,
+      "step": 8283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.462740339769323e-06,
+      "loss": 0.9526,
+      "step": 8284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.462134626565125e-06,
+      "loss": 0.7585,
+      "step": 8285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.461528865657625e-06,
+      "loss": 0.939,
+      "step": 8286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.460923057058562e-06,
+      "loss": 0.8711,
+      "step": 8287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4603172007796744e-06,
+      "loss": 0.8746,
+      "step": 8288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.459711296832699e-06,
+      "loss": 0.9772,
+      "step": 8289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4591053452293745e-06,
+      "loss": 0.9502,
+      "step": 8290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.458499345981442e-06,
+      "loss": 0.9865,
+      "step": 8291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.457893299100642e-06,
+      "loss": 1.1996,
+      "step": 8292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.457287204598716e-06,
+      "loss": 1.0785,
+      "step": 8293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.45668106248741e-06,
+      "loss": 0.862,
+      "step": 8294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.456074872778465e-06,
+      "loss": 0.8617,
+      "step": 8295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.455468635483626e-06,
+      "loss": 0.8839,
+      "step": 8296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.454862350614638e-06,
+      "loss": 0.788,
+      "step": 8297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.45425601818325e-06,
+      "loss": 1.0295,
+      "step": 8298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.453649638201209e-06,
+      "loss": 0.8262,
+      "step": 8299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.453043210680262e-06,
+      "loss": 1.0207,
+      "step": 8300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.45243673563216e-06,
+      "loss": 0.8704,
+      "step": 8301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.451830213068653e-06,
+      "loss": 1.0893,
+      "step": 8302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.451223643001489e-06,
+      "loss": 0.9016,
+      "step": 8303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4506170254424245e-06,
+      "loss": 0.941,
+      "step": 8304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.450010360403212e-06,
+      "loss": 0.931,
+      "step": 8305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.449403647895605e-06,
+      "loss": 0.8292,
+      "step": 8306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.448796887931355e-06,
+      "loss": 1.0029,
+      "step": 8307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.448190080522224e-06,
+      "loss": 0.8545,
+      "step": 8308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.447583225679964e-06,
+      "loss": 0.8458,
+      "step": 8309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.446976323416335e-06,
+      "loss": 0.8052,
+      "step": 8310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4463693737430945e-06,
+      "loss": 0.9844,
+      "step": 8311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.445762376672001e-06,
+      "loss": 0.9861,
+      "step": 8312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.445155332214817e-06,
+      "loss": 0.7996,
+      "step": 8313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.444548240383304e-06,
+      "loss": 0.9267,
+      "step": 8314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.443941101189221e-06,
+      "loss": 0.9587,
+      "step": 8315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.443333914644336e-06,
+      "loss": 0.8426,
+      "step": 8316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.44272668076041e-06,
+      "loss": 0.9125,
+      "step": 8317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.442119399549208e-06,
+      "loss": 0.9458,
+      "step": 8318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.441512071022497e-06,
+      "loss": 1.0382,
+      "step": 8319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4409046951920435e-06,
+      "loss": 1.0894,
+      "step": 8320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.440297272069615e-06,
+      "loss": 1.1362,
+      "step": 8321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.439689801666981e-06,
+      "loss": 1.2009,
+      "step": 8322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.439082283995911e-06,
+      "loss": 1.089,
+      "step": 8323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.438474719068174e-06,
+      "loss": 1.119,
+      "step": 8324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.437867106895542e-06,
+      "loss": 0.7693,
+      "step": 8325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.437259447489788e-06,
+      "loss": 0.9841,
+      "step": 8326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.436651740862686e-06,
+      "loss": 1.0222,
+      "step": 8327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.436043987026009e-06,
+      "loss": 0.9948,
+      "step": 8328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.435436185991531e-06,
+      "loss": 1.0305,
+      "step": 8329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.434828337771031e-06,
+      "loss": 0.9556,
+      "step": 8330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.434220442376284e-06,
+      "loss": 0.9518,
+      "step": 8331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.433612499819067e-06,
+      "loss": 0.8169,
+      "step": 8332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4330045101111615e-06,
+      "loss": 0.9549,
+      "step": 8333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.432396473264344e-06,
+      "loss": 0.9294,
+      "step": 8334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4317883892903975e-06,
+      "loss": 0.9417,
+      "step": 8335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.431180258201103e-06,
+      "loss": 0.9997,
+      "step": 8336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.430572080008242e-06,
+      "loss": 0.924,
+      "step": 8337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.429963854723598e-06,
+      "loss": 0.8165,
+      "step": 8338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.429355582358956e-06,
+      "loss": 0.9887,
+      "step": 8339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4287472629261e-06,
+      "loss": 0.8866,
+      "step": 8340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.428138896436817e-06,
+      "loss": 0.9333,
+      "step": 8341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.427530482902894e-06,
+      "loss": 0.8972,
+      "step": 8342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4269220223361175e-06,
+      "loss": 0.771,
+      "step": 8343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.426313514748278e-06,
+      "loss": 1.054,
+      "step": 8344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.425704960151166e-06,
+      "loss": 0.9801,
+      "step": 8345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.425096358556568e-06,
+      "loss": 1.1485,
+      "step": 8346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.42448770997628e-06,
+      "loss": 0.8439,
+      "step": 8347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4238790144220914e-06,
+      "loss": 0.9178,
+      "step": 8348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.423270271905798e-06,
+      "loss": 0.8988,
+      "step": 8349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.422661482439191e-06,
+      "loss": 0.9012,
+      "step": 8350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.422052646034069e-06,
+      "loss": 0.884,
+      "step": 8351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.421443762702225e-06,
+      "loss": 1.0372,
+      "step": 8352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.420834832455459e-06,
+      "loss": 0.8645,
+      "step": 8353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.420225855305567e-06,
+      "loss": 0.9793,
+      "step": 8354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.419616831264347e-06,
+      "loss": 1.0018,
+      "step": 8355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4190077603435996e-06,
+      "loss": 0.9897,
+      "step": 8356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.418398642555127e-06,
+      "loss": 0.9202,
+      "step": 8357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.417789477910728e-06,
+      "loss": 0.8014,
+      "step": 8358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.417180266422207e-06,
+      "loss": 0.9005,
+      "step": 8359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.416571008101366e-06,
+      "loss": 0.9003,
+      "step": 8360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.41596170296001e-06,
+      "loss": 1.0428,
+      "step": 8361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.415352351009945e-06,
+      "loss": 1.0233,
+      "step": 8362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.414742952262975e-06,
+      "loss": 0.9678,
+      "step": 8363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.41413350673091e-06,
+      "loss": 0.8189,
+      "step": 8364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.413524014425554e-06,
+      "loss": 0.8554,
+      "step": 8365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4129144753587175e-06,
+      "loss": 1.0722,
+      "step": 8366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.41230488954221e-06,
+      "loss": 1.0972,
+      "step": 8367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.411695256987844e-06,
+      "loss": 0.919,
+      "step": 8368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.411085577707429e-06,
+      "loss": 1.0067,
+      "step": 8369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4104758517127784e-06,
+      "loss": 1.2552,
+      "step": 8370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.409866079015704e-06,
+      "loss": 1.0714,
+      "step": 8371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.409256259628022e-06,
+      "loss": 1.0287,
+      "step": 8372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.408646393561546e-06,
+      "loss": 0.8748,
+      "step": 8373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.408036480828093e-06,
+      "loss": 0.9384,
+      "step": 8374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.407426521439478e-06,
+      "loss": 1.0022,
+      "step": 8375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.406816515407521e-06,
+      "loss": 0.7621,
+      "step": 8376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.40620646274404e-06,
+      "loss": 1.0768,
+      "step": 8377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.405596363460855e-06,
+      "loss": 0.9816,
+      "step": 8378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.404986217569785e-06,
+      "loss": 0.959,
+      "step": 8379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.404376025082654e-06,
+      "loss": 0.9423,
+      "step": 8380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.403765786011281e-06,
+      "loss": 0.9337,
+      "step": 8381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4031555003674925e-06,
+      "loss": 0.9741,
+      "step": 8382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.402545168163111e-06,
+      "loss": 1.1374,
+      "step": 8383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.401934789409962e-06,
+      "loss": 0.9287,
+      "step": 8384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.401324364119872e-06,
+      "loss": 1.0987,
+      "step": 8385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.400713892304666e-06,
+      "loss": 1.1285,
+      "step": 8386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.400103373976173e-06,
+      "loss": 0.9311,
+      "step": 8387
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.399492809146223e-06,
+      "loss": 1.075,
+      "step": 8388
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3988821978266424e-06,
+      "loss": 0.7976,
+      "step": 8389
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.398271540029264e-06,
+      "loss": 0.876,
+      "step": 8390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.397660835765918e-06,
+      "loss": 0.8866,
+      "step": 8391
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.397050085048438e-06,
+      "loss": 1.0748,
+      "step": 8392
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.396439287888657e-06,
+      "loss": 0.913,
+      "step": 8393
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.395828444298407e-06,
+      "loss": 1.0783,
+      "step": 8394
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.395217554289524e-06,
+      "loss": 1.0186,
+      "step": 8395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.394606617873845e-06,
+      "loss": 0.9003,
+      "step": 8396
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.393995635063207e-06,
+      "loss": 0.9583,
+      "step": 8397
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.393384605869445e-06,
+      "loss": 0.8787,
+      "step": 8398
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3927735303044e-06,
+      "loss": 1.0255,
+      "step": 8399
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3921624083799105e-06,
+      "loss": 0.9929,
+      "step": 8400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.391551240107817e-06,
+      "loss": 0.8612,
+      "step": 8401
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.390940025499961e-06,
+      "loss": 0.945,
+      "step": 8402
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3903287645681846e-06,
+      "loss": 0.9553,
+      "step": 8403
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3897174573243305e-06,
+      "loss": 1.171,
+      "step": 8404
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.389106103780242e-06,
+      "loss": 1.1296,
+      "step": 8405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.388494703947765e-06,
+      "loss": 1.032,
+      "step": 8406
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.387883257838745e-06,
+      "loss": 0.821,
+      "step": 8407
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.387271765465027e-06,
+      "loss": 0.9951,
+      "step": 8408
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.386660226838462e-06,
+      "loss": 1.0891,
+      "step": 8409
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.386048641970895e-06,
+      "loss": 1.0727,
+      "step": 8410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.385437010874179e-06,
+      "loss": 1.0647,
+      "step": 8411
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.38482533356016e-06,
+      "loss": 0.8715,
+      "step": 8412
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.384213610040693e-06,
+      "loss": 1.0241,
+      "step": 8413
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.383601840327625e-06,
+      "loss": 0.8412,
+      "step": 8414
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.382990024432815e-06,
+      "loss": 0.9942,
+      "step": 8415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.382378162368114e-06,
+      "loss": 0.8667,
+      "step": 8416
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3817662541453745e-06,
+      "loss": 1.0654,
+      "step": 8417
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.381154299776455e-06,
+      "loss": 0.9791,
+      "step": 8418
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.38054229927321e-06,
+      "loss": 0.9074,
+      "step": 8419
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3799302526475e-06,
+      "loss": 1.0757,
+      "step": 8420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.37931815991118e-06,
+      "loss": 1.1437,
+      "step": 8421
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.378706021076111e-06,
+      "loss": 1.0065,
+      "step": 8422
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.378093836154152e-06,
+      "loss": 0.8125,
+      "step": 8423
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.377481605157164e-06,
+      "loss": 0.9355,
+      "step": 8424
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.376869328097009e-06,
+      "loss": 0.8413,
+      "step": 8425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.376257004985551e-06,
+      "loss": 1.0399,
+      "step": 8426
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.375644635834652e-06,
+      "loss": 0.9958,
+      "step": 8427
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.375032220656178e-06,
+      "loss": 1.0491,
+      "step": 8428
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.374419759461992e-06,
+      "loss": 0.9314,
+      "step": 8429
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.373807252263963e-06,
+      "loss": 1.0828,
+      "step": 8430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3731946990739565e-06,
+      "loss": 1.0604,
+      "step": 8431
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.372582099903841e-06,
+      "loss": 1.0566,
+      "step": 8432
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.371969454765486e-06,
+      "loss": 0.7879,
+      "step": 8433
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3713567636707605e-06,
+      "loss": 0.9274,
+      "step": 8434
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.370744026631536e-06,
+      "loss": 0.8877,
+      "step": 8435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.370131243659683e-06,
+      "loss": 0.9886,
+      "step": 8436
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.369518414767074e-06,
+      "loss": 0.9518,
+      "step": 8437
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.368905539965584e-06,
+      "loss": 0.9397,
+      "step": 8438
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.368292619267087e-06,
+      "loss": 0.9413,
+      "step": 8439
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.367679652683457e-06,
+      "loss": 0.9254,
+      "step": 8440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.367066640226571e-06,
+      "loss": 0.9978,
+      "step": 8441
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.366453581908306e-06,
+      "loss": 1.0421,
+      "step": 8442
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.36584047774054e-06,
+      "loss": 1.0955,
+      "step": 8443
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.365227327735151e-06,
+      "loss": 0.9313,
+      "step": 8444
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.364614131904019e-06,
+      "loss": 0.8282,
+      "step": 8445
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.364000890259024e-06,
+      "loss": 0.9274,
+      "step": 8446
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.36338760281205e-06,
+      "loss": 0.8765,
+      "step": 8447
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.362774269574975e-06,
+      "loss": 0.8667,
+      "step": 8448
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.362160890559685e-06,
+      "loss": 0.9941,
+      "step": 8449
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.361547465778065e-06,
+      "loss": 1.0076,
+      "step": 8450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.360933995241996e-06,
+      "loss": 0.9452,
+      "step": 8451
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.36032047896337e-06,
+      "loss": 0.8489,
+      "step": 8452
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.359706916954068e-06,
+      "loss": 1.0666,
+      "step": 8453
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.359093309225979e-06,
+      "loss": 0.906,
+      "step": 8454
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.358479655790994e-06,
+      "loss": 0.9707,
+      "step": 8455
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.357865956660999e-06,
+      "loss": 0.7443,
+      "step": 8456
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.357252211847886e-06,
+      "loss": 1.0269,
+      "step": 8457
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.356638421363548e-06,
+      "loss": 1.0243,
+      "step": 8458
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.356024585219874e-06,
+      "loss": 1.0121,
+      "step": 8459
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.355410703428757e-06,
+      "loss": 0.7783,
+      "step": 8460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3547967760020935e-06,
+      "loss": 0.8963,
+      "step": 8461
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.354182802951776e-06,
+      "loss": 0.871,
+      "step": 8462
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.353568784289701e-06,
+      "loss": 0.9628,
+      "step": 8463
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.352954720027765e-06,
+      "loss": 0.9774,
+      "step": 8464
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.352340610177864e-06,
+      "loss": 0.9834,
+      "step": 8465
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.351726454751898e-06,
+      "loss": 1.0361,
+      "step": 8466
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.351112253761765e-06,
+      "loss": 0.9467,
+      "step": 8467
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3504980072193665e-06,
+      "loss": 1.0427,
+      "step": 8468
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.349883715136601e-06,
+      "loss": 1.147,
+      "step": 8469
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3492693775253725e-06,
+      "loss": 0.9626,
+      "step": 8470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3486549943975825e-06,
+      "loss": 0.9674,
+      "step": 8471
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.348040565765134e-06,
+      "loss": 0.9372,
+      "step": 8472
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.347426091639932e-06,
+      "loss": 0.9807,
+      "step": 8473
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.346811572033883e-06,
+      "loss": 1.1564,
+      "step": 8474
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.346197006958892e-06,
+      "loss": 0.977,
+      "step": 8475
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.345582396426866e-06,
+      "loss": 1.15,
+      "step": 8476
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.344967740449713e-06,
+      "loss": 0.8944,
+      "step": 8477
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.344353039039341e-06,
+      "loss": 1.1147,
+      "step": 8478
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.343738292207663e-06,
+      "loss": 1.0165,
+      "step": 8479
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.343123499966585e-06,
+      "loss": 0.8841,
+      "step": 8480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.342508662328022e-06,
+      "loss": 1.0002,
+      "step": 8481
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3418937793038844e-06,
+      "loss": 0.7562,
+      "step": 8482
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.341278850906086e-06,
+      "loss": 1.0029,
+      "step": 8483
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.340663877146541e-06,
+      "loss": 0.8805,
+      "step": 8484
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.340048858037165e-06,
+      "loss": 1.0974,
+      "step": 8485
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3394337935898725e-06,
+      "loss": 0.8126,
+      "step": 8486
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.338818683816582e-06,
+      "loss": 0.9188,
+      "step": 8487
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3382035287292085e-06,
+      "loss": 1.1237,
+      "step": 8488
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.337588328339672e-06,
+      "loss": 0.7822,
+      "step": 8489
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.336973082659893e-06,
+      "loss": 0.9616,
+      "step": 8490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.33635779170179e-06,
+      "loss": 1.0159,
+      "step": 8491
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.335742455477286e-06,
+      "loss": 0.9465,
+      "step": 8492
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.335127073998301e-06,
+      "loss": 0.8692,
+      "step": 8493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.334511647276759e-06,
+      "loss": 1.099,
+      "step": 8494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.333896175324582e-06,
+      "loss": 0.8668,
+      "step": 8495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.333280658153697e-06,
+      "loss": 1.0466,
+      "step": 8496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.332665095776028e-06,
+      "loss": 0.9262,
+      "step": 8497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.332049488203503e-06,
+      "loss": 1.0123,
+      "step": 8498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.331433835448047e-06,
+      "loss": 0.9781,
+      "step": 8499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.33081813752159e-06,
+      "loss": 0.8857,
+      "step": 8500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.33020239443606e-06,
+      "loss": 0.8955,
+      "step": 8501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.329586606203387e-06,
+      "loss": 0.8166,
+      "step": 8502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.328970772835502e-06,
+      "loss": 1.0313,
+      "step": 8503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.328354894344338e-06,
+      "loss": 1.0586,
+      "step": 8504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3277389707418245e-06,
+      "loss": 1.0763,
+      "step": 8505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.327123002039897e-06,
+      "loss": 0.8781,
+      "step": 8506
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.326506988250488e-06,
+      "loss": 0.9683,
+      "step": 8507
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3258909293855356e-06,
+      "loss": 0.7788,
+      "step": 8508
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.325274825456973e-06,
+      "loss": 0.9807,
+      "step": 8509
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.324658676476738e-06,
+      "loss": 0.9838,
+      "step": 8510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.324042482456769e-06,
+      "loss": 1.0582,
+      "step": 8511
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.323426243409003e-06,
+      "loss": 0.9826,
+      "step": 8512
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.322809959345383e-06,
+      "loss": 1.0815,
+      "step": 8513
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.322193630277845e-06,
+      "loss": 0.9686,
+      "step": 8514
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.321577256218333e-06,
+      "loss": 1.0073,
+      "step": 8515
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3209608371787875e-06,
+      "loss": 0.7557,
+      "step": 8516
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.320344373171153e-06,
+      "loss": 0.8566,
+      "step": 8517
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.319727864207371e-06,
+      "loss": 0.9693,
+      "step": 8518
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.319111310299389e-06,
+      "loss": 0.8375,
+      "step": 8519
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.318494711459152e-06,
+      "loss": 0.9013,
+      "step": 8520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.317878067698606e-06,
+      "loss": 0.874,
+      "step": 8521
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.317261379029698e-06,
+      "loss": 0.9878,
+      "step": 8522
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.316644645464376e-06,
+      "loss": 1.0478,
+      "step": 8523
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3160278670145905e-06,
+      "loss": 0.9102,
+      "step": 8524
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.315411043692288e-06,
+      "loss": 0.7804,
+      "step": 8525
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.314794175509425e-06,
+      "loss": 0.9668,
+      "step": 8526
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.314177262477947e-06,
+      "loss": 0.9468,
+      "step": 8527
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.313560304609812e-06,
+      "loss": 0.8346,
+      "step": 8528
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.312943301916969e-06,
+      "loss": 1.212,
+      "step": 8529
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.312326254411373e-06,
+      "loss": 0.7936,
+      "step": 8530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.311709162104983e-06,
+      "loss": 0.7783,
+      "step": 8531
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.311092025009751e-06,
+      "loss": 0.9437,
+      "step": 8532
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.310474843137637e-06,
+      "loss": 0.9628,
+      "step": 8533
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.309857616500595e-06,
+      "loss": 0.964,
+      "step": 8534
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.309240345110587e-06,
+      "loss": 0.9143,
+      "step": 8535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.30862302897957e-06,
+      "loss": 1.0577,
+      "step": 8536
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.308005668119507e-06,
+      "loss": 0.9197,
+      "step": 8537
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3073882625423566e-06,
+      "loss": 0.9702,
+      "step": 8538
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.306770812260082e-06,
+      "loss": 1.1443,
+      "step": 8539
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.306153317284647e-06,
+      "loss": 0.9151,
+      "step": 8540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3055357776280125e-06,
+      "loss": 0.8112,
+      "step": 8541
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.304918193302146e-06,
+      "loss": 1.0135,
+      "step": 8542
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.304300564319013e-06,
+      "loss": 0.8639,
+      "step": 8543
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.303682890690581e-06,
+      "loss": 0.9959,
+      "step": 8544
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.303065172428814e-06,
+      "loss": 0.9341,
+      "step": 8545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.302447409545681e-06,
+      "loss": 1.1278,
+      "step": 8546
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.301829602053152e-06,
+      "loss": 0.9843,
+      "step": 8547
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.301211749963197e-06,
+      "loss": 0.879,
+      "step": 8548
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.300593853287786e-06,
+      "loss": 1.0166,
+      "step": 8549
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.299975912038893e-06,
+      "loss": 0.9079,
+      "step": 8550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2993579262284854e-06,
+      "loss": 1.051,
+      "step": 8551
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2987398958685405e-06,
+      "loss": 0.9224,
+      "step": 8552
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.298121820971032e-06,
+      "loss": 0.9384,
+      "step": 8553
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.297503701547935e-06,
+      "loss": 0.8166,
+      "step": 8554
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.296885537611225e-06,
+      "loss": 0.9121,
+      "step": 8555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.296267329172877e-06,
+      "loss": 0.9866,
+      "step": 8556
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.295649076244872e-06,
+      "loss": 0.7281,
+      "step": 8557
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.295030778839186e-06,
+      "loss": 1.0446,
+      "step": 8558
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2944124369678e-06,
+      "loss": 1.0401,
+      "step": 8559
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.293794050642693e-06,
+      "loss": 0.9836,
+      "step": 8560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.293175619875848e-06,
+      "loss": 1.0365,
+      "step": 8561
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.292557144679244e-06,
+      "loss": 0.7582,
+      "step": 8562
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.291938625064865e-06,
+      "loss": 1.0582,
+      "step": 8563
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.291320061044695e-06,
+      "loss": 0.856,
+      "step": 8564
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2907014526307195e-06,
+      "loss": 1.0123,
+      "step": 8565
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.290082799834924e-06,
+      "loss": 0.8593,
+      "step": 8566
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.289464102669291e-06,
+      "loss": 0.7902,
+      "step": 8567
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.288845361145812e-06,
+      "loss": 0.9756,
+      "step": 8568
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2882265752764736e-06,
+      "loss": 1.0575,
+      "step": 8569
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.287607745073264e-06,
+      "loss": 0.9031,
+      "step": 8570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.286988870548173e-06,
+      "loss": 1.0227,
+      "step": 8571
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.28636995171319e-06,
+      "loss": 0.7996,
+      "step": 8572
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.28575098858031e-06,
+      "loss": 0.8179,
+      "step": 8573
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.285131981161523e-06,
+      "loss": 0.9047,
+      "step": 8574
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.284512929468822e-06,
+      "loss": 1.1353,
+      "step": 8575
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2838938335142e-06,
+      "loss": 0.9136,
+      "step": 8576
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.283274693309655e-06,
+      "loss": 0.781,
+      "step": 8577
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.282655508867178e-06,
+      "loss": 1.0884,
+      "step": 8578
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.282036280198771e-06,
+      "loss": 0.9505,
+      "step": 8579
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.281417007316427e-06,
+      "loss": 0.9649,
+      "step": 8580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.280797690232148e-06,
+      "loss": 0.9741,
+      "step": 8581
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2801783289579285e-06,
+      "loss": 0.8167,
+      "step": 8582
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.279558923505773e-06,
+      "loss": 0.8834,
+      "step": 8583
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.27893947388768e-06,
+      "loss": 1.1045,
+      "step": 8584
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.278319980115651e-06,
+      "loss": 0.7929,
+      "step": 8585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.27770044220169e-06,
+      "loss": 0.9892,
+      "step": 8586
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2770808601577995e-06,
+      "loss": 1.0682,
+      "step": 8587
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.276461233995984e-06,
+      "loss": 1.0424,
+      "step": 8588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2758415637282476e-06,
+      "loss": 0.8684,
+      "step": 8589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.275221849366599e-06,
+      "loss": 0.9116,
+      "step": 8590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.274602090923042e-06,
+      "loss": 0.973,
+      "step": 8591
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.273982288409585e-06,
+      "loss": 0.7955,
+      "step": 8592
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2733624418382375e-06,
+      "loss": 0.8993,
+      "step": 8593
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.272742551221007e-06,
+      "loss": 1.0234,
+      "step": 8594
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.272122616569907e-06,
+      "loss": 0.9474,
+      "step": 8595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.271502637896946e-06,
+      "loss": 0.8863,
+      "step": 8596
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.270882615214136e-06,
+      "loss": 0.9802,
+      "step": 8597
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2702625485334924e-06,
+      "loss": 0.8723,
+      "step": 8598
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.269642437867025e-06,
+      "loss": 0.9652,
+      "step": 8599
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.26902228322675e-06,
+      "loss": 0.9367,
+      "step": 8600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2684020846246835e-06,
+      "loss": 0.9242,
+      "step": 8601
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2677818420728415e-06,
+      "loss": 0.8677,
+      "step": 8602
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.267161555583242e-06,
+      "loss": 0.8786,
+      "step": 8603
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2665412251679e-06,
+      "loss": 1.04,
+      "step": 8604
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.265920850838835e-06,
+      "loss": 1.1801,
+      "step": 8605
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.265300432608069e-06,
+      "loss": 1.0168,
+      "step": 8606
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.264679970487622e-06,
+      "loss": 1.0371,
+      "step": 8607
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.264059464489512e-06,
+      "loss": 0.9222,
+      "step": 8608
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.263438914625764e-06,
+      "loss": 0.961,
+      "step": 8609
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.262818320908403e-06,
+      "loss": 1.1208,
+      "step": 8610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.262197683349446e-06,
+      "loss": 0.7158,
+      "step": 8611
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.261577001960924e-06,
+      "loss": 1.0589,
+      "step": 8612
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.260956276754861e-06,
+      "loss": 1.0289,
+      "step": 8613
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.260335507743283e-06,
+      "loss": 1.0493,
+      "step": 8614
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.259714694938218e-06,
+      "loss": 1.1243,
+      "step": 8615
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.259093838351693e-06,
+      "loss": 0.9194,
+      "step": 8616
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.258472937995736e-06,
+      "loss": 0.8205,
+      "step": 8617
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.25785199388238e-06,
+      "loss": 0.9245,
+      "step": 8618
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.257231006023653e-06,
+      "loss": 1.0177,
+      "step": 8619
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.256609974431586e-06,
+      "loss": 0.8594,
+      "step": 8620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.255988899118215e-06,
+      "loss": 0.888,
+      "step": 8621
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2553677800955695e-06,
+      "loss": 1.1347,
+      "step": 8622
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.254746617375684e-06,
+      "loss": 0.8424,
+      "step": 8623
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.254125410970595e-06,
+      "loss": 1.1228,
+      "step": 8624
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.253504160892337e-06,
+      "loss": 0.7685,
+      "step": 8625
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.252882867152948e-06,
+      "loss": 0.8181,
+      "step": 8626
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.252261529764465e-06,
+      "loss": 0.869,
+      "step": 8627
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.251640148738926e-06,
+      "loss": 1.0243,
+      "step": 8628
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.251018724088367e-06,
+      "loss": 0.9321,
+      "step": 8629
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.250397255824832e-06,
+      "loss": 1.0215,
+      "step": 8630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.249775743960362e-06,
+      "loss": 0.8883,
+      "step": 8631
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.249154188506997e-06,
+      "loss": 0.9562,
+      "step": 8632
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.24853258947678e-06,
+      "loss": 0.8262,
+      "step": 8633
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.247910946881752e-06,
+      "loss": 1.0904,
+      "step": 8634
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2472892607339605e-06,
+      "loss": 0.8427,
+      "step": 8635
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.246667531045451e-06,
+      "loss": 0.978,
+      "step": 8636
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2460457578282665e-06,
+      "loss": 0.8783,
+      "step": 8637
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.245423941094454e-06,
+      "loss": 0.9644,
+      "step": 8638
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.244802080856064e-06,
+      "loss": 1.0289,
+      "step": 8639
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.244180177125139e-06,
+      "loss": 0.8072,
+      "step": 8640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.243558229913736e-06,
+      "loss": 0.9451,
+      "step": 8641
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2429362392339e-06,
+      "loss": 0.9795,
+      "step": 8642
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.242314205097684e-06,
+      "loss": 0.8566,
+      "step": 8643
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.241692127517137e-06,
+      "loss": 0.981,
+      "step": 8644
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.241070006504316e-06,
+      "loss": 0.9431,
+      "step": 8645
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.24044784207127e-06,
+      "loss": 1.0592,
+      "step": 8646
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.239825634230057e-06,
+      "loss": 1.1067,
+      "step": 8647
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.239203382992729e-06,
+      "loss": 0.9904,
+      "step": 8648
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.238581088371344e-06,
+      "loss": 1.0486,
+      "step": 8649
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2379587503779584e-06,
+      "loss": 0.9593,
+      "step": 8650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.237336369024628e-06,
+      "loss": 0.885,
+      "step": 8651
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2367139443234136e-06,
+      "loss": 1.0761,
+      "step": 8652
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.236091476286374e-06,
+      "loss": 0.9587,
+      "step": 8653
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.235468964925571e-06,
+      "loss": 0.9083,
+      "step": 8654
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.234846410253062e-06,
+      "loss": 1.0109,
+      "step": 8655
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.234223812280911e-06,
+      "loss": 0.8792,
+      "step": 8656
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2336011710211815e-06,
+      "loss": 0.8053,
+      "step": 8657
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2329784864859345e-06,
+      "loss": 0.9304,
+      "step": 8658
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.232355758687237e-06,
+      "loss": 0.9538,
+      "step": 8659
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.231732987637153e-06,
+      "loss": 0.9016,
+      "step": 8660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.231110173347748e-06,
+      "loss": 0.8322,
+      "step": 8661
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2304873158310896e-06,
+      "loss": 0.8089,
+      "step": 8662
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.229864415099246e-06,
+      "loss": 1.202,
+      "step": 8663
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.229241471164284e-06,
+      "loss": 0.7971,
+      "step": 8664
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.228618484038276e-06,
+      "loss": 0.9671,
+      "step": 8665
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.227995453733289e-06,
+      "loss": 0.9697,
+      "step": 8666
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.227372380261396e-06,
+      "loss": 0.9797,
+      "step": 8667
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2267492636346685e-06,
+      "loss": 1.0085,
+      "step": 8668
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.226126103865178e-06,
+      "loss": 0.9695,
+      "step": 8669
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2255029009650005e-06,
+      "loss": 0.9129,
+      "step": 8670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.224879654946209e-06,
+      "loss": 1.0031,
+      "step": 8671
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.224256365820878e-06,
+      "loss": 0.9807,
+      "step": 8672
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2236330336010855e-06,
+      "loss": 0.8437,
+      "step": 8673
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.223009658298906e-06,
+      "loss": 0.9187,
+      "step": 8674
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.222386239926419e-06,
+      "loss": 0.9562,
+      "step": 8675
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2217627784957035e-06,
+      "loss": 0.8359,
+      "step": 8676
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.221139274018838e-06,
+      "loss": 0.9569,
+      "step": 8677
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.220515726507903e-06,
+      "loss": 1.0924,
+      "step": 8678
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.21989213597498e-06,
+      "loss": 0.9645,
+      "step": 8679
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2192685024321485e-06,
+      "loss": 0.8608,
+      "step": 8680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2186448258914955e-06,
+      "loss": 0.9752,
+      "step": 8681
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2180211063651e-06,
+      "loss": 0.9013,
+      "step": 8682
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.217397343865051e-06,
+      "loss": 0.9034,
+      "step": 8683
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2167735384034295e-06,
+      "loss": 0.9055,
+      "step": 8684
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.216149689992326e-06,
+      "loss": 1.0763,
+      "step": 8685
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.215525798643823e-06,
+      "loss": 1.1136,
+      "step": 8686
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2149018643700095e-06,
+      "loss": 0.9086,
+      "step": 8687
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.214277887182976e-06,
+      "loss": 0.9396,
+      "step": 8688
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.213653867094811e-06,
+      "loss": 0.8823,
+      "step": 8689
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.213029804117604e-06,
+      "loss": 1.0466,
+      "step": 8690
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.212405698263446e-06,
+      "loss": 1.1119,
+      "step": 8691
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.21178154954443e-06,
+      "loss": 1.0146,
+      "step": 8692
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.211157357972648e-06,
+      "loss": 1.0357,
+      "step": 8693
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2105331235601935e-06,
+      "loss": 0.8918,
+      "step": 8694
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.209908846319162e-06,
+      "loss": 0.6939,
+      "step": 8695
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.209284526261648e-06,
+      "loss": 1.0259,
+      "step": 8696
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.208660163399747e-06,
+      "loss": 1.0044,
+      "step": 8697
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.208035757745555e-06,
+      "loss": 0.9743,
+      "step": 8698
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.207411309311173e-06,
+      "loss": 0.9034,
+      "step": 8699
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.206786818108698e-06,
+      "loss": 0.9734,
+      "step": 8700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.206162284150227e-06,
+      "loss": 1.0606,
+      "step": 8701
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.205537707447864e-06,
+      "loss": 1.0821,
+      "step": 8702
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.204913088013708e-06,
+      "loss": 0.9171,
+      "step": 8703
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.204288425859859e-06,
+      "loss": 0.8808,
+      "step": 8704
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.203663720998423e-06,
+      "loss": 0.9619,
+      "step": 8705
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.203038973441503e-06,
+      "loss": 1.0201,
+      "step": 8706
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.202414183201202e-06,
+      "loss": 0.9305,
+      "step": 8707
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.201789350289625e-06,
+      "loss": 0.9693,
+      "step": 8708
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2011644747188805e-06,
+      "loss": 1.0141,
+      "step": 8709
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.200539556501072e-06,
+      "loss": 0.8386,
+      "step": 8710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.1999145956483105e-06,
+      "loss": 1.081,
+      "step": 8711
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.199289592172703e-06,
+      "loss": 1.0613,
+      "step": 8712
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.198664546086357e-06,
+      "loss": 0.8934,
+      "step": 8713
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.198039457401384e-06,
+      "loss": 0.7775,
+      "step": 8714
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.1974143261298955e-06,
+      "loss": 1.0198,
+      "step": 8715
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.196789152284002e-06,
+      "loss": 0.8647,
+      "step": 8716
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.1961639358758185e-06,
+      "loss": 0.9801,
+      "step": 8717
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.195538676917457e-06,
+      "loss": 1.0668,
+      "step": 8718
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.194913375421031e-06,
+      "loss": 1.0454,
+      "step": 8719
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.194288031398656e-06,
+      "loss": 0.9009,
+      "step": 8720
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.193662644862449e-06,
+      "loss": 1.1406,
+      "step": 8721
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.193037215824524e-06,
+      "loss": 0.8818,
+      "step": 8722
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.192411744297002e-06,
+      "loss": 1.018,
+      "step": 8723
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.191786230291999e-06,
+      "loss": 1.0455,
+      "step": 8724
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.191160673821637e-06,
+      "loss": 0.9404,
+      "step": 8725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.190535074898032e-06,
+      "loss": 1.049,
+      "step": 8726
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1899094335333065e-06,
+      "loss": 0.9627,
+      "step": 8727
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.189283749739584e-06,
+      "loss": 1.0717,
+      "step": 8728
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.188658023528985e-06,
+      "loss": 1.0477,
+      "step": 8729
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1880322549136325e-06,
+      "loss": 1.0308,
+      "step": 8730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.187406443905652e-06,
+      "loss": 0.8808,
+      "step": 8731
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1867805905171675e-06,
+      "loss": 0.8959,
+      "step": 8732
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1861546947603045e-06,
+      "loss": 0.9972,
+      "step": 8733
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1855287566471905e-06,
+      "loss": 1.0589,
+      "step": 8734
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.184902776189953e-06,
+      "loss": 0.9813,
+      "step": 8735
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.184276753400721e-06,
+      "loss": 1.0711,
+      "step": 8736
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.183650688291621e-06,
+      "loss": 1.0044,
+      "step": 8737
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.183024580874785e-06,
+      "loss": 1.1295,
+      "step": 8738
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.182398431162343e-06,
+      "loss": 1.0098,
+      "step": 8739
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.181772239166426e-06,
+      "loss": 0.8752,
+      "step": 8740
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.181146004899166e-06,
+      "loss": 0.885,
+      "step": 8741
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.180519728372699e-06,
+      "loss": 0.9902,
+      "step": 8742
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1798934095991545e-06,
+      "loss": 1.0193,
+      "step": 8743
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1792670485906716e-06,
+      "loss": 0.9726,
+      "step": 8744
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.178640645359382e-06,
+      "loss": 0.8715,
+      "step": 8745
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.178014199917425e-06,
+      "loss": 1.0678,
+      "step": 8746
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.177387712276938e-06,
+      "loss": 1.0106,
+      "step": 8747
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.176761182450056e-06,
+      "loss": 1.0824,
+      "step": 8748
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.176134610448921e-06,
+      "loss": 1.0869,
+      "step": 8749
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.175507996285672e-06,
+      "loss": 0.8231,
+      "step": 8750
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.174881339972448e-06,
+      "loss": 0.9781,
+      "step": 8751
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.174254641521392e-06,
+      "loss": 0.9415,
+      "step": 8752
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.173627900944646e-06,
+      "loss": 1.0223,
+      "step": 8753
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.173001118254353e-06,
+      "loss": 0.8763,
+      "step": 8754
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.172374293462654e-06,
+      "loss": 0.983,
+      "step": 8755
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.171747426581697e-06,
+      "loss": 0.8728,
+      "step": 8756
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.171120517623625e-06,
+      "loss": 0.9536,
+      "step": 8757
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.170493566600587e-06,
+      "loss": 0.7878,
+      "step": 8758
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.169866573524729e-06,
+      "loss": 1.0003,
+      "step": 8759
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.169239538408198e-06,
+      "loss": 0.9482,
+      "step": 8760
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.168612461263142e-06,
+      "loss": 0.9989,
+      "step": 8761
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.167985342101711e-06,
+      "loss": 0.8388,
+      "step": 8762
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1673581809360575e-06,
+      "loss": 1.0332,
+      "step": 8763
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.166730977778331e-06,
+      "loss": 0.9954,
+      "step": 8764
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1661037326406825e-06,
+      "loss": 0.8319,
+      "step": 8765
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.165476445535266e-06,
+      "loss": 1.0183,
+      "step": 8766
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.164849116474234e-06,
+      "loss": 1.0301,
+      "step": 8767
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.164221745469742e-06,
+      "loss": 0.9465,
+      "step": 8768
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.163594332533944e-06,
+      "loss": 0.9241,
+      "step": 8769
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.162966877678996e-06,
+      "loss": 1.136,
+      "step": 8770
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1623393809170565e-06,
+      "loss": 0.9095,
+      "step": 8771
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.161711842260281e-06,
+      "loss": 0.8874,
+      "step": 8772
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.16108426172083e-06,
+      "loss": 0.845,
+      "step": 8773
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.160456639310862e-06,
+      "loss": 0.8525,
+      "step": 8774
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.159828975042536e-06,
+      "loss": 0.9306,
+      "step": 8775
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1592012689280135e-06,
+      "loss": 0.8887,
+      "step": 8776
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.158573520979456e-06,
+      "loss": 0.8853,
+      "step": 8777
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.157945731209026e-06,
+      "loss": 0.9671,
+      "step": 8778
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.157317899628888e-06,
+      "loss": 0.7614,
+      "step": 8779
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1566900262512025e-06,
+      "loss": 0.9857,
+      "step": 8780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.15606211108814e-06,
+      "loss": 0.8331,
+      "step": 8781
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.155434154151861e-06,
+      "loss": 0.9493,
+      "step": 8782
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.154806155454537e-06,
+      "loss": 1.1412,
+      "step": 8783
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.15417811500833e-06,
+      "loss": 0.9255,
+      "step": 8784
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.153550032825411e-06,
+      "loss": 0.9416,
+      "step": 8785
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.152921908917949e-06,
+      "loss": 0.8676,
+      "step": 8786
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.152293743298113e-06,
+      "loss": 1.1736,
+      "step": 8787
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.151665535978075e-06,
+      "loss": 1.0195,
+      "step": 8788
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.151037286970004e-06,
+      "loss": 0.8055,
+      "step": 8789
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1504089962860746e-06,
+      "loss": 0.8561,
+      "step": 8790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.149780663938457e-06,
+      "loss": 0.929,
+      "step": 8791
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.149152289939328e-06,
+      "loss": 0.9798,
+      "step": 8792
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.148523874300861e-06,
+      "loss": 0.8462,
+      "step": 8793
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.147895417035232e-06,
+      "loss": 1.0054,
+      "step": 8794
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.147266918154616e-06,
+      "loss": 0.8716,
+      "step": 8795
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.14663837767119e-06,
+      "loss": 1.0422,
+      "step": 8796
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.146009795597132e-06,
+      "loss": 0.7339,
+      "step": 8797
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.145381171944621e-06,
+      "loss": 0.8872,
+      "step": 8798
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.144752506725839e-06,
+      "loss": 0.9436,
+      "step": 8799
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.14412379995296e-06,
+      "loss": 0.9401,
+      "step": 8800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.143495051638172e-06,
+      "loss": 1.2306,
+      "step": 8801
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.142866261793651e-06,
+      "loss": 0.9286,
+      "step": 8802
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.142237430431584e-06,
+      "loss": 1.0664,
+      "step": 8803
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.141608557564153e-06,
+      "loss": 0.9826,
+      "step": 8804
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.14097964320354e-06,
+      "loss": 0.9518,
+      "step": 8805
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.140350687361934e-06,
+      "loss": 1.1159,
+      "step": 8806
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1397216900515175e-06,
+      "loss": 0.9207,
+      "step": 8807
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.139092651284479e-06,
+      "loss": 0.8904,
+      "step": 8808
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.138463571073005e-06,
+      "loss": 0.8727,
+      "step": 8809
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.137834449429286e-06,
+      "loss": 0.8821,
+      "step": 8810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.137205286365508e-06,
+      "loss": 0.9483,
+      "step": 8811
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.136576081893863e-06,
+      "loss": 0.9367,
+      "step": 8812
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1359468360265416e-06,
+      "loss": 1.0954,
+      "step": 8813
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1353175487757355e-06,
+      "loss": 1.1657,
+      "step": 8814
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.134688220153634e-06,
+      "loss": 1.0002,
+      "step": 8815
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.134058850172433e-06,
+      "loss": 0.7091,
+      "step": 8816
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.133429438844327e-06,
+      "loss": 0.9063,
+      "step": 8817
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.132799986181509e-06,
+      "loss": 0.9587,
+      "step": 8818
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1321704921961756e-06,
+      "loss": 1.1794,
+      "step": 8819
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.13154095690052e-06,
+      "loss": 0.9841,
+      "step": 8820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1309113803067445e-06,
+      "loss": 1.0839,
+      "step": 8821
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.130281762427043e-06,
+      "loss": 1.0268,
+      "step": 8822
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.129652103273616e-06,
+      "loss": 1.0486,
+      "step": 8823
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.129022402858664e-06,
+      "loss": 0.8619,
+      "step": 8824
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.128392661194385e-06,
+      "loss": 0.9055,
+      "step": 8825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.127762878292979e-06,
+      "loss": 0.9622,
+      "step": 8826
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.127133054166653e-06,
+      "loss": 0.9695,
+      "step": 8827
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.126503188827605e-06,
+      "loss": 0.938,
+      "step": 8828
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1258732822880404e-06,
+      "loss": 1.1692,
+      "step": 8829
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.125243334560164e-06,
+      "loss": 0.9295,
+      "step": 8830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1246133456561794e-06,
+      "loss": 0.881,
+      "step": 8831
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.123983315588292e-06,
+      "loss": 0.772,
+      "step": 8832
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.123353244368711e-06,
+      "loss": 0.9348,
+      "step": 8833
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.122723132009643e-06,
+      "loss": 0.811,
+      "step": 8834
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.122092978523295e-06,
+      "loss": 0.9898,
+      "step": 8835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.121462783921877e-06,
+      "loss": 0.9907,
+      "step": 8836
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.120832548217599e-06,
+      "loss": 0.9036,
+      "step": 8837
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.120202271422671e-06,
+      "loss": 1.0507,
+      "step": 8838
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.119571953549305e-06,
+      "loss": 0.9878,
+      "step": 8839
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1189415946097126e-06,
+      "loss": 1.039,
+      "step": 8840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.118311194616109e-06,
+      "loss": 1.1019,
+      "step": 8841
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.117680753580705e-06,
+      "loss": 1.0117,
+      "step": 8842
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.117050271515717e-06,
+      "loss": 0.8424,
+      "step": 8843
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.116419748433361e-06,
+      "loss": 1.0049,
+      "step": 8844
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1157891843458515e-06,
+      "loss": 0.9457,
+      "step": 8845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1151585792654075e-06,
+      "loss": 0.8461,
+      "step": 8846
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.114527933204246e-06,
+      "loss": 0.9394,
+      "step": 8847
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.113897246174585e-06,
+      "loss": 0.7345,
+      "step": 8848
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.113266518188643e-06,
+      "loss": 0.9212,
+      "step": 8849
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.112635749258643e-06,
+      "loss": 1.005,
+      "step": 8850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1120049393968045e-06,
+      "loss": 0.7737,
+      "step": 8851
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.11137408861535e-06,
+      "loss": 0.9297,
+      "step": 8852
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.110743196926501e-06,
+      "loss": 0.9713,
+      "step": 8853
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.110112264342481e-06,
+      "loss": 0.8251,
+      "step": 8854
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.109481290875514e-06,
+      "loss": 0.943,
+      "step": 8855
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.108850276537826e-06,
+      "loss": 1.1493,
+      "step": 8856
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.108219221341642e-06,
+      "loss": 0.8201,
+      "step": 8857
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.107588125299189e-06,
+      "loss": 0.8222,
+      "step": 8858
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.106956988422695e-06,
+      "loss": 1.0952,
+      "step": 8859
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.106325810724387e-06,
+      "loss": 0.9446,
+      "step": 8860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.105694592216494e-06,
+      "loss": 0.8456,
+      "step": 8861
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.105063332911246e-06,
+      "loss": 0.817,
+      "step": 8862
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1044320328208734e-06,
+      "loss": 0.9757,
+      "step": 8863
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.103800691957608e-06,
+      "loss": 0.9624,
+      "step": 8864
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.103169310333681e-06,
+      "loss": 1.0366,
+      "step": 8865
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.102537887961325e-06,
+      "loss": 0.9267,
+      "step": 8866
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.101906424852774e-06,
+      "loss": 0.9052,
+      "step": 8867
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1012749210202645e-06,
+      "loss": 1.0285,
+      "step": 8868
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.100643376476028e-06,
+      "loss": 0.8311,
+      "step": 8869
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1000117912323054e-06,
+      "loss": 0.8378,
+      "step": 8870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.099380165301329e-06,
+      "loss": 0.8446,
+      "step": 8871
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0987484986953385e-06,
+      "loss": 0.9915,
+      "step": 8872
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.09811679142657e-06,
+      "loss": 0.8805,
+      "step": 8873
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.097485043507267e-06,
+      "loss": 1.0212,
+      "step": 8874
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.096853254949666e-06,
+      "loss": 0.9238,
+      "step": 8875
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.09622142576601e-06,
+      "loss": 0.9045,
+      "step": 8876
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.095589555968539e-06,
+      "loss": 0.9592,
+      "step": 8877
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.094957645569494e-06,
+      "loss": 0.789,
+      "step": 8878
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.094325694581122e-06,
+      "loss": 1.0304,
+      "step": 8879
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0936937030156626e-06,
+      "loss": 0.9504,
+      "step": 8880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.093061670885364e-06,
+      "loss": 0.9705,
+      "step": 8881
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.092429598202471e-06,
+      "loss": 1.1086,
+      "step": 8882
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.091797484979228e-06,
+      "loss": 1.0218,
+      "step": 8883
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.091165331227883e-06,
+      "loss": 1.003,
+      "step": 8884
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0905331369606845e-06,
+      "loss": 0.885,
+      "step": 8885
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.089900902189882e-06,
+      "loss": 0.8736,
+      "step": 8886
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.089268626927721e-06,
+      "loss": 0.9493,
+      "step": 8887
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.088636311186457e-06,
+      "loss": 0.91,
+      "step": 8888
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.088003954978336e-06,
+      "loss": 1.0085,
+      "step": 8889
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.087371558315613e-06,
+      "loss": 0.9949,
+      "step": 8890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.086739121210539e-06,
+      "loss": 0.9117,
+      "step": 8891
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.086106643675368e-06,
+      "loss": 0.8567,
+      "step": 8892
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.085474125722355e-06,
+      "loss": 1.1559,
+      "step": 8893
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.084841567363754e-06,
+      "loss": 1.1206,
+      "step": 8894
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.084208968611819e-06,
+      "loss": 0.8956,
+      "step": 8895
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.083576329478809e-06,
+      "loss": 0.9246,
+      "step": 8896
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.08294364997698e-06,
+      "loss": 0.973,
+      "step": 8897
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.08231093011859e-06,
+      "loss": 0.7927,
+      "step": 8898
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.081678169915899e-06,
+      "loss": 0.8055,
+      "step": 8899
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.081045369381165e-06,
+      "loss": 1.0149,
+      "step": 8900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.08041252852665e-06,
+      "loss": 0.8082,
+      "step": 8901
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.079779647364612e-06,
+      "loss": 0.9036,
+      "step": 8902
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.079146725907317e-06,
+      "loss": 0.8604,
+      "step": 8903
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.078513764167025e-06,
+      "loss": 0.8956,
+      "step": 8904
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.077880762156e-06,
+      "loss": 0.9263,
+      "step": 8905
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.077247719886508e-06,
+      "loss": 1.0135,
+      "step": 8906
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.076614637370811e-06,
+      "loss": 0.9702,
+      "step": 8907
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.075981514621176e-06,
+      "loss": 0.9192,
+      "step": 8908
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.075348351649872e-06,
+      "loss": 0.8403,
+      "step": 8909
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.074715148469164e-06,
+      "loss": 0.972,
+      "step": 8910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.07408190509132e-06,
+      "loss": 0.9725,
+      "step": 8911
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.07344862152861e-06,
+      "loss": 0.8659,
+      "step": 8912
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.072815297793303e-06,
+      "loss": 0.9077,
+      "step": 8913
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.07218193389767e-06,
+      "loss": 0.9149,
+      "step": 8914
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.071548529853982e-06,
+      "loss": 0.9743,
+      "step": 8915
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.070915085674511e-06,
+      "loss": 1.0771,
+      "step": 8916
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.07028160137153e-06,
+      "loss": 0.9573,
+      "step": 8917
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.069648076957313e-06,
+      "loss": 0.8785,
+      "step": 8918
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.069014512444132e-06,
+      "loss": 0.8376,
+      "step": 8919
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.068380907844267e-06,
+      "loss": 1.042,
+      "step": 8920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.06774726316999e-06,
+      "loss": 1.0807,
+      "step": 8921
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.067113578433579e-06,
+      "loss": 0.9762,
+      "step": 8922
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0664798536473116e-06,
+      "loss": 0.7782,
+      "step": 8923
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.065846088823465e-06,
+      "loss": 0.9356,
+      "step": 8924
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0652122839743195e-06,
+      "loss": 0.8701,
+      "step": 8925
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.064578439112154e-06,
+      "loss": 0.8531,
+      "step": 8926
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.06394455424925e-06,
+      "loss": 0.9383,
+      "step": 8927
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.06331062939789e-06,
+      "loss": 0.8417,
+      "step": 8928
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0626766645703515e-06,
+      "loss": 0.9642,
+      "step": 8929
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.062042659778922e-06,
+      "loss": 1.0601,
+      "step": 8930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.061408615035883e-06,
+      "loss": 0.9359,
+      "step": 8931
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.06077453035352e-06,
+      "loss": 0.9297,
+      "step": 8932
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.06014040574412e-06,
+      "loss": 1.0401,
+      "step": 8933
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.059506241219964e-06,
+      "loss": 0.8136,
+      "step": 8934
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.058872036793344e-06,
+      "loss": 0.8754,
+      "step": 8935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0582377924765445e-06,
+      "loss": 0.9371,
+      "step": 8936
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.057603508281853e-06,
+      "loss": 1.0042,
+      "step": 8937
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.056969184221561e-06,
+      "loss": 1.0209,
+      "step": 8938
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.056334820307959e-06,
+      "loss": 0.8058,
+      "step": 8939
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.055700416553336e-06,
+      "loss": 0.8932,
+      "step": 8940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.055065972969983e-06,
+      "loss": 0.9309,
+      "step": 8941
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.054431489570193e-06,
+      "loss": 0.9096,
+      "step": 8942
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.05379696636626e-06,
+      "loss": 1.0219,
+      "step": 8943
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.053162403370476e-06,
+      "loss": 0.756,
+      "step": 8944
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.052527800595136e-06,
+      "loss": 0.9824,
+      "step": 8945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0518931580525365e-06,
+      "loss": 0.7824,
+      "step": 8946
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.051258475754971e-06,
+      "loss": 0.9114,
+      "step": 8947
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.050623753714737e-06,
+      "loss": 0.9933,
+      "step": 8948
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.049988991944134e-06,
+      "loss": 0.9375,
+      "step": 8949
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0493541904554605e-06,
+      "loss": 1.1241,
+      "step": 8950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.048719349261014e-06,
+      "loss": 0.9985,
+      "step": 8951
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.048084468373095e-06,
+      "loss": 0.8225,
+      "step": 8952
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.047449547804003e-06,
+      "loss": 1.132,
+      "step": 8953
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.046814587566041e-06,
+      "loss": 0.9166,
+      "step": 8954
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.046179587671511e-06,
+      "loss": 0.9566,
+      "step": 8955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.045544548132715e-06,
+      "loss": 1.1084,
+      "step": 8956
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.044909468961958e-06,
+      "loss": 0.9156,
+      "step": 8957
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.044274350171543e-06,
+      "loss": 1.0281,
+      "step": 8958
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.043639191773777e-06,
+      "loss": 1.048,
+      "step": 8959
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.043003993780963e-06,
+      "loss": 0.9499,
+      "step": 8960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.042368756205413e-06,
+      "loss": 0.9799,
+      "step": 8961
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.04173347905943e-06,
+      "loss": 0.8004,
+      "step": 8962
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.041098162355325e-06,
+      "loss": 0.9822,
+      "step": 8963
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.040462806105406e-06,
+      "loss": 1.1246,
+      "step": 8964
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.039827410321982e-06,
+      "loss": 1.0654,
+      "step": 8965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0391919750173655e-06,
+      "loss": 0.9419,
+      "step": 8966
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.038556500203867e-06,
+      "loss": 0.9663,
+      "step": 8967
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.037920985893798e-06,
+      "loss": 0.998,
+      "step": 8968
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.037285432099473e-06,
+      "loss": 0.8821,
+      "step": 8969
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0366498388332045e-06,
+      "loss": 1.1349,
+      "step": 8970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0360142061073065e-06,
+      "loss": 0.9355,
+      "step": 8971
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0353785339340965e-06,
+      "loss": 0.9009,
+      "step": 8972
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.034742822325889e-06,
+      "loss": 0.9824,
+      "step": 8973
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.034107071295e-06,
+      "loss": 0.8917,
+      "step": 8974
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.03347128085375e-06,
+      "loss": 1.0655,
+      "step": 8975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.032835451014453e-06,
+      "loss": 0.9239,
+      "step": 8976
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.03219958178943e-06,
+      "loss": 0.9499,
+      "step": 8977
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.031563673191002e-06,
+      "loss": 0.9331,
+      "step": 8978
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.030927725231489e-06,
+      "loss": 0.8712,
+      "step": 8979
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.030291737923212e-06,
+      "loss": 1.0244,
+      "step": 8980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.029655711278492e-06,
+      "loss": 0.8838,
+      "step": 8981
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.029019645309655e-06,
+      "loss": 0.9153,
+      "step": 8982
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0283835400290205e-06,
+      "loss": 1.0015,
+      "step": 8983
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.027747395448916e-06,
+      "loss": 0.9107,
+      "step": 8984
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.027111211581665e-06,
+      "loss": 0.8855,
+      "step": 8985
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.026474988439596e-06,
+      "loss": 0.8775,
+      "step": 8986
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.025838726035032e-06,
+      "loss": 0.9008,
+      "step": 8987
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.025202424380303e-06,
+      "loss": 0.8956,
+      "step": 8988
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.024566083487736e-06,
+      "loss": 0.9023,
+      "step": 8989
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.023929703369661e-06,
+      "loss": 0.8864,
+      "step": 8990
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.023293284038406e-06,
+      "loss": 1.045,
+      "step": 8991
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.022656825506305e-06,
+      "loss": 0.983,
+      "step": 8992
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.022020327785684e-06,
+      "loss": 0.9527,
+      "step": 8993
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.021383790888879e-06,
+      "loss": 1.0602,
+      "step": 8994
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.020747214828221e-06,
+      "loss": 0.9525,
+      "step": 8995
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0201105996160445e-06,
+      "loss": 0.8873,
+      "step": 8996
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0194739452646835e-06,
+      "loss": 0.7391,
+      "step": 8997
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.018837251786474e-06,
+      "loss": 1.0821,
+      "step": 8998
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0182005191937485e-06,
+      "loss": 0.9561,
+      "step": 8999
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.017563747498847e-06,
+      "loss": 0.8888,
+      "step": 9000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.016926936714105e-06,
+      "loss": 0.9324,
+      "step": 9001
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.016290086851861e-06,
+      "loss": 1.0938,
+      "step": 9002
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.015653197924454e-06,
+      "loss": 0.9243,
+      "step": 9003
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.015016269944225e-06,
+      "loss": 0.8402,
+      "step": 9004
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0143793029235105e-06,
+      "loss": 1.0703,
+      "step": 9005
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.013742296874655e-06,
+      "loss": 0.7458,
+      "step": 9006
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.013105251809999e-06,
+      "loss": 1.0296,
+      "step": 9007
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.012468167741887e-06,
+      "loss": 1.0388,
+      "step": 9008
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.01183104468266e-06,
+      "loss": 1.0155,
+      "step": 9009
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.011193882644663e-06,
+      "loss": 0.9446,
+      "step": 9010
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.010556681640239e-06,
+      "loss": 0.8427,
+      "step": 9011
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.009919441681736e-06,
+      "loss": 0.9024,
+      "step": 9012
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.009282162781501e-06,
+      "loss": 0.934,
+      "step": 9013
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.008644844951879e-06,
+      "loss": 1.058,
+      "step": 9014
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.00800748820522e-06,
+      "loss": 0.9326,
+      "step": 9015
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.007370092553869e-06,
+      "loss": 0.9636,
+      "step": 9016
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.00673265801018e-06,
+      "loss": 0.8811,
+      "step": 9017
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0060951845865e-06,
+      "loss": 0.9666,
+      "step": 9018
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0054576722951815e-06,
+      "loss": 0.9343,
+      "step": 9019
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0048201211485745e-06,
+      "loss": 1.0464,
+      "step": 9020
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.004182531159034e-06,
+      "loss": 1.1237,
+      "step": 9021
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.00354490233891e-06,
+      "loss": 0.9465,
+      "step": 9022
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.002907234700559e-06,
+      "loss": 0.7005,
+      "step": 9023
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.002269528256334e-06,
+      "loss": 1.159,
+      "step": 9024
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.00163178301859e-06,
+      "loss": 0.9726,
+      "step": 9025
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.000993998999685e-06,
+      "loss": 1.0811,
+      "step": 9026
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.000356176211978e-06,
+      "loss": 0.9288,
+      "step": 9027
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9997183146678205e-06,
+      "loss": 0.7952,
+      "step": 9028
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.999080414379576e-06,
+      "loss": 0.9735,
+      "step": 9029
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.998442475359599e-06,
+      "loss": 0.9927,
+      "step": 9030
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.997804497620255e-06,
+      "loss": 0.832,
+      "step": 9031
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.997166481173903e-06,
+      "loss": 0.9378,
+      "step": 9032
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9965284260329016e-06,
+      "loss": 0.8615,
+      "step": 9033
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.995890332209615e-06,
+      "loss": 0.9197,
+      "step": 9034
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.995252199716406e-06,
+      "loss": 0.918,
+      "step": 9035
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.994614028565638e-06,
+      "loss": 0.994,
+      "step": 9036
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.993975818769676e-06,
+      "loss": 0.9155,
+      "step": 9037
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.993337570340886e-06,
+      "loss": 0.8581,
+      "step": 9038
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.992699283291632e-06,
+      "loss": 0.9326,
+      "step": 9039
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.992060957634281e-06,
+      "loss": 1.0071,
+      "step": 9040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.991422593381199e-06,
+      "loss": 0.9333,
+      "step": 9041
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.99078419054476e-06,
+      "loss": 0.9327,
+      "step": 9042
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.990145749137326e-06,
+      "loss": 0.9309,
+      "step": 9043
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.989507269171271e-06,
+      "loss": 1.0745,
+      "step": 9044
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9888687506589625e-06,
+      "loss": 0.9089,
+      "step": 9045
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.988230193612775e-06,
+      "loss": 0.8909,
+      "step": 9046
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9875915980450774e-06,
+      "loss": 0.807,
+      "step": 9047
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.986952963968243e-06,
+      "loss": 0.9759,
+      "step": 9048
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.986314291394647e-06,
+      "loss": 1.0361,
+      "step": 9049
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9856755803366615e-06,
+      "loss": 0.949,
+      "step": 9050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.985036830806662e-06,
+      "loss": 0.9353,
+      "step": 9051
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.984398042817023e-06,
+      "loss": 0.9188,
+      "step": 9052
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.983759216380122e-06,
+      "loss": 1.0258,
+      "step": 9053
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.983120351508338e-06,
+      "loss": 0.9152,
+      "step": 9054
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.982481448214047e-06,
+      "loss": 0.9665,
+      "step": 9055
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.981842506509626e-06,
+      "loss": 0.9106,
+      "step": 9056
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.981203526407458e-06,
+      "loss": 0.9001,
+      "step": 9057
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.98056450791992e-06,
+      "loss": 0.981,
+      "step": 9058
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.979925451059393e-06,
+      "loss": 0.9147,
+      "step": 9059
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.979286355838261e-06,
+      "loss": 0.8395,
+      "step": 9060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.978647222268904e-06,
+      "loss": 0.9074,
+      "step": 9061
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.978008050363707e-06,
+      "loss": 0.8791,
+      "step": 9062
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.977368840135052e-06,
+      "loss": 0.9146,
+      "step": 9063
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.976729591595323e-06,
+      "loss": 0.9418,
+      "step": 9064
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.976090304756907e-06,
+      "loss": 0.9003,
+      "step": 9065
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.975450979632191e-06,
+      "loss": 1.0385,
+      "step": 9066
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.97481161623356e-06,
+      "loss": 0.9375,
+      "step": 9067
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.974172214573402e-06,
+      "loss": 0.9149,
+      "step": 9068
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.973532774664104e-06,
+      "loss": 0.7783,
+      "step": 9069
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.972893296518057e-06,
+      "loss": 0.8922,
+      "step": 9070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9722537801476495e-06,
+      "loss": 0.9568,
+      "step": 9071
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.971614225565273e-06,
+      "loss": 1.1278,
+      "step": 9072
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.970974632783318e-06,
+      "loss": 1.2002,
+      "step": 9073
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.970335001814177e-06,
+      "loss": 1.1326,
+      "step": 9074
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.969695332670241e-06,
+      "loss": 0.8256,
+      "step": 9075
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.969055625363907e-06,
+      "loss": 1.1186,
+      "step": 9076
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.968415879907564e-06,
+      "loss": 0.9797,
+      "step": 9077
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.967776096313612e-06,
+      "loss": 0.9786,
+      "step": 9078
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.967136274594444e-06,
+      "loss": 1.0443,
+      "step": 9079
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.966496414762456e-06,
+      "loss": 0.7374,
+      "step": 9080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9658565168300464e-06,
+      "loss": 0.964,
+      "step": 9081
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.965216580809613e-06,
+      "loss": 0.8924,
+      "step": 9082
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9645766067135535e-06,
+      "loss": 0.8943,
+      "step": 9083
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9639365945542695e-06,
+      "loss": 0.994,
+      "step": 9084
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.963296544344158e-06,
+      "loss": 0.9968,
+      "step": 9085
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.962656456095622e-06,
+      "loss": 0.9615,
+      "step": 9086
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.962016329821062e-06,
+      "loss": 0.9145,
+      "step": 9087
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.96137616553288e-06,
+      "loss": 0.9339,
+      "step": 9088
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9607359632434816e-06,
+      "loss": 1.1234,
+      "step": 9089
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.960095722965268e-06,
+      "loss": 0.9168,
+      "step": 9090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9594554447106435e-06,
+      "loss": 0.7707,
+      "step": 9091
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.958815128492014e-06,
+      "loss": 0.8648,
+      "step": 9092
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.958174774321787e-06,
+      "loss": 0.8452,
+      "step": 9093
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.957534382212367e-06,
+      "loss": 0.8002,
+      "step": 9094
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.956893952176163e-06,
+      "loss": 0.9992,
+      "step": 9095
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.956253484225581e-06,
+      "loss": 1.0034,
+      "step": 9096
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.955612978373034e-06,
+      "loss": 1.0002,
+      "step": 9097
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.954972434630928e-06,
+      "loss": 0.9461,
+      "step": 9098
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.954331853011673e-06,
+      "loss": 1.0643,
+      "step": 9099
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9536912335276845e-06,
+      "loss": 1.1265,
+      "step": 9100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.953050576191369e-06,
+      "loss": 0.894,
+      "step": 9101
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.952409881015143e-06,
+      "loss": 1.0944,
+      "step": 9102
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.951769148011419e-06,
+      "loss": 0.8872,
+      "step": 9103
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9511283771926095e-06,
+      "loss": 1.0343,
+      "step": 9104
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.95048756857113e-06,
+      "loss": 0.995,
+      "step": 9105
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.949846722159397e-06,
+      "loss": 0.9361,
+      "step": 9106
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.949205837969826e-06,
+      "loss": 1.1317,
+      "step": 9107
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.948564916014834e-06,
+      "loss": 0.9521,
+      "step": 9108
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.947923956306839e-06,
+      "loss": 1.1423,
+      "step": 9109
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.947282958858259e-06,
+      "loss": 0.8707,
+      "step": 9110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.946641923681513e-06,
+      "loss": 0.8292,
+      "step": 9111
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.946000850789022e-06,
+      "loss": 1.0266,
+      "step": 9112
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.945359740193206e-06,
+      "loss": 0.8712,
+      "step": 9113
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.944718591906485e-06,
+      "loss": 0.9242,
+      "step": 9114
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.944077405941284e-06,
+      "loss": 1.0738,
+      "step": 9115
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.943436182310024e-06,
+      "loss": 0.9351,
+      "step": 9116
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.942794921025127e-06,
+      "loss": 1.0095,
+      "step": 9117
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.942153622099021e-06,
+      "loss": 0.9613,
+      "step": 9118
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.941512285544129e-06,
+      "loss": 0.819,
+      "step": 9119
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9408709113728766e-06,
+      "loss": 1.067,
+      "step": 9120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.940229499597691e-06,
+      "loss": 0.9228,
+      "step": 9121
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.939588050230998e-06,
+      "loss": 0.9389,
+      "step": 9122
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.938946563285225e-06,
+      "loss": 0.8332,
+      "step": 9123
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.938305038772805e-06,
+      "loss": 0.8746,
+      "step": 9124
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9376634767061635e-06,
+      "loss": 0.8642,
+      "step": 9125
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.937021877097732e-06,
+      "loss": 0.8588,
+      "step": 9126
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.936380239959941e-06,
+      "loss": 0.9279,
+      "step": 9127
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.935738565305221e-06,
+      "loss": 0.9489,
+      "step": 9128
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.935096853146006e-06,
+      "loss": 0.9977,
+      "step": 9129
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.934455103494728e-06,
+      "loss": 0.9111,
+      "step": 9130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.933813316363822e-06,
+      "loss": 0.8222,
+      "step": 9131
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.933171491765722e-06,
+      "loss": 0.9321,
+      "step": 9132
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.932529629712863e-06,
+      "loss": 1.0441,
+      "step": 9133
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.931887730217679e-06,
+      "loss": 1.02,
+      "step": 9134
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.93124579329261e-06,
+      "loss": 1.104,
+      "step": 9135
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.930603818950091e-06,
+      "loss": 1.0461,
+      "step": 9136
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.929961807202562e-06,
+      "loss": 0.929,
+      "step": 9137
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.929319758062461e-06,
+      "loss": 1.0678,
+      "step": 9138
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.928677671542226e-06,
+      "loss": 0.9451,
+      "step": 9139
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.928035547654299e-06,
+      "loss": 1.0719,
+      "step": 9140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.927393386411121e-06,
+      "loss": 1.0916,
+      "step": 9141
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.926751187825133e-06,
+      "loss": 0.9341,
+      "step": 9142
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.926108951908779e-06,
+      "loss": 0.845,
+      "step": 9143
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9254666786745e-06,
+      "loss": 0.9332,
+      "step": 9144
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9248243681347405e-06,
+      "loss": 0.9606,
+      "step": 9145
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.924182020301944e-06,
+      "loss": 0.9353,
+      "step": 9146
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9235396351885594e-06,
+      "loss": 1.0603,
+      "step": 9147
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.922897212807031e-06,
+      "loss": 1.1122,
+      "step": 9148
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.922254753169805e-06,
+      "loss": 0.9538,
+      "step": 9149
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.921612256289328e-06,
+      "loss": 1.1912,
+      "step": 9150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.92096972217805e-06,
+      "loss": 0.994,
+      "step": 9151
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.92032715084842e-06,
+      "loss": 0.9593,
+      "step": 9152
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.919684542312886e-06,
+      "loss": 0.99,
+      "step": 9153
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9190418965839e-06,
+      "loss": 1.154,
+      "step": 9154
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.918399213673914e-06,
+      "loss": 0.9367,
+      "step": 9155
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.917756493595377e-06,
+      "loss": 0.8488,
+      "step": 9156
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.917113736360742e-06,
+      "loss": 0.861,
+      "step": 9157
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.916470941982465e-06,
+      "loss": 0.98,
+      "step": 9158
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.915828110472997e-06,
+      "loss": 0.9235,
+      "step": 9159
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9151852418447955e-06,
+      "loss": 1.0205,
+      "step": 9160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.914542336110312e-06,
+      "loss": 1.0319,
+      "step": 9161
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.913899393282007e-06,
+      "loss": 0.7953,
+      "step": 9162
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.913256413372333e-06,
+      "loss": 1.025,
+      "step": 9163
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.912613396393752e-06,
+      "loss": 0.8581,
+      "step": 9164
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.911970342358721e-06,
+      "loss": 1.0326,
+      "step": 9165
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.911327251279698e-06,
+      "loss": 0.9967,
+      "step": 9166
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.910684123169142e-06,
+      "loss": 0.8576,
+      "step": 9167
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.910040958039515e-06,
+      "loss": 1.0431,
+      "step": 9168
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9093977559032775e-06,
+      "loss": 0.8379,
+      "step": 9169
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.908754516772893e-06,
+      "loss": 0.8843,
+      "step": 9170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9081112406608205e-06,
+      "loss": 1.1725,
+      "step": 9171
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.907467927579528e-06,
+      "loss": 0.9087,
+      "step": 9172
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.906824577541475e-06,
+      "loss": 0.9448,
+      "step": 9173
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.906181190559129e-06,
+      "loss": 0.9623,
+      "step": 9174
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.905537766644954e-06,
+      "loss": 0.8863,
+      "step": 9175
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.904894305811418e-06,
+      "loss": 0.9251,
+      "step": 9176
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.904250808070987e-06,
+      "loss": 0.6848,
+      "step": 9177
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.903607273436128e-06,
+      "loss": 0.9335,
+      "step": 9178
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9029637019193104e-06,
+      "loss": 0.8209,
+      "step": 9179
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.902320093533004e-06,
+      "loss": 0.9475,
+      "step": 9180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.901676448289674e-06,
+      "loss": 0.9396,
+      "step": 9181
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9010327662017964e-06,
+      "loss": 0.9898,
+      "step": 9182
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.900389047281839e-06,
+      "loss": 1.0505,
+      "step": 9183
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.899745291542277e-06,
+      "loss": 1.0171,
+      "step": 9184
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.8991014989955795e-06,
+      "loss": 0.8771,
+      "step": 9185
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.898457669654221e-06,
+      "loss": 0.9749,
+      "step": 9186
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.897813803530675e-06,
+      "loss": 0.8743,
+      "step": 9187
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.89716990063742e-06,
+      "loss": 0.9994,
+      "step": 9188
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.8965259609869265e-06,
+      "loss": 1.1108,
+      "step": 9189
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.895881984591672e-06,
+      "loss": 0.9584,
+      "step": 9190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.8952379714641345e-06,
+      "loss": 1.0938,
+      "step": 9191
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.894593921616792e-06,
+      "loss": 1.034,
+      "step": 9192
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.893949835062121e-06,
+      "loss": 1.0737,
+      "step": 9193
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.893305711812603e-06,
+      "loss": 0.9013,
+      "step": 9194
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.892661551880716e-06,
+      "loss": 0.8926,
+      "step": 9195
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8920173552789404e-06,
+      "loss": 1.0613,
+      "step": 9196
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.891373122019757e-06,
+      "loss": 0.9722,
+      "step": 9197
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.890728852115649e-06,
+      "loss": 0.8929,
+      "step": 9198
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.890084545579099e-06,
+      "loss": 0.9683,
+      "step": 9199
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.889440202422589e-06,
+      "loss": 0.8289,
+      "step": 9200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.888795822658605e-06,
+      "loss": 1.0361,
+      "step": 9201
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.888151406299629e-06,
+      "loss": 1.1743,
+      "step": 9202
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.887506953358147e-06,
+      "loss": 0.8086,
+      "step": 9203
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.886862463846648e-06,
+      "loss": 0.9505,
+      "step": 9204
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8862179377776165e-06,
+      "loss": 0.9598,
+      "step": 9205
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.885573375163539e-06,
+      "loss": 1.0077,
+      "step": 9206
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.884928776016906e-06,
+      "loss": 0.9562,
+      "step": 9207
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.884284140350206e-06,
+      "loss": 0.9847,
+      "step": 9208
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.883639468175926e-06,
+      "loss": 0.8539,
+      "step": 9209
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.882994759506561e-06,
+      "loss": 0.8802,
+      "step": 9210
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.882350014354597e-06,
+      "loss": 1.1881,
+      "step": 9211
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.881705232732532e-06,
+      "loss": 0.7712,
+      "step": 9212
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.881060414652852e-06,
+      "loss": 0.9012,
+      "step": 9213
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.880415560128055e-06,
+      "loss": 1.146,
+      "step": 9214
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.879770669170632e-06,
+      "loss": 0.9872,
+      "step": 9215
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.879125741793078e-06,
+      "loss": 0.9587,
+      "step": 9216
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.87848077800789e-06,
+      "loss": 0.9063,
+      "step": 9217
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.877835777827564e-06,
+      "loss": 0.9303,
+      "step": 9218
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.877190741264595e-06,
+      "loss": 0.9861,
+      "step": 9219
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.87654566833148e-06,
+      "loss": 1.0677,
+      "step": 9220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.875900559040719e-06,
+      "loss": 0.7927,
+      "step": 9221
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8752554134048114e-06,
+      "loss": 0.7853,
+      "step": 9222
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.874610231436254e-06,
+      "loss": 1.0109,
+      "step": 9223
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8739650131475495e-06,
+      "loss": 0.8954,
+      "step": 9224
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.873319758551198e-06,
+      "loss": 1.0171,
+      "step": 9225
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8726744676597015e-06,
+      "loss": 0.9058,
+      "step": 9226
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.872029140485561e-06,
+      "loss": 1.0875,
+      "step": 9227
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.871383777041281e-06,
+      "loss": 0.9996,
+      "step": 9228
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8707383773393645e-06,
+      "loss": 1.0438,
+      "step": 9229
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.870092941392317e-06,
+      "loss": 1.0754,
+      "step": 9230
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8694474692126425e-06,
+      "loss": 0.9892,
+      "step": 9231
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.868801960812847e-06,
+      "loss": 0.8735,
+      "step": 9232
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8681564162054386e-06,
+      "loss": 1.1261,
+      "step": 9233
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.867510835402922e-06,
+      "loss": 1.0839,
+      "step": 9234
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8668652184178076e-06,
+      "loss": 0.8883,
+      "step": 9235
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.866219565262603e-06,
+      "loss": 0.9565,
+      "step": 9236
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.865573875949817e-06,
+      "loss": 0.9908,
+      "step": 9237
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.864928150491961e-06,
+      "loss": 0.9183,
+      "step": 9238
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.864282388901544e-06,
+      "loss": 0.8332,
+      "step": 9239
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.86363659119108e-06,
+      "loss": 0.7811,
+      "step": 9240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.862990757373078e-06,
+      "loss": 0.8807,
+      "step": 9241
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.862344887460055e-06,
+      "loss": 0.9533,
+      "step": 9242
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.861698981464521e-06,
+      "loss": 0.978,
+      "step": 9243
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.861053039398991e-06,
+      "loss": 0.9047,
+      "step": 9244
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.860407061275981e-06,
+      "loss": 0.9996,
+      "step": 9245
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.859761047108007e-06,
+      "loss": 0.8772,
+      "step": 9246
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.859114996907583e-06,
+      "loss": 0.8344,
+      "step": 9247
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.858468910687228e-06,
+      "loss": 0.9555,
+      "step": 9248
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.85782278845946e-06,
+      "loss": 0.97,
+      "step": 9249
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.857176630236797e-06,
+      "loss": 1.0448,
+      "step": 9250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.856530436031759e-06,
+      "loss": 0.9036,
+      "step": 9251
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.855884205856863e-06,
+      "loss": 0.9792,
+      "step": 9252
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.855237939724633e-06,
+      "loss": 1.0946,
+      "step": 9253
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.854591637647588e-06,
+      "loss": 1.1558,
+      "step": 9254
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.85394529963825e-06,
+      "loss": 0.8129,
+      "step": 9255
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.853298925709143e-06,
+      "loss": 0.9849,
+      "step": 9256
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.852652515872791e-06,
+      "loss": 1.1066,
+      "step": 9257
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8520060701417145e-06,
+      "loss": 0.9612,
+      "step": 9258
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.851359588528442e-06,
+      "loss": 0.8692,
+      "step": 9259
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8507130710454975e-06,
+      "loss": 0.9426,
+      "step": 9260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.850066517705407e-06,
+      "loss": 1.0548,
+      "step": 9261
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.849419928520698e-06,
+      "loss": 0.9833,
+      "step": 9262
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.848773303503895e-06,
+      "loss": 1.0828,
+      "step": 9263
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.848126642667531e-06,
+      "loss": 0.9523,
+      "step": 9264
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.847479946024131e-06,
+      "loss": 0.9463,
+      "step": 9265
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8468332135862285e-06,
+      "loss": 1.0393,
+      "step": 9266
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.846186445366348e-06,
+      "loss": 0.9375,
+      "step": 9267
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.845539641377026e-06,
+      "loss": 0.992,
+      "step": 9268
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.844892801630791e-06,
+      "loss": 1.0172,
+      "step": 9269
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.844245926140178e-06,
+      "loss": 1.0168,
+      "step": 9270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.843599014917717e-06,
+      "loss": 0.8747,
+      "step": 9271
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.842952067975944e-06,
+      "loss": 1.1409,
+      "step": 9272
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.842305085327393e-06,
+      "loss": 0.9487,
+      "step": 9273
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.841658066984598e-06,
+      "loss": 0.9759,
+      "step": 9274
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.841011012960096e-06,
+      "loss": 0.9155,
+      "step": 9275
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8403639232664235e-06,
+      "loss": 1.062,
+      "step": 9276
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8397167979161176e-06,
+      "loss": 0.978,
+      "step": 9277
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.839069636921716e-06,
+      "loss": 0.9778,
+      "step": 9278
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8384224402957565e-06,
+      "loss": 0.938,
+      "step": 9279
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8377752080507795e-06,
+      "loss": 0.9081,
+      "step": 9280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.837127940199327e-06,
+      "loss": 0.9733,
+      "step": 9281
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.836480636753935e-06,
+      "loss": 0.9707,
+      "step": 9282
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.835833297727148e-06,
+      "loss": 0.9026,
+      "step": 9283
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.835185923131506e-06,
+      "loss": 0.8632,
+      "step": 9284
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.834538512979554e-06,
+      "loss": 0.9252,
+      "step": 9285
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.833891067283835e-06,
+      "loss": 0.9039,
+      "step": 9286
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.833243586056892e-06,
+      "loss": 0.8218,
+      "step": 9287
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.832596069311271e-06,
+      "loss": 0.8318,
+      "step": 9288
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.831948517059515e-06,
+      "loss": 0.8495,
+      "step": 9289
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.831300929314175e-06,
+      "loss": 1.03,
+      "step": 9290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.830653306087794e-06,
+      "loss": 1.0607,
+      "step": 9291
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.830005647392919e-06,
+      "loss": 0.8116,
+      "step": 9292
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8293579532421e-06,
+      "loss": 0.861,
+      "step": 9293
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.828710223647885e-06,
+      "loss": 1.0478,
+      "step": 9294
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.828062458622826e-06,
+      "loss": 0.942,
+      "step": 9295
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.827414658179472e-06,
+      "loss": 0.9829,
+      "step": 9296
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.826766822330372e-06,
+      "loss": 0.9189,
+      "step": 9297
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.82611895108808e-06,
+      "loss": 0.8829,
+      "step": 9298
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.825471044465147e-06,
+      "loss": 1.2103,
+      "step": 9299
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.824823102474127e-06,
+      "loss": 0.8624,
+      "step": 9300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8241751251275734e-06,
+      "loss": 0.7688,
+      "step": 9301
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.82352711243804e-06,
+      "loss": 0.9326,
+      "step": 9302
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.822879064418082e-06,
+      "loss": 1.016,
+      "step": 9303
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8222309810802565e-06,
+      "loss": 0.9108,
+      "step": 9304
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.82158286243712e-06,
+      "loss": 0.9288,
+      "step": 9305
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.820934708501227e-06,
+      "loss": 1.0794,
+      "step": 9306
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.820286519285137e-06,
+      "loss": 0.9541,
+      "step": 9307
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.819638294801409e-06,
+      "loss": 1.0435,
+      "step": 9308
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.818990035062602e-06,
+      "loss": 1.0224,
+      "step": 9309
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.818341740081276e-06,
+      "loss": 0.9836,
+      "step": 9310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.817693409869991e-06,
+      "loss": 1.0226,
+      "step": 9311
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.817045044441308e-06,
+      "loss": 0.855,
+      "step": 9312
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.816396643807791e-06,
+      "loss": 1.0775,
+      "step": 9313
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.815748207981999e-06,
+      "loss": 1.3298,
+      "step": 9314
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.815099736976498e-06,
+      "loss": 0.8942,
+      "step": 9315
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.814451230803852e-06,
+      "loss": 0.8375,
+      "step": 9316
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.813802689476625e-06,
+      "loss": 0.9265,
+      "step": 9317
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.813154113007383e-06,
+      "loss": 0.9212,
+      "step": 9318
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.812505501408691e-06,
+      "loss": 0.9821,
+      "step": 9319
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.811856854693114e-06,
+      "loss": 0.9032,
+      "step": 9320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.811208172873223e-06,
+      "loss": 0.9969,
+      "step": 9321
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.810559455961585e-06,
+      "loss": 0.9357,
+      "step": 9322
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.809910703970768e-06,
+      "loss": 0.9086,
+      "step": 9323
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.809261916913341e-06,
+      "loss": 0.8074,
+      "step": 9324
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.808613094801875e-06,
+      "loss": 0.9463,
+      "step": 9325
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.80796423764894e-06,
+      "loss": 1.0026,
+      "step": 9326
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.807315345467109e-06,
+      "loss": 0.9463,
+      "step": 9327
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.806666418268954e-06,
+      "loss": 0.9871,
+      "step": 9328
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8060174560670456e-06,
+      "loss": 0.8687,
+      "step": 9329
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.805368458873959e-06,
+      "loss": 0.9502,
+      "step": 9330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.804719426702268e-06,
+      "loss": 0.7894,
+      "step": 9331
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.804070359564546e-06,
+      "loss": 0.9894,
+      "step": 9332
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8034212574733714e-06,
+      "loss": 0.8043,
+      "step": 9333
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8027721204413186e-06,
+      "loss": 0.8034,
+      "step": 9334
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.802122948480966e-06,
+      "loss": 0.8985,
+      "step": 9335
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8014737416048895e-06,
+      "loss": 0.9632,
+      "step": 9336
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.800824499825668e-06,
+      "loss": 1.0086,
+      "step": 9337
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.80017522315588e-06,
+      "loss": 0.9296,
+      "step": 9338
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.799525911608105e-06,
+      "loss": 0.8684,
+      "step": 9339
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.798876565194924e-06,
+      "loss": 0.812,
+      "step": 9340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.798227183928919e-06,
+      "loss": 0.8538,
+      "step": 9341
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7975777678226675e-06,
+      "loss": 0.9398,
+      "step": 9342
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.796928316888756e-06,
+      "loss": 1.0651,
+      "step": 9343
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.796278831139765e-06,
+      "loss": 0.9279,
+      "step": 9344
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.795629310588281e-06,
+      "loss": 0.9162,
+      "step": 9345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.794979755246885e-06,
+      "loss": 0.9658,
+      "step": 9346
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.794330165128165e-06,
+      "loss": 0.888,
+      "step": 9347
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7936805402447036e-06,
+      "loss": 0.8903,
+      "step": 9348
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.793030880609088e-06,
+      "loss": 0.9444,
+      "step": 9349
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.792381186233907e-06,
+      "loss": 0.927,
+      "step": 9350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.791731457131747e-06,
+      "loss": 0.9717,
+      "step": 9351
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.791081693315196e-06,
+      "loss": 1.1729,
+      "step": 9352
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.790431894796844e-06,
+      "loss": 0.702,
+      "step": 9353
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.789782061589281e-06,
+      "loss": 1.1471,
+      "step": 9354
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.789132193705095e-06,
+      "loss": 1.0246,
+      "step": 9355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7884822911568795e-06,
+      "loss": 1.0877,
+      "step": 9356
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.787832353957225e-06,
+      "loss": 1.1032,
+      "step": 9357
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.787182382118724e-06,
+      "loss": 1.1337,
+      "step": 9358
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.78653237565397e-06,
+      "loss": 0.9869,
+      "step": 9359
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7858823345755565e-06,
+      "loss": 0.7967,
+      "step": 9360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.785232258896078e-06,
+      "loss": 0.9065,
+      "step": 9361
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7845821486281286e-06,
+      "loss": 1.1167,
+      "step": 9362
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.783932003784305e-06,
+      "loss": 0.949,
+      "step": 9363
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.783281824377203e-06,
+      "loss": 0.9411,
+      "step": 9364
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.782631610419422e-06,
+      "loss": 0.9001,
+      "step": 9365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.781981361923557e-06,
+      "loss": 0.8742,
+      "step": 9366
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.781331078902207e-06,
+      "loss": 0.8498,
+      "step": 9367
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7806807613679716e-06,
+      "loss": 0.8484,
+      "step": 9368
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7800304093334514e-06,
+      "loss": 0.9762,
+      "step": 9369
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.779380022811244e-06,
+      "loss": 0.9113,
+      "step": 9370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.778729601813954e-06,
+      "loss": 1.0478,
+      "step": 9371
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.778079146354179e-06,
+      "loss": 0.8902,
+      "step": 9372
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.777428656444526e-06,
+      "loss": 1.1071,
+      "step": 9373
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.776778132097596e-06,
+      "loss": 1.0045,
+      "step": 9374
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.776127573325993e-06,
+      "loss": 1.0244,
+      "step": 9375
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.775476980142319e-06,
+      "loss": 1.0804,
+      "step": 9376
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.774826352559183e-06,
+      "loss": 0.9347,
+      "step": 9377
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7741756905891885e-06,
+      "loss": 0.9903,
+      "step": 9378
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.773524994244942e-06,
+      "loss": 1.0043,
+      "step": 9379
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.772874263539052e-06,
+      "loss": 1.0117,
+      "step": 9380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.772223498484125e-06,
+      "loss": 0.9902,
+      "step": 9381
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.77157269909277e-06,
+      "loss": 0.9002,
+      "step": 9382
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.770921865377595e-06,
+      "loss": 0.8146,
+      "step": 9383
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.770270997351211e-06,
+      "loss": 0.8956,
+      "step": 9384
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.769620095026229e-06,
+      "loss": 0.874,
+      "step": 9385
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.768969158415258e-06,
+      "loss": 0.9741,
+      "step": 9386
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.768318187530911e-06,
+      "loss": 1.0472,
+      "step": 9387
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.767667182385801e-06,
+      "loss": 1.0871,
+      "step": 9388
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.76701614299254e-06,
+      "loss": 0.9128,
+      "step": 9389
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.766365069363739e-06,
+      "loss": 1.2078,
+      "step": 9390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.765713961512019e-06,
+      "loss": 0.968,
+      "step": 9391
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.76506281944999e-06,
+      "loss": 0.9176,
+      "step": 9392
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.764411643190271e-06,
+      "loss": 0.777,
+      "step": 9393
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.763760432745475e-06,
+      "loss": 1.0473,
+      "step": 9394
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.763109188128222e-06,
+      "loss": 0.8834,
+      "step": 9395
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.762457909351125e-06,
+      "loss": 0.9835,
+      "step": 9396
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.76180659642681e-06,
+      "loss": 1.0443,
+      "step": 9397
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.76115524936789e-06,
+      "loss": 0.8781,
+      "step": 9398
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7605038681869876e-06,
+      "loss": 0.9363,
+      "step": 9399
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.75985245289672e-06,
+      "loss": 0.982,
+      "step": 9400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.759201003509712e-06,
+      "loss": 0.8462,
+      "step": 9401
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.758549520038582e-06,
+      "loss": 0.8318,
+      "step": 9402
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.757898002495956e-06,
+      "loss": 1.0262,
+      "step": 9403
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.757246450894455e-06,
+      "loss": 1.0518,
+      "step": 9404
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.756594865246702e-06,
+      "loss": 0.939,
+      "step": 9405
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.755943245565321e-06,
+      "loss": 0.9268,
+      "step": 9406
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.755291591862938e-06,
+      "loss": 0.9373,
+      "step": 9407
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.754639904152179e-06,
+      "loss": 1.0015,
+      "step": 9408
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7539881824456695e-06,
+      "loss": 0.9716,
+      "step": 9409
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.753336426756039e-06,
+      "loss": 1.2549,
+      "step": 9410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.752684637095911e-06,
+      "loss": 0.9868,
+      "step": 9411
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.752032813477916e-06,
+      "loss": 1.0988,
+      "step": 9412
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.751380955914681e-06,
+      "loss": 0.8957,
+      "step": 9413
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.750729064418839e-06,
+      "loss": 0.7572,
+      "step": 9414
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.75007713900302e-06,
+      "loss": 0.936,
+      "step": 9415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.749425179679852e-06,
+      "loss": 0.9944,
+      "step": 9416
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.748773186461968e-06,
+      "loss": 0.9523,
+      "step": 9417
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.748121159362e-06,
+      "loss": 0.9475,
+      "step": 9418
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.747469098392581e-06,
+      "loss": 0.8416,
+      "step": 9419
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.746817003566345e-06,
+      "loss": 0.7842,
+      "step": 9420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.746164874895926e-06,
+      "loss": 0.9278,
+      "step": 9421
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.745512712393958e-06,
+      "loss": 0.8267,
+      "step": 9422
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.744860516073078e-06,
+      "loss": 1.0782,
+      "step": 9423
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.74420828594592e-06,
+      "loss": 0.9978,
+      "step": 9424
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.743556022025121e-06,
+      "loss": 0.9982,
+      "step": 9425
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.742903724323323e-06,
+      "loss": 0.8876,
+      "step": 9426
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.742251392853158e-06,
+      "loss": 0.9365,
+      "step": 9427
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.74159902762727e-06,
+      "loss": 0.8788,
+      "step": 9428
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.740946628658295e-06,
+      "loss": 0.9328,
+      "step": 9429
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.740294195958873e-06,
+      "loss": 0.8161,
+      "step": 9430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.739641729541645e-06,
+      "loss": 1.1326,
+      "step": 9431
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.738989229419255e-06,
+      "loss": 0.9262,
+      "step": 9432
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.738336695604342e-06,
+      "loss": 0.8618,
+      "step": 9433
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.737684128109551e-06,
+      "loss": 0.9931,
+      "step": 9434
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.737031526947523e-06,
+      "loss": 0.8517,
+      "step": 9435
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.736378892130903e-06,
+      "loss": 0.9428,
+      "step": 9436
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.735726223672336e-06,
+      "loss": 0.9472,
+      "step": 9437
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.735073521584467e-06,
+      "loss": 0.8886,
+      "step": 9438
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.734420785879944e-06,
+      "loss": 1.0996,
+      "step": 9439
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.73376801657141e-06,
+      "loss": 0.8721,
+      "step": 9440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.733115213671515e-06,
+      "loss": 0.8748,
+      "step": 9441
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.732462377192905e-06,
+      "loss": 1.0493,
+      "step": 9442
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7318095071482294e-06,
+      "loss": 0.9508,
+      "step": 9443
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.731156603550137e-06,
+      "loss": 0.9149,
+      "step": 9444
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.73050366641128e-06,
+      "loss": 0.961,
+      "step": 9445
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.729850695744306e-06,
+      "loss": 0.9191,
+      "step": 9446
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.729197691561866e-06,
+      "loss": 0.8834,
+      "step": 9447
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.728544653876615e-06,
+      "loss": 0.9743,
+      "step": 9448
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.727891582701202e-06,
+      "loss": 0.8621,
+      "step": 9449
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.727238478048282e-06,
+      "loss": 1.117,
+      "step": 9450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.726585339930509e-06,
+      "loss": 1.0157,
+      "step": 9451
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.725932168360536e-06,
+      "loss": 0.8843,
+      "step": 9452
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.725278963351019e-06,
+      "loss": 1.0325,
+      "step": 9453
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7246257249146115e-06,
+      "loss": 0.9683,
+      "step": 9454
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.723972453063974e-06,
+      "loss": 1.0882,
+      "step": 9455
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.723319147811761e-06,
+      "loss": 0.8971,
+      "step": 9456
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.72266580917063e-06,
+      "loss": 0.9797,
+      "step": 9457
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.72201243715324e-06,
+      "loss": 0.9142,
+      "step": 9458
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.72135903177225e-06,
+      "loss": 0.833,
+      "step": 9459
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.720705593040317e-06,
+      "loss": 0.7244,
+      "step": 9460
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.720052120970106e-06,
+      "loss": 0.9358,
+      "step": 9461
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.719398615574274e-06,
+      "loss": 0.957,
+      "step": 9462
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.718745076865485e-06,
+      "loss": 0.9203,
+      "step": 9463
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.718091504856399e-06,
+      "loss": 0.7423,
+      "step": 9464
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.71743789955968e-06,
+      "loss": 1.0999,
+      "step": 9465
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.716784260987991e-06,
+      "loss": 1.077,
+      "step": 9466
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.716130589153997e-06,
+      "loss": 0.9654,
+      "step": 9467
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.715476884070362e-06,
+      "loss": 1.1181,
+      "step": 9468
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.714823145749751e-06,
+      "loss": 0.7761,
+      "step": 9469
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.714169374204832e-06,
+      "loss": 1.0425,
+      "step": 9470
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.713515569448268e-06,
+      "loss": 0.9512,
+      "step": 9471
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.71286173149273e-06,
+      "loss": 0.9568,
+      "step": 9472
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.712207860350884e-06,
+      "loss": 0.8965,
+      "step": 9473
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.711553956035399e-06,
+      "loss": 0.9547,
+      "step": 9474
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7109000185589454e-06,
+      "loss": 0.8562,
+      "step": 9475
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.710246047934191e-06,
+      "loss": 0.9996,
+      "step": 9476
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.709592044173808e-06,
+      "loss": 0.9046,
+      "step": 9477
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.708938007290468e-06,
+      "loss": 1.1642,
+      "step": 9478
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7082839372968404e-06,
+      "loss": 0.8062,
+      "step": 9479
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.707629834205598e-06,
+      "loss": 0.8774,
+      "step": 9480
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.706975698029417e-06,
+      "loss": 0.9351,
+      "step": 9481
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7063215287809676e-06,
+      "loss": 0.9477,
+      "step": 9482
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.705667326472926e-06,
+      "loss": 1.0137,
+      "step": 9483
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.705013091117967e-06,
+      "loss": 0.8592,
+      "step": 9484
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.704358822728765e-06,
+      "loss": 0.9987,
+      "step": 9485
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.703704521317999e-06,
+      "loss": 1.0935,
+      "step": 9486
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.703050186898343e-06,
+      "loss": 0.8284,
+      "step": 9487
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.702395819482478e-06,
+      "loss": 1.0884,
+      "step": 9488
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.701741419083078e-06,
+      "loss": 0.924,
+      "step": 9489
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.701086985712825e-06,
+      "loss": 0.9774,
+      "step": 9490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.700432519384396e-06,
+      "loss": 0.8506,
+      "step": 9491
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.699778020110475e-06,
+      "loss": 1.058,
+      "step": 9492
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.69912348790374e-06,
+      "loss": 0.9986,
+      "step": 9493
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6984689227768715e-06,
+      "loss": 0.7565,
+      "step": 9494
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.697814324742554e-06,
+      "loss": 0.9263,
+      "step": 9495
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.697159693813469e-06,
+      "loss": 0.962,
+      "step": 9496
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6965050300023005e-06,
+      "loss": 0.9866,
+      "step": 9497
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.695850333321731e-06,
+      "loss": 0.856,
+      "step": 9498
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.695195603784446e-06,
+      "loss": 0.9238,
+      "step": 9499
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6945408414031295e-06,
+      "loss": 1.0562,
+      "step": 9500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.693886046190471e-06,
+      "loss": 0.8686,
+      "step": 9501
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.693231218159153e-06,
+      "loss": 0.9305,
+      "step": 9502
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.692576357321865e-06,
+      "loss": 0.8527,
+      "step": 9503
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.691921463691294e-06,
+      "loss": 1.0855,
+      "step": 9504
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.691266537280128e-06,
+      "loss": 0.9421,
+      "step": 9505
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.690611578101057e-06,
+      "loss": 0.8352,
+      "step": 9506
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.68995658616677e-06,
+      "loss": 1.0215,
+      "step": 9507
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.689301561489956e-06,
+      "loss": 1.103,
+      "step": 9508
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.688646504083309e-06,
+      "loss": 0.8405,
+      "step": 9509
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.687991413959519e-06,
+      "loss": 1.0855,
+      "step": 9510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.687336291131279e-06,
+      "loss": 0.6633,
+      "step": 9511
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6866811356112805e-06,
+      "loss": 0.8733,
+      "step": 9512
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.686025947412217e-06,
+      "loss": 0.9476,
+      "step": 9513
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.685370726546784e-06,
+      "loss": 0.8424,
+      "step": 9514
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6847154730276745e-06,
+      "loss": 0.8653,
+      "step": 9515
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.684060186867586e-06,
+      "loss": 0.9959,
+      "step": 9516
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.683404868079213e-06,
+      "loss": 0.9801,
+      "step": 9517
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.682749516675251e-06,
+      "loss": 1.0042,
+      "step": 9518
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6820941326684e-06,
+      "loss": 0.9052,
+      "step": 9519
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.681438716071358e-06,
+      "loss": 0.9478,
+      "step": 9520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.680783266896821e-06,
+      "loss": 0.9482,
+      "step": 9521
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.680127785157489e-06,
+      "loss": 0.9882,
+      "step": 9522
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.679472270866063e-06,
+      "loss": 1.0944,
+      "step": 9523
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.678816724035242e-06,
+      "loss": 1.0223,
+      "step": 9524
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.678161144677728e-06,
+      "loss": 0.9389,
+      "step": 9525
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.677505532806223e-06,
+      "loss": 0.9405,
+      "step": 9526
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.67684988843343e-06,
+      "loss": 0.9296,
+      "step": 9527
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.676194211572049e-06,
+      "loss": 1.0382,
+      "step": 9528
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.675538502234785e-06,
+      "loss": 0.8313,
+      "step": 9529
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.674882760434344e-06,
+      "loss": 0.6524,
+      "step": 9530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.674226986183428e-06,
+      "loss": 0.904,
+      "step": 9531
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.673571179494746e-06,
+      "loss": 0.936,
+      "step": 9532
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.672915340381001e-06,
+      "loss": 1.0704,
+      "step": 9533
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.672259468854901e-06,
+      "loss": 0.8324,
+      "step": 9534
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.671603564929151e-06,
+      "loss": 1.0172,
+      "step": 9535
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.670947628616463e-06,
+      "loss": 0.9534,
+      "step": 9536
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.670291659929544e-06,
+      "loss": 0.9109,
+      "step": 9537
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.669635658881103e-06,
+      "loss": 0.8574,
+      "step": 9538
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.668979625483849e-06,
+      "loss": 0.9282,
+      "step": 9539
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.668323559750494e-06,
+      "loss": 0.8548,
+      "step": 9540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.667667461693749e-06,
+      "loss": 1.0147,
+      "step": 9541
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.667011331326324e-06,
+      "loss": 1.0595,
+      "step": 9542
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.666355168660933e-06,
+      "loss": 1.0939,
+      "step": 9543
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.665698973710289e-06,
+      "loss": 0.9577,
+      "step": 9544
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6650427464871045e-06,
+      "loss": 1.1366,
+      "step": 9545
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.664386487004095e-06,
+      "loss": 0.8315,
+      "step": 9546
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.663730195273974e-06,
+      "loss": 0.8986,
+      "step": 9547
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.663073871309459e-06,
+      "loss": 1.0005,
+      "step": 9548
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.662417515123265e-06,
+      "loss": 0.9126,
+      "step": 9549
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.661761126728108e-06,
+      "loss": 0.9737,
+      "step": 9550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.661104706136707e-06,
+      "loss": 0.8597,
+      "step": 9551
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.660448253361779e-06,
+      "loss": 1.0131,
+      "step": 9552
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6597917684160405e-06,
+      "loss": 1.3243,
+      "step": 9553
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.659135251312215e-06,
+      "loss": 0.9649,
+      "step": 9554
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.65847870206302e-06,
+      "loss": 1.0254,
+      "step": 9555
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.657822120681177e-06,
+      "loss": 1.1216,
+      "step": 9556
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.657165507179406e-06,
+      "loss": 1.0416,
+      "step": 9557
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.656508861570429e-06,
+      "loss": 0.9042,
+      "step": 9558
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.655852183866969e-06,
+      "loss": 1.0563,
+      "step": 9559
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6551954740817486e-06,
+      "loss": 0.7265,
+      "step": 9560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.65453873222749e-06,
+      "loss": 1.0109,
+      "step": 9561
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.65388195831692e-06,
+      "loss": 0.9045,
+      "step": 9562
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.653225152362762e-06,
+      "loss": 0.9356,
+      "step": 9563
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.652568314377741e-06,
+      "loss": 1.0537,
+      "step": 9564
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6519114443745844e-06,
+      "loss": 0.8588,
+      "step": 9565
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.651254542366019e-06,
+      "loss": 0.8242,
+      "step": 9566
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.650597608364771e-06,
+      "loss": 0.8296,
+      "step": 9567
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.649940642383569e-06,
+      "loss": 1.0352,
+      "step": 9568
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.649283644435141e-06,
+      "loss": 1.0143,
+      "step": 9569
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.648626614532217e-06,
+      "loss": 1.0277,
+      "step": 9570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.647969552687525e-06,
+      "loss": 0.881,
+      "step": 9571
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.647312458913797e-06,
+      "loss": 1.1653,
+      "step": 9572
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.646655333223765e-06,
+      "loss": 0.9887,
+      "step": 9573
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.64599817563016e-06,
+      "loss": 0.7834,
+      "step": 9574
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.645340986145712e-06,
+      "loss": 0.8783,
+      "step": 9575
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.644683764783154e-06,
+      "loss": 0.991,
+      "step": 9576
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.644026511555224e-06,
+      "loss": 0.934,
+      "step": 9577
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.643369226474652e-06,
+      "loss": 0.958,
+      "step": 9578
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6427119095541745e-06,
+      "loss": 0.9248,
+      "step": 9579
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.642054560806527e-06,
+      "loss": 0.7909,
+      "step": 9580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.641397180244444e-06,
+      "loss": 0.9969,
+      "step": 9581
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.640739767880662e-06,
+      "loss": 0.9829,
+      "step": 9582
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.640082323727921e-06,
+      "loss": 0.8892,
+      "step": 9583
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6394248477989566e-06,
+      "loss": 1.025,
+      "step": 9584
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.638767340106506e-06,
+      "loss": 0.9845,
+      "step": 9585
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.638109800663312e-06,
+      "loss": 0.9451,
+      "step": 9586
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.637452229482111e-06,
+      "loss": 1.1642,
+      "step": 9587
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.636794626575645e-06,
+      "loss": 0.7642,
+      "step": 9588
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.636136991956654e-06,
+      "loss": 1.0135,
+      "step": 9589
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6354793256378816e-06,
+      "loss": 1.0165,
+      "step": 9590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.634821627632066e-06,
+      "loss": 0.8359,
+      "step": 9591
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.634163897951954e-06,
+      "loss": 1.0098,
+      "step": 9592
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.633506136610285e-06,
+      "loss": 0.8933,
+      "step": 9593
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.632848343619806e-06,
+      "loss": 0.8689,
+      "step": 9594
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6321905189932625e-06,
+      "loss": 0.9957,
+      "step": 9595
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.631532662743396e-06,
+      "loss": 1.0095,
+      "step": 9596
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.630874774882956e-06,
+      "loss": 0.8353,
+      "step": 9597
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.630216855424686e-06,
+      "loss": 0.8537,
+      "step": 9598
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.629558904381334e-06,
+      "loss": 0.8173,
+      "step": 9599
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.628900921765649e-06,
+      "loss": 0.8302,
+      "step": 9600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.628242907590376e-06,
+      "loss": 0.9204,
+      "step": 9601
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.627584861868268e-06,
+      "loss": 0.9516,
+      "step": 9602
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6269267846120726e-06,
+      "loss": 0.8132,
+      "step": 9603
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.626268675834539e-06,
+      "loss": 0.8421,
+      "step": 9604
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.625610535548418e-06,
+      "loss": 1.022,
+      "step": 9605
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.624952363766462e-06,
+      "loss": 1.0559,
+      "step": 9606
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6242941605014234e-06,
+      "loss": 0.9367,
+      "step": 9607
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.623635925766053e-06,
+      "loss": 0.969,
+      "step": 9608
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.622977659573105e-06,
+      "loss": 1.0673,
+      "step": 9609
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.622319361935334e-06,
+      "loss": 0.7295,
+      "step": 9610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.621661032865492e-06,
+      "loss": 0.8988,
+      "step": 9611
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.621002672376335e-06,
+      "loss": 0.9354,
+      "step": 9612
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6203442804806205e-06,
+      "loss": 0.943,
+      "step": 9613
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.619685857191103e-06,
+      "loss": 0.8626,
+      "step": 9614
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.61902740252054e-06,
+      "loss": 0.7963,
+      "step": 9615
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.618368916481686e-06,
+      "loss": 0.9955,
+      "step": 9616
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6177103990873015e-06,
+      "loss": 0.949,
+      "step": 9617
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.617051850350147e-06,
+      "loss": 1.0337,
+      "step": 9618
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.616393270282978e-06,
+      "loss": 0.858,
+      "step": 9619
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.615734658898557e-06,
+      "loss": 1.0842,
+      "step": 9620
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.615076016209644e-06,
+      "loss": 1.004,
+      "step": 9621
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.614417342228998e-06,
+      "loss": 0.9287,
+      "step": 9622
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.613758636969383e-06,
+      "loss": 0.9336,
+      "step": 9623
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.61309990044356e-06,
+      "loss": 1.0101,
+      "step": 9624
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.612441132664293e-06,
+      "loss": 0.9318,
+      "step": 9625
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.611782333644346e-06,
+      "loss": 0.8057,
+      "step": 9626
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.61112350339648e-06,
+      "loss": 0.9992,
+      "step": 9627
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.610464641933462e-06,
+      "loss": 0.882,
+      "step": 9628
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6098057492680565e-06,
+      "loss": 0.9513,
+      "step": 9629
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.609146825413031e-06,
+      "loss": 1.0358,
+      "step": 9630
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.608487870381149e-06,
+      "loss": 0.9594,
+      "step": 9631
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6078288841851816e-06,
+      "loss": 1.101,
+      "step": 9632
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.607169866837892e-06,
+      "loss": 1.1792,
+      "step": 9633
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.606510818352053e-06,
+      "loss": 0.9958,
+      "step": 9634
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.605851738740431e-06,
+      "loss": 1.0212,
+      "step": 9635
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.605192628015795e-06,
+      "loss": 0.831,
+      "step": 9636
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.604533486190916e-06,
+      "loss": 1.0203,
+      "step": 9637
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6038743132785645e-06,
+      "loss": 0.9639,
+      "step": 9638
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.603215109291513e-06,
+      "loss": 0.9219,
+      "step": 9639
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.602555874242531e-06,
+      "loss": 0.8885,
+      "step": 9640
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.601896608144393e-06,
+      "loss": 1.1109,
+      "step": 9641
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.601237311009871e-06,
+      "loss": 1.0281,
+      "step": 9642
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6005779828517415e-06,
+      "loss": 0.9477,
+      "step": 9643
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.599918623682775e-06,
+      "loss": 0.7732,
+      "step": 9644
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.599259233515748e-06,
+      "loss": 0.8656,
+      "step": 9645
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.598599812363435e-06,
+      "loss": 0.9267,
+      "step": 9646
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.597940360238615e-06,
+      "loss": 0.9643,
+      "step": 9647
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.597280877154063e-06,
+      "loss": 1.0399,
+      "step": 9648
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.5966213631225565e-06,
+      "loss": 1.0475,
+      "step": 9649
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.595961818156873e-06,
+      "loss": 0.9715,
+      "step": 9650
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.59530224226979e-06,
+      "loss": 1.1189,
+      "step": 9651
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.594642635474089e-06,
+      "loss": 1.0998,
+      "step": 9652
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.593982997782549e-06,
+      "loss": 1.0607,
+      "step": 9653
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.593323329207949e-06,
+      "loss": 1.0462,
+      "step": 9654
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.592663629763072e-06,
+      "loss": 1.1398,
+      "step": 9655
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.592003899460698e-06,
+      "loss": 0.873,
+      "step": 9656
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5913441383136086e-06,
+      "loss": 0.7779,
+      "step": 9657
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.590684346334588e-06,
+      "loss": 0.9122,
+      "step": 9658
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.590024523536421e-06,
+      "loss": 1.1189,
+      "step": 9659
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.589364669931888e-06,
+      "loss": 0.864,
+      "step": 9660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.588704785533775e-06,
+      "loss": 0.9276,
+      "step": 9661
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.588044870354867e-06,
+      "loss": 0.8991,
+      "step": 9662
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.58738492440795e-06,
+      "loss": 0.8683,
+      "step": 9663
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5867249477058105e-06,
+      "loss": 0.6791,
+      "step": 9664
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.586064940261235e-06,
+      "loss": 0.9603,
+      "step": 9665
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.585404902087011e-06,
+      "loss": 0.9804,
+      "step": 9666
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5847448331959265e-06,
+      "loss": 0.8623,
+      "step": 9667
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.584084733600771e-06,
+      "loss": 1.0467,
+      "step": 9668
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.583424603314331e-06,
+      "loss": 0.9564,
+      "step": 9669
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5827644423494005e-06,
+      "loss": 1.2295,
+      "step": 9670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.582104250718767e-06,
+      "loss": 0.9746,
+      "step": 9671
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5814440284352236e-06,
+      "loss": 0.8381,
+      "step": 9672
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.58078377551156e-06,
+      "loss": 0.8836,
+      "step": 9673
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.580123491960571e-06,
+      "loss": 1.035,
+      "step": 9674
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.579463177795045e-06,
+      "loss": 0.9148,
+      "step": 9675
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.57880283302778e-06,
+      "loss": 1.1849,
+      "step": 9676
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.578142457671567e-06,
+      "loss": 0.933,
+      "step": 9677
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5774820517392035e-06,
+      "loss": 0.9391,
+      "step": 9678
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.576821615243482e-06,
+      "loss": 0.9053,
+      "step": 9679
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5761611481971985e-06,
+      "loss": 1.1582,
+      "step": 9680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5755006506131515e-06,
+      "loss": 1.0225,
+      "step": 9681
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.574840122504137e-06,
+      "loss": 1.0512,
+      "step": 9682
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5741795638829516e-06,
+      "loss": 0.9697,
+      "step": 9683
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.573518974762395e-06,
+      "loss": 1.0505,
+      "step": 9684
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5728583551552645e-06,
+      "loss": 0.6767,
+      "step": 9685
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.57219770507436e-06,
+      "loss": 1.0256,
+      "step": 9686
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.571537024532483e-06,
+      "loss": 1.0793,
+      "step": 9687
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5708763135424314e-06,
+      "loss": 0.8974,
+      "step": 9688
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5702155721170095e-06,
+      "loss": 0.8725,
+      "step": 9689
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.569554800269014e-06,
+      "loss": 1.012,
+      "step": 9690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.568893998011252e-06,
+      "loss": 1.0595,
+      "step": 9691
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5682331653565255e-06,
+      "loss": 0.9278,
+      "step": 9692
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.567572302317635e-06,
+      "loss": 0.9494,
+      "step": 9693
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.566911408907388e-06,
+      "loss": 1.0573,
+      "step": 9694
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.566250485138589e-06,
+      "loss": 0.9347,
+      "step": 9695
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.565589531024041e-06,
+      "loss": 0.8921,
+      "step": 9696
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.564928546576551e-06,
+      "loss": 0.9742,
+      "step": 9697
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.564267531808924e-06,
+      "loss": 0.9243,
+      "step": 9698
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.56360648673397e-06,
+      "loss": 1.0691,
+      "step": 9699
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.562945411364494e-06,
+      "loss": 1.0717,
+      "step": 9700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.562284305713307e-06,
+      "loss": 0.9567,
+      "step": 9701
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.561623169793214e-06,
+      "loss": 0.8913,
+      "step": 9702
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.560962003617028e-06,
+      "loss": 0.9974,
+      "step": 9703
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.560300807197555e-06,
+      "loss": 0.9817,
+      "step": 9704
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.55963958054761e-06,
+      "loss": 1.1147,
+      "step": 9705
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.558978323680001e-06,
+      "loss": 1.0285,
+      "step": 9706
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.558317036607541e-06,
+      "loss": 0.8405,
+      "step": 9707
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5576557193430415e-06,
+      "loss": 0.8436,
+      "step": 9708
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5569943718993144e-06,
+      "loss": 0.8204,
+      "step": 9709
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5563329942891765e-06,
+      "loss": 0.9348,
+      "step": 9710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.55567158652544e-06,
+      "loss": 0.9558,
+      "step": 9711
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.555010148620918e-06,
+      "loss": 0.9311,
+      "step": 9712
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.554348680588428e-06,
+      "loss": 0.8514,
+      "step": 9713
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.553687182440786e-06,
+      "loss": 0.9459,
+      "step": 9714
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.553025654190804e-06,
+      "loss": 1.0543,
+      "step": 9715
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5523640958513046e-06,
+      "loss": 0.9282,
+      "step": 9716
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.551702507435104e-06,
+      "loss": 0.8994,
+      "step": 9717
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.551040888955018e-06,
+      "loss": 0.8268,
+      "step": 9718
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.550379240423866e-06,
+      "loss": 0.896,
+      "step": 9719
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.549717561854468e-06,
+      "loss": 0.8781,
+      "step": 9720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.549055853259644e-06,
+      "loss": 0.8776,
+      "step": 9721
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.548394114652215e-06,
+      "loss": 0.9514,
+      "step": 9722
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5477323460450006e-06,
+      "loss": 0.9852,
+      "step": 9723
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5470705474508235e-06,
+      "loss": 0.8578,
+      "step": 9724
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.546408718882505e-06,
+      "loss": 1.0483,
+      "step": 9725
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.545746860352869e-06,
+      "loss": 0.9662,
+      "step": 9726
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.545084971874738e-06,
+      "loss": 0.8412,
+      "step": 9727
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.544423053460935e-06,
+      "loss": 1.1094,
+      "step": 9728
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5437611051242874e-06,
+      "loss": 0.905,
+      "step": 9729
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.543099126877618e-06,
+      "loss": 0.9568,
+      "step": 9730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.542437118733753e-06,
+      "loss": 1.0568,
+      "step": 9731
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.541775080705519e-06,
+      "loss": 1.1129,
+      "step": 9732
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.541113012805741e-06,
+      "loss": 0.9832,
+      "step": 9733
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5404509150472494e-06,
+      "loss": 1.0138,
+      "step": 9734
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.53978878744287e-06,
+      "loss": 1.0269,
+      "step": 9735
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.539126630005432e-06,
+      "loss": 0.7744,
+      "step": 9736
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.538464442747766e-06,
+      "loss": 1.078,
+      "step": 9737
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5378022256827e-06,
+      "loss": 0.7543,
+      "step": 9738
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.537139978823064e-06,
+      "loss": 1.0571,
+      "step": 9739
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.536477702181689e-06,
+      "loss": 0.9778,
+      "step": 9740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.535815395771408e-06,
+      "loss": 0.7681,
+      "step": 9741
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.535153059605052e-06,
+      "loss": 0.6646,
+      "step": 9742
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.534490693695455e-06,
+      "loss": 0.8803,
+      "step": 9743
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.533828298055447e-06,
+      "loss": 1.0287,
+      "step": 9744
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.533165872697864e-06,
+      "loss": 0.9374,
+      "step": 9745
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.53250341763554e-06,
+      "loss": 0.8187,
+      "step": 9746
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.53184093288131e-06,
+      "loss": 0.9212,
+      "step": 9747
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.53117841844801e-06,
+      "loss": 0.8191,
+      "step": 9748
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5305158743484755e-06,
+      "loss": 1.0202,
+      "step": 9749
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.529853300595541e-06,
+      "loss": 0.8818,
+      "step": 9750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.529190697202048e-06,
+      "loss": 0.9004,
+      "step": 9751
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.528528064180833e-06,
+      "loss": 0.8562,
+      "step": 9752
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.527865401544732e-06,
+      "loss": 0.963,
+      "step": 9753
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.527202709306585e-06,
+      "loss": 0.9982,
+      "step": 9754
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.526539987479232e-06,
+      "loss": 0.962,
+      "step": 9755
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.525877236075514e-06,
+      "loss": 0.8042,
+      "step": 9756
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5252144551082705e-06,
+      "loss": 1.0984,
+      "step": 9757
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5245516445903435e-06,
+      "loss": 0.8859,
+      "step": 9758
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.523888804534573e-06,
+      "loss": 0.9909,
+      "step": 9759
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.523225934953803e-06,
+      "loss": 0.9587,
+      "step": 9760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.522563035860876e-06,
+      "loss": 1.0566,
+      "step": 9761
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.521900107268635e-06,
+      "loss": 0.9778,
+      "step": 9762
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5212371491899254e-06,
+      "loss": 1.0182,
+      "step": 9763
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.520574161637591e-06,
+      "loss": 1.0292,
+      "step": 9764
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5199111446244765e-06,
+      "loss": 1.1474,
+      "step": 9765
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.519248098163429e-06,
+      "loss": 0.945,
+      "step": 9766
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.518585022267294e-06,
+      "loss": 0.9188,
+      "step": 9767
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5179219169489185e-06,
+      "loss": 0.8511,
+      "step": 9768
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.517258782221149e-06,
+      "loss": 1.0305,
+      "step": 9769
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5165956180968375e-06,
+      "loss": 0.8507,
+      "step": 9770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.515932424588829e-06,
+      "loss": 0.8707,
+      "step": 9771
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5152692017099725e-06,
+      "loss": 0.962,
+      "step": 9772
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5146059494731184e-06,
+      "loss": 1.0475,
+      "step": 9773
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.51394266789112e-06,
+      "loss": 1.0362,
+      "step": 9774
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.513279356976825e-06,
+      "loss": 1.0212,
+      "step": 9775
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.512616016743085e-06,
+      "loss": 0.84,
+      "step": 9776
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.511952647202753e-06,
+      "loss": 0.7991,
+      "step": 9777
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.511289248368682e-06,
+      "loss": 0.8346,
+      "step": 9778
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5106258202537245e-06,
+      "loss": 0.9707,
+      "step": 9779
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.509962362870733e-06,
+      "loss": 0.8934,
+      "step": 9780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.509298876232565e-06,
+      "loss": 0.8676,
+      "step": 9781
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.508635360352074e-06,
+      "loss": 1.0646,
+      "step": 9782
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5079718152421135e-06,
+      "loss": 0.8045,
+      "step": 9783
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.507308240915542e-06,
+      "loss": 1.0446,
+      "step": 9784
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.506644637385215e-06,
+      "loss": 1.0002,
+      "step": 9785
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.50598100466399e-06,
+      "loss": 1.0456,
+      "step": 9786
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.505317342764726e-06,
+      "loss": 0.9561,
+      "step": 9787
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.504653651700278e-06,
+      "loss": 0.9219,
+      "step": 9788
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5039899314835075e-06,
+      "loss": 0.9646,
+      "step": 9789
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.503326182127273e-06,
+      "loss": 1.0606,
+      "step": 9790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.502662403644433e-06,
+      "loss": 0.8174,
+      "step": 9791
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5019985960478524e-06,
+      "loss": 0.8972,
+      "step": 9792
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.50133475935039e-06,
+      "loss": 0.9954,
+      "step": 9793
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.500670893564905e-06,
+      "loss": 0.8015,
+      "step": 9794
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.500006998704264e-06,
+      "loss": 1.0409,
+      "step": 9795
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.499343074781326e-06,
+      "loss": 1.191,
+      "step": 9796
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.498679121808956e-06,
+      "loss": 0.8401,
+      "step": 9797
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4980151398000195e-06,
+      "loss": 0.9189,
+      "step": 9798
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.497351128767378e-06,
+      "loss": 1.0036,
+      "step": 9799
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.496687088723899e-06,
+      "loss": 0.9627,
+      "step": 9800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.496023019682447e-06,
+      "loss": 0.9478,
+      "step": 9801
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.495358921655887e-06,
+      "loss": 1.0312,
+      "step": 9802
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4946947946570886e-06,
+      "loss": 1.0171,
+      "step": 9803
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.494030638698917e-06,
+      "loss": 1.0705,
+      "step": 9804
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.493366453794241e-06,
+      "loss": 0.9798,
+      "step": 9805
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.492702239955929e-06,
+      "loss": 1.145,
+      "step": 9806
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.492037997196849e-06,
+      "loss": 1.0091,
+      "step": 9807
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.491373725529871e-06,
+      "loss": 0.9847,
+      "step": 9808
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.490709424967867e-06,
+      "loss": 1.0195,
+      "step": 9809
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.490045095523705e-06,
+      "loss": 0.9185,
+      "step": 9810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.489380737210257e-06,
+      "loss": 1.0436,
+      "step": 9811
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.488716350040396e-06,
+      "loss": 0.8747,
+      "step": 9812
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.488051934026993e-06,
+      "loss": 0.8561,
+      "step": 9813
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.487387489182922e-06,
+      "loss": 1.0011,
+      "step": 9814
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.486723015521053e-06,
+      "loss": 0.9034,
+      "step": 9815
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.486058513054265e-06,
+      "loss": 0.9186,
+      "step": 9816
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4853939817954316e-06,
+      "loss": 1.0305,
+      "step": 9817
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.484729421757425e-06,
+      "loss": 0.9055,
+      "step": 9818
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.484064832953122e-06,
+      "loss": 0.9763,
+      "step": 9819
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4834002153954e-06,
+      "loss": 0.8006,
+      "step": 9820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.482735569097136e-06,
+      "loss": 0.9469,
+      "step": 9821
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4820708940712065e-06,
+      "loss": 0.8303,
+      "step": 9822
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.48140619033049e-06,
+      "loss": 1.0293,
+      "step": 9823
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.480741457887865e-06,
+      "loss": 0.8535,
+      "step": 9824
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.480076696756209e-06,
+      "loss": 1.0096,
+      "step": 9825
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4794119069484016e-06,
+      "loss": 0.741,
+      "step": 9826
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.478747088477326e-06,
+      "loss": 1.1844,
+      "step": 9827
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.478082241355861e-06,
+      "loss": 0.8458,
+      "step": 9828
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.477417365596887e-06,
+      "loss": 0.901,
+      "step": 9829
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.476752461213287e-06,
+      "loss": 0.9745,
+      "step": 9830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4760875282179435e-06,
+      "loss": 0.9451,
+      "step": 9831
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.475422566623737e-06,
+      "loss": 0.9691,
+      "step": 9832
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.474757576443555e-06,
+      "loss": 0.9133,
+      "step": 9833
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.474092557690279e-06,
+      "loss": 0.913,
+      "step": 9834
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.473427510376795e-06,
+      "loss": 1.1417,
+      "step": 9835
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.472762434515985e-06,
+      "loss": 0.8318,
+      "step": 9836
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.472097330120738e-06,
+      "loss": 0.9681,
+      "step": 9837
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4714321972039395e-06,
+      "loss": 0.9105,
+      "step": 9838
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.470767035778476e-06,
+      "loss": 0.9921,
+      "step": 9839
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.470101845857235e-06,
+      "loss": 0.9955,
+      "step": 9840
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.469436627453104e-06,
+      "loss": 0.8883,
+      "step": 9841
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.468771380578971e-06,
+      "loss": 0.9803,
+      "step": 9842
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.468106105247725e-06,
+      "loss": 0.9832,
+      "step": 9843
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4674408014722575e-06,
+      "loss": 1.0053,
+      "step": 9844
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.466775469265458e-06,
+      "loss": 1.0656,
+      "step": 9845
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.466110108640216e-06,
+      "loss": 0.9716,
+      "step": 9846
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.465444719609424e-06,
+      "loss": 0.9401,
+      "step": 9847
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.464779302185972e-06,
+      "loss": 1.0043,
+      "step": 9848
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.464113856382752e-06,
+      "loss": 0.9648,
+      "step": 9849
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4634483822126605e-06,
+      "loss": 1.2133,
+      "step": 9850
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.462782879688587e-06,
+      "loss": 0.8165,
+      "step": 9851
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.462117348823428e-06,
+      "loss": 1.1079,
+      "step": 9852
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.461451789630077e-06,
+      "loss": 0.9898,
+      "step": 9853
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.460786202121429e-06,
+      "loss": 1.0342,
+      "step": 9854
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4601205863103776e-06,
+      "loss": 1.0375,
+      "step": 9855
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.459454942209823e-06,
+      "loss": 1.0237,
+      "step": 9856
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4587892698326595e-06,
+      "loss": 1.0108,
+      "step": 9857
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4581235691917855e-06,
+      "loss": 0.9124,
+      "step": 9858
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.457457840300098e-06,
+      "loss": 1.078,
+      "step": 9859
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.456792083170494e-06,
+      "loss": 0.7929,
+      "step": 9860
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.456126297815875e-06,
+      "loss": 1.0621,
+      "step": 9861
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.455460484249138e-06,
+      "loss": 1.0864,
+      "step": 9862
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.454794642483187e-06,
+      "loss": 1.0051,
+      "step": 9863
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.454128772530917e-06,
+      "loss": 0.9174,
+      "step": 9864
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.453462874405234e-06,
+      "loss": 0.8875,
+      "step": 9865
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.452796948119036e-06,
+      "loss": 0.8988,
+      "step": 9866
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.452130993685227e-06,
+      "loss": 0.7461,
+      "step": 9867
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.451465011116711e-06,
+      "loss": 0.9608,
+      "step": 9868
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.450799000426388e-06,
+      "loss": 0.7733,
+      "step": 9869
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4501329616271634e-06,
+      "loss": 0.8183,
+      "step": 9870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4494668947319426e-06,
+      "loss": 0.9068,
+      "step": 9871
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.448800799753628e-06,
+      "loss": 0.9486,
+      "step": 9872
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.448134676705128e-06,
+      "loss": 1.0709,
+      "step": 9873
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.447468525599347e-06,
+      "loss": 0.814,
+      "step": 9874
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4468023464491906e-06,
+      "loss": 0.7618,
+      "step": 9875
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.446136139267568e-06,
+      "loss": 0.7997,
+      "step": 9876
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.445469904067385e-06,
+      "loss": 0.9,
+      "step": 9877
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.444803640861552e-06,
+      "loss": 1.0208,
+      "step": 9878
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.444137349662974e-06,
+      "loss": 1.0217,
+      "step": 9879
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.443471030484565e-06,
+      "loss": 0.8884,
+      "step": 9880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.44280468333923e-06,
+      "loss": 0.9262,
+      "step": 9881
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4421383082398824e-06,
+      "loss": 0.9733,
+      "step": 9882
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.441471905199432e-06,
+      "loss": 0.9866,
+      "step": 9883
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.44080547423079e-06,
+      "loss": 0.9776,
+      "step": 9884
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.440139015346869e-06,
+      "loss": 1.0845,
+      "step": 9885
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.439472528560582e-06,
+      "loss": 1.1796,
+      "step": 9886
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.438806013884841e-06,
+      "loss": 0.9927,
+      "step": 9887
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4381394713325595e-06,
+      "loss": 0.9552,
+      "step": 9888
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4374729009166524e-06,
+      "loss": 0.8478,
+      "step": 9889
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.436806302650033e-06,
+      "loss": 1.1859,
+      "step": 9890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.436139676545618e-06,
+      "loss": 0.958,
+      "step": 9891
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4354730226163234e-06,
+      "loss": 0.8735,
+      "step": 9892
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4348063408750625e-06,
+      "loss": 0.9283,
+      "step": 9893
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.434139631334756e-06,
+      "loss": 0.9119,
+      "step": 9894
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.433472894008318e-06,
+      "loss": 1.0825,
+      "step": 9895
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.432806128908669e-06,
+      "loss": 1.0019,
+      "step": 9896
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.432139336048726e-06,
+      "loss": 0.92,
+      "step": 9897
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.431472515441408e-06,
+      "loss": 1.0012,
+      "step": 9898
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4308056670996355e-06,
+      "loss": 0.9877,
+      "step": 9899
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.430138791036328e-06,
+      "loss": 0.8572,
+      "step": 9900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.429471887264405e-06,
+      "loss": 0.9016,
+      "step": 9901
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.428804955796788e-06,
+      "loss": 1.0355,
+      "step": 9902
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.428137996646399e-06,
+      "loss": 0.8003,
+      "step": 9903
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.427471009826161e-06,
+      "loss": 0.8401,
+      "step": 9904
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.426803995348997e-06,
+      "loss": 0.8795,
+      "step": 9905
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.426136953227828e-06,
+      "loss": 0.9997,
+      "step": 9906
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.425469883475579e-06,
+      "loss": 0.8982,
+      "step": 9907
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.424802786105176e-06,
+      "loss": 0.7696,
+      "step": 9908
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.42413566112954e-06,
+      "loss": 1.0188,
+      "step": 9909
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.423468508561599e-06,
+      "loss": 0.808,
+      "step": 9910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.42280132841428e-06,
+      "loss": 0.973,
+      "step": 9911
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.422134120700506e-06,
+      "loss": 0.8725,
+      "step": 9912
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.421466885433206e-06,
+      "loss": 0.8765,
+      "step": 9913
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.420799622625309e-06,
+      "loss": 0.9148,
+      "step": 9914
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.420132332289742e-06,
+      "loss": 0.8908,
+      "step": 9915
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.419465014439432e-06,
+      "loss": 0.9974,
+      "step": 9916
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.418797669087309e-06,
+      "loss": 0.9477,
+      "step": 9917
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.418130296246305e-06,
+      "loss": 0.9371,
+      "step": 9918
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.417462895929345e-06,
+      "loss": 0.8909,
+      "step": 9919
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.416795468149365e-06,
+      "loss": 0.9568,
+      "step": 9920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.416128012919294e-06,
+      "loss": 0.916,
+      "step": 9921
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.415460530252065e-06,
+      "loss": 0.9603,
+      "step": 9922
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.414793020160608e-06,
+      "loss": 0.9706,
+      "step": 9923
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.414125482657856e-06,
+      "loss": 0.945,
+      "step": 9924
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4134579177567445e-06,
+      "loss": 0.9843,
+      "step": 9925
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.412790325470208e-06,
+      "loss": 0.9669,
+      "step": 9926
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.412122705811178e-06,
+      "loss": 1.0283,
+      "step": 9927
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4114550587925905e-06,
+      "loss": 1.1306,
+      "step": 9928
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.410787384427383e-06,
+      "loss": 0.9772,
+      "step": 9929
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.410119682728488e-06,
+      "loss": 0.9104,
+      "step": 9930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.409451953708845e-06,
+      "loss": 0.9882,
+      "step": 9931
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.40878419738139e-06,
+      "loss": 1.1016,
+      "step": 9932
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.408116413759062e-06,
+      "loss": 1.1154,
+      "step": 9933
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.407448602854797e-06,
+      "loss": 0.8757,
+      "step": 9934
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.406780764681534e-06,
+      "loss": 0.9557,
+      "step": 9935
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.406112899252212e-06,
+      "loss": 1.0457,
+      "step": 9936
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.405445006579773e-06,
+      "loss": 1.0536,
+      "step": 9937
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4047770866771564e-06,
+      "loss": 0.8314,
+      "step": 9938
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4041091395573015e-06,
+      "loss": 0.8781,
+      "step": 9939
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.403441165233151e-06,
+      "loss": 1.0494,
+      "step": 9940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.402773163717646e-06,
+      "loss": 0.9708,
+      "step": 9941
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4021051350237285e-06,
+      "loss": 1.0639,
+      "step": 9942
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.401437079164344e-06,
+      "loss": 1.0188,
+      "step": 9943
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.400768996152431e-06,
+      "loss": 0.8927,
+      "step": 9944
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4001008860009385e-06,
+      "loss": 0.8815,
+      "step": 9945
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3994327487228095e-06,
+      "loss": 0.9319,
+      "step": 9946
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3987645843309865e-06,
+      "loss": 0.9373,
+      "step": 9947
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.398096392838418e-06,
+      "loss": 1.203,
+      "step": 9948
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.397428174258048e-06,
+      "loss": 0.9511,
+      "step": 9949
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.396759928602825e-06,
+      "loss": 0.9323,
+      "step": 9950
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.396091655885695e-06,
+      "loss": 0.8127,
+      "step": 9951
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.395423356119606e-06,
+      "loss": 0.9553,
+      "step": 9952
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.394755029317507e-06,
+      "loss": 1.0343,
+      "step": 9953
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.394086675492343e-06,
+      "loss": 0.8715,
+      "step": 9954
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.393418294657069e-06,
+      "loss": 1.0178,
+      "step": 9955
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3927498868246295e-06,
+      "loss": 0.8194,
+      "step": 9956
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3920814520079784e-06,
+      "loss": 0.8661,
+      "step": 9957
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.391412990220064e-06,
+      "loss": 1.1019,
+      "step": 9958
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.390744501473838e-06,
+      "loss": 0.8932,
+      "step": 9959
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.390075985782254e-06,
+      "loss": 0.9259,
+      "step": 9960
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.389407443158262e-06,
+      "loss": 0.9633,
+      "step": 9961
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.388738873614817e-06,
+      "loss": 1.0435,
+      "step": 9962
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.38807027716487e-06,
+      "loss": 1.0429,
+      "step": 9963
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.387401653821378e-06,
+      "loss": 1.1204,
+      "step": 9964
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.386733003597292e-06,
+      "loss": 0.9386,
+      "step": 9965
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.38606432650557e-06,
+      "loss": 1.0051,
+      "step": 9966
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.385395622559165e-06,
+      "loss": 1.0099,
+      "step": 9967
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.384726891771035e-06,
+      "loss": 1.036,
+      "step": 9968
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.384058134154135e-06,
+      "loss": 0.9019,
+      "step": 9969
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.383389349721423e-06,
+      "loss": 0.8881,
+      "step": 9970
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.382720538485856e-06,
+      "loss": 0.8247,
+      "step": 9971
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.382051700460392e-06,
+      "loss": 0.9635,
+      "step": 9972
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.381382835657991e-06,
+      "loss": 0.9343,
+      "step": 9973
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.38071394409161e-06,
+      "loss": 0.9084,
+      "step": 9974
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3800450257742085e-06,
+      "loss": 0.9014,
+      "step": 9975
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.379376080718749e-06,
+      "loss": 0.8818,
+      "step": 9976
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.37870710893819e-06,
+      "loss": 1.0392,
+      "step": 9977
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.378038110445494e-06,
+      "loss": 0.863,
+      "step": 9978
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.377369085253622e-06,
+      "loss": 0.8758,
+      "step": 9979
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.376700033375537e-06,
+      "loss": 0.8212,
+      "step": 9980
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3760309548242e-06,
+      "loss": 0.9216,
+      "step": 9981
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.375361849612577e-06,
+      "loss": 0.8861,
+      "step": 9982
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.374692717753628e-06,
+      "loss": 0.958,
+      "step": 9983
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.37402355926032e-06,
+      "loss": 0.9366,
+      "step": 9984
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.373354374145619e-06,
+      "loss": 0.9506,
+      "step": 9985
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3726851624224875e-06,
+      "loss": 0.9822,
+      "step": 9986
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.372015924103892e-06,
+      "loss": 0.9418,
+      "step": 9987
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.371346659202799e-06,
+      "loss": 0.9386,
+      "step": 9988
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.370677367732176e-06,
+      "loss": 1.0862,
+      "step": 9989
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.37000804970499e-06,
+      "loss": 0.8956,
+      "step": 9990
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.369338705134209e-06,
+      "loss": 0.8004,
+      "step": 9991
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.368669334032801e-06,
+      "loss": 1.0348,
+      "step": 9992
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3679999364137346e-06,
+      "loss": 0.9112,
+      "step": 9993
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.36733051228998e-06,
+      "loss": 0.9493,
+      "step": 9994
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.366661061674506e-06,
+      "loss": 0.8809,
+      "step": 9995
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.365991584580286e-06,
+      "loss": 0.9615,
+      "step": 9996
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.365322081020288e-06,
+      "loss": 1.1088,
+      "step": 9997
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.364652551007484e-06,
+      "loss": 1.0052,
+      "step": 9998
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.363982994554847e-06,
+      "loss": 1.0202,
+      "step": 9999
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.363313411675349e-06,
+      "loss": 1.0325,
+      "step": 10000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.362643802381963e-06,
+      "loss": 0.9965,
+      "step": 10001
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.361974166687661e-06,
+      "loss": 0.858,
+      "step": 10002
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.36130450460542e-06,
+      "loss": 0.8598,
+      "step": 10003
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.360634816148212e-06,
+      "loss": 0.8627,
+      "step": 10004
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.359965101329014e-06,
+      "loss": 0.7967,
+      "step": 10005
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3592953601607986e-06,
+      "loss": 1.0665,
+      "step": 10006
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.358625592656546e-06,
+      "loss": 1.0197,
+      "step": 10007
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.357955798829231e-06,
+      "loss": 0.9434,
+      "step": 10008
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.357285978691829e-06,
+      "loss": 0.8629,
+      "step": 10009
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.356616132257321e-06,
+      "loss": 1.0795,
+      "step": 10010
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.355946259538681e-06,
+      "loss": 1.0528,
+      "step": 10011
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.35527636054889e-06,
+      "loss": 0.8802,
+      "step": 10012
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.354606435300928e-06,
+      "loss": 0.8983,
+      "step": 10013
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.353936483807772e-06,
+      "loss": 0.7263,
+      "step": 10014
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.353266506082405e-06,
+      "loss": 0.9591,
+      "step": 10015
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3525965021378065e-06,
+      "loss": 0.8688,
+      "step": 10016
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.351926471986957e-06,
+      "loss": 0.965,
+      "step": 10017
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.351256415642839e-06,
+      "loss": 0.9602,
+      "step": 10018
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.350586333118435e-06,
+      "loss": 1.0495,
+      "step": 10019
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.349916224426726e-06,
+      "loss": 0.9302,
+      "step": 10020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.349246089580699e-06,
+      "loss": 1.0357,
+      "step": 10021
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.348575928593333e-06,
+      "loss": 0.7968,
+      "step": 10022
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.347905741477613e-06,
+      "loss": 0.9273,
+      "step": 10023
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.347235528246527e-06,
+      "loss": 0.8405,
+      "step": 10024
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.346565288913058e-06,
+      "loss": 0.7732,
+      "step": 10025
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3458950234901915e-06,
+      "loss": 0.8564,
+      "step": 10026
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.345224731990916e-06,
+      "loss": 1.0573,
+      "step": 10027
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.344554414428214e-06,
+      "loss": 1.0562,
+      "step": 10028
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3438840708150765e-06,
+      "loss": 0.9489,
+      "step": 10029
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.34321370116449e-06,
+      "loss": 1.0413,
+      "step": 10030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3425433054894424e-06,
+      "loss": 0.9857,
+      "step": 10031
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.341872883802923e-06,
+      "loss": 0.941,
+      "step": 10032
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.341202436117919e-06,
+      "loss": 0.8635,
+      "step": 10033
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.340531962447425e-06,
+      "loss": 0.9773,
+      "step": 10034
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.339861462804427e-06,
+      "loss": 0.9588,
+      "step": 10035
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.339190937201916e-06,
+      "loss": 0.9625,
+      "step": 10036
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.338520385652886e-06,
+      "loss": 0.9674,
+      "step": 10037
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.337849808170326e-06,
+      "loss": 0.9155,
+      "step": 10038
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.337179204767231e-06,
+      "loss": 0.9157,
+      "step": 10039
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.336508575456591e-06,
+      "loss": 0.9803,
+      "step": 10040
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.335837920251401e-06,
+      "loss": 0.945,
+      "step": 10041
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.335167239164653e-06,
+      "loss": 1.0143,
+      "step": 10042
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.334496532209344e-06,
+      "loss": 1.024,
+      "step": 10043
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3338257993984675e-06,
+      "loss": 0.8503,
+      "step": 10044
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.333155040745018e-06,
+      "loss": 0.9431,
+      "step": 10045
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.332484256261994e-06,
+      "loss": 0.9251,
+      "step": 10046
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.331813445962387e-06,
+      "loss": 0.9853,
+      "step": 10047
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.331142609859197e-06,
+      "loss": 1.1271,
+      "step": 10048
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.330471747965422e-06,
+      "loss": 0.9035,
+      "step": 10049
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.329800860294059e-06,
+      "loss": 0.9273,
+      "step": 10050
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.329129946858105e-06,
+      "loss": 0.9302,
+      "step": 10051
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.328459007670558e-06,
+      "loss": 1.1989,
+      "step": 10052
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.327788042744423e-06,
+      "loss": 0.9544,
+      "step": 10053
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.327117052092693e-06,
+      "loss": 0.9239,
+      "step": 10054
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.326446035728373e-06,
+      "loss": 1.014,
+      "step": 10055
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.325774993664461e-06,
+      "loss": 0.9706,
+      "step": 10056
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.325103925913959e-06,
+      "loss": 0.9358,
+      "step": 10057
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3244328324898684e-06,
+      "loss": 0.8984,
+      "step": 10058
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.323761713405193e-06,
+      "loss": 1.0688,
+      "step": 10059
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.323090568672935e-06,
+      "loss": 1.0792,
+      "step": 10060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3224193983060975e-06,
+      "loss": 0.8643,
+      "step": 10061
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.321748202317683e-06,
+      "loss": 0.9565,
+      "step": 10062
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.321076980720698e-06,
+      "loss": 1.0687,
+      "step": 10063
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.320405733528146e-06,
+      "loss": 0.9998,
+      "step": 10064
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.319734460753033e-06,
+      "loss": 1.0256,
+      "step": 10065
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.319063162408363e-06,
+      "loss": 0.9298,
+      "step": 10066
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.318391838507144e-06,
+      "loss": 0.9524,
+      "step": 10067
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.317720489062382e-06,
+      "loss": 1.0258,
+      "step": 10068
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.317049114087085e-06,
+      "loss": 1.2092,
+      "step": 10069
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.316377713594258e-06,
+      "loss": 0.9222,
+      "step": 10070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.315706287596914e-06,
+      "loss": 0.9911,
+      "step": 10071
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.315034836108058e-06,
+      "loss": 0.8697,
+      "step": 10072
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3143633591407005e-06,
+      "loss": 1.0427,
+      "step": 10073
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3136918567078505e-06,
+      "loss": 1.0551,
+      "step": 10074
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.313020328822521e-06,
+      "loss": 1.0778,
+      "step": 10075
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3123487754977175e-06,
+      "loss": 0.9505,
+      "step": 10076
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.311677196746455e-06,
+      "loss": 0.9915,
+      "step": 10077
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.311005592581745e-06,
+      "loss": 0.8469,
+      "step": 10078
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.310333963016599e-06,
+      "loss": 0.8649,
+      "step": 10079
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.309662308064031e-06,
+      "loss": 0.9804,
+      "step": 10080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.308990627737049e-06,
+      "loss": 1.0472,
+      "step": 10081
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.308318922048674e-06,
+      "loss": 0.8993,
+      "step": 10082
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.307647191011915e-06,
+      "loss": 0.9487,
+      "step": 10083
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.306975434639789e-06,
+      "loss": 1.0688,
+      "step": 10084
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.306303652945312e-06,
+      "loss": 1.0688,
+      "step": 10085
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.305631845941495e-06,
+      "loss": 1.0933,
+      "step": 10086
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.304960013641358e-06,
+      "loss": 0.9259,
+      "step": 10087
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.304288156057917e-06,
+      "loss": 0.9101,
+      "step": 10088
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.303616273204189e-06,
+      "loss": 0.9374,
+      "step": 10089
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.302944365093192e-06,
+      "loss": 1.0227,
+      "step": 10090
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.302272431737944e-06,
+      "loss": 0.9167,
+      "step": 10091
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.301600473151462e-06,
+      "loss": 1.0486,
+      "step": 10092
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3009284893467655e-06,
+      "loss": 1.0202,
+      "step": 10093
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.300256480336876e-06,
+      "loss": 1.0598,
+      "step": 10094
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.299584446134812e-06,
+      "loss": 1.0325,
+      "step": 10095
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.298912386753593e-06,
+      "loss": 0.8233,
+      "step": 10096
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.298240302206242e-06,
+      "loss": 0.8296,
+      "step": 10097
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.29756819250578e-06,
+      "loss": 0.9836,
+      "step": 10098
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.296896057665227e-06,
+      "loss": 0.8457,
+      "step": 10099
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.296223897697608e-06,
+      "loss": 1.1175,
+      "step": 10100
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.295551712615945e-06,
+      "loss": 0.9252,
+      "step": 10101
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.2948795024332625e-06,
+      "loss": 0.94,
+      "step": 10102
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.294207267162582e-06,
+      "loss": 0.9466,
+      "step": 10103
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.293535006816929e-06,
+      "loss": 1.0862,
+      "step": 10104
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.292862721409329e-06,
+      "loss": 1.0135,
+      "step": 10105
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.292190410952807e-06,
+      "loss": 1.0611,
+      "step": 10106
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.29151807546039e-06,
+      "loss": 1.0813,
+      "step": 10107
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.290845714945102e-06,
+      "loss": 1.0038,
+      "step": 10108
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.290173329419972e-06,
+      "loss": 0.8749,
+      "step": 10109
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.289500918898025e-06,
+      "loss": 0.8813,
+      "step": 10110
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.2888284833922905e-06,
+      "loss": 0.8522,
+      "step": 10111
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.288156022915798e-06,
+      "loss": 0.87,
+      "step": 10112
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.287483537481573e-06,
+      "loss": 0.9954,
+      "step": 10113
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.2868110271026485e-06,
+      "loss": 0.9604,
+      "step": 10114
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.286138491792051e-06,
+      "loss": 0.9241,
+      "step": 10115
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.285465931562811e-06,
+      "loss": 1.067,
+      "step": 10116
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.284793346427962e-06,
+      "loss": 1.1157,
+      "step": 10117
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.284120736400533e-06,
+      "loss": 1.0491,
+      "step": 10118
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.283448101493556e-06,
+      "loss": 1.1925,
+      "step": 10119
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.282775441720062e-06,
+      "loss": 0.8892,
+      "step": 10120
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.282102757093086e-06,
+      "loss": 0.9367,
+      "step": 10121
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.281430047625661e-06,
+      "loss": 0.9358,
+      "step": 10122
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.280757313330818e-06,
+      "loss": 0.8549,
+      "step": 10123
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.280084554221591e-06,
+      "loss": 0.9292,
+      "step": 10124
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.279411770311019e-06,
+      "loss": 0.9629,
+      "step": 10125
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.278738961612133e-06,
+      "loss": 0.9476,
+      "step": 10126
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.27806612813797e-06,
+      "loss": 1.0458,
+      "step": 10127
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2773932699015635e-06,
+      "loss": 0.758,
+      "step": 10128
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.276720386915956e-06,
+      "loss": 0.9312,
+      "step": 10129
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.276047479194177e-06,
+      "loss": 0.9321,
+      "step": 10130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.27537454674927e-06,
+      "loss": 1.0147,
+      "step": 10131
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.27470158959427e-06,
+      "loss": 0.9028,
+      "step": 10132
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.274028607742215e-06,
+      "loss": 0.9246,
+      "step": 10133
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.273355601206143e-06,
+      "loss": 1.129,
+      "step": 10134
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.272682569999099e-06,
+      "loss": 0.9692,
+      "step": 10135
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2720095141341165e-06,
+      "loss": 0.8735,
+      "step": 10136
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.27133643362424e-06,
+      "loss": 0.98,
+      "step": 10137
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.270663328482507e-06,
+      "loss": 0.8572,
+      "step": 10138
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.26999019872196e-06,
+      "loss": 0.9421,
+      "step": 10139
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.26931704435564e-06,
+      "loss": 0.9916,
+      "step": 10140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.268643865396593e-06,
+      "loss": 0.7116,
+      "step": 10141
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.267970661857858e-06,
+      "loss": 0.8499,
+      "step": 10142
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.267297433752478e-06,
+      "loss": 0.7435,
+      "step": 10143
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2666241810934995e-06,
+      "loss": 0.8195,
+      "step": 10144
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.265950903893963e-06,
+      "loss": 0.9883,
+      "step": 10145
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2652776021669144e-06,
+      "loss": 0.8362,
+      "step": 10146
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.264604275925401e-06,
+      "loss": 0.8064,
+      "step": 10147
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.263930925182467e-06,
+      "loss": 0.8997,
+      "step": 10148
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.263257549951157e-06,
+      "loss": 0.9646,
+      "step": 10149
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.262584150244518e-06,
+      "loss": 1.0039,
+      "step": 10150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.261910726075597e-06,
+      "loss": 0.8736,
+      "step": 10151
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.261237277457443e-06,
+      "loss": 1.0126,
+      "step": 10152
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.260563804403103e-06,
+      "loss": 0.9212,
+      "step": 10153
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.259890306925627e-06,
+      "loss": 1.0235,
+      "step": 10154
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.25921678503806e-06,
+      "loss": 1.1759,
+      "step": 10155
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.258543238753454e-06,
+      "loss": 0.9394,
+      "step": 10156
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.257869668084857e-06,
+      "loss": 1.065,
+      "step": 10157
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.257196073045322e-06,
+      "loss": 0.9626,
+      "step": 10158
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.256522453647897e-06,
+      "loss": 0.9074,
+      "step": 10159
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2558488099056356e-06,
+      "loss": 0.9347,
+      "step": 10160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.255175141831588e-06,
+      "loss": 0.896,
+      "step": 10161
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.254501449438808e-06,
+      "loss": 0.887,
+      "step": 10162
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2538277327403445e-06,
+      "loss": 1.0192,
+      "step": 10163
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.253153991749254e-06,
+      "loss": 1.0337,
+      "step": 10164
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.25248022647859e-06,
+      "loss": 1.078,
+      "step": 10165
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.251806436941405e-06,
+      "loss": 1.0148,
+      "step": 10166
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.251132623150754e-06,
+      "loss": 1.022,
+      "step": 10167
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.250458785119693e-06,
+      "loss": 1.1102,
+      "step": 10168
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.249784922861275e-06,
+      "loss": 1.0329,
+      "step": 10169
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.24911103638856e-06,
+      "loss": 0.9515,
+      "step": 10170
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2484371257146e-06,
+      "loss": 0.8925,
+      "step": 10171
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.247763190852456e-06,
+      "loss": 0.9283,
+      "step": 10172
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.247089231815182e-06,
+      "loss": 0.848,
+      "step": 10173
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.246415248615836e-06,
+      "loss": 0.9759,
+      "step": 10174
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.245741241267479e-06,
+      "loss": 0.8669,
+      "step": 10175
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.245067209783166e-06,
+      "loss": 0.8029,
+      "step": 10176
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.24439315417596e-06,
+      "loss": 0.9598,
+      "step": 10177
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2437190744589195e-06,
+      "loss": 0.9351,
+      "step": 10178
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.243044970645103e-06,
+      "loss": 1.0506,
+      "step": 10179
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.242370842747571e-06,
+      "loss": 0.9551,
+      "step": 10180
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2416966907793876e-06,
+      "loss": 0.8917,
+      "step": 10181
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.241022514753612e-06,
+      "loss": 1.0668,
+      "step": 10182
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2403483146833065e-06,
+      "loss": 0.8943,
+      "step": 10183
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.239674090581534e-06,
+      "loss": 0.9322,
+      "step": 10184
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.238999842461358e-06,
+      "loss": 0.8452,
+      "step": 10185
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.23832557033584e-06,
+      "loss": 0.9924,
+      "step": 10186
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.237651274218045e-06,
+      "loss": 0.9854,
+      "step": 10187
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.236976954121038e-06,
+      "loss": 1.0419,
+      "step": 10188
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.236302610057883e-06,
+      "loss": 1.0135,
+      "step": 10189
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.235628242041645e-06,
+      "loss": 0.9166,
+      "step": 10190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.23495385008539e-06,
+      "loss": 0.911,
+      "step": 10191
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.234279434202183e-06,
+      "loss": 0.8962,
+      "step": 10192
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2336049944050935e-06,
+      "loss": 1.008,
+      "step": 10193
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.232930530707187e-06,
+      "loss": 0.9965,
+      "step": 10194
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.232256043121531e-06,
+      "loss": 0.9422,
+      "step": 10195
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.231581531661194e-06,
+      "loss": 0.7182,
+      "step": 10196
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.230906996339243e-06,
+      "loss": 1.0256,
+      "step": 10197
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.230232437168748e-06,
+      "loss": 1.1137,
+      "step": 10198
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.22955785416278e-06,
+      "loss": 0.8523,
+      "step": 10199
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2288832473344074e-06,
+      "loss": 0.8848,
+      "step": 10200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2282086166967005e-06,
+      "loss": 1.0717,
+      "step": 10201
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.22753396226273e-06,
+      "loss": 0.93,
+      "step": 10202
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.226859284045567e-06,
+      "loss": 0.9697,
+      "step": 10203
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.226184582058284e-06,
+      "loss": 0.9063,
+      "step": 10204
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.225509856313954e-06,
+      "loss": 0.9686,
+      "step": 10205
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.224835106825646e-06,
+      "loss": 0.9814,
+      "step": 10206
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.224160333606438e-06,
+      "loss": 0.9512,
+      "step": 10207
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.223485536669401e-06,
+      "loss": 0.8263,
+      "step": 10208
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.222810716027608e-06,
+      "loss": 0.8651,
+      "step": 10209
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2221358716941345e-06,
+      "loss": 0.9454,
+      "step": 10210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.221461003682058e-06,
+      "loss": 0.9693,
+      "step": 10211
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.22078611200445e-06,
+      "loss": 0.9887,
+      "step": 10212
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.220111196674388e-06,
+      "loss": 0.7682,
+      "step": 10213
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.21943625770495e-06,
+      "loss": 0.8644,
+      "step": 10214
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.218761295109209e-06,
+      "loss": 0.7713,
+      "step": 10215
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.218086308900245e-06,
+      "loss": 0.8076,
+      "step": 10216
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.217411299091135e-06,
+      "loss": 0.9668,
+      "step": 10217
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.216736265694958e-06,
+      "loss": 1.0829,
+      "step": 10218
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.21606120872479e-06,
+      "loss": 0.8031,
+      "step": 10219
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.215386128193714e-06,
+      "loss": 1.0786,
+      "step": 10220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.214711024114807e-06,
+      "loss": 0.787,
+      "step": 10221
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.214035896501149e-06,
+      "loss": 0.908,
+      "step": 10222
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.213360745365821e-06,
+      "loss": 0.9038,
+      "step": 10223
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.212685570721904e-06,
+      "loss": 1.0452,
+      "step": 10224
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.212010372582479e-06,
+      "loss": 0.8909,
+      "step": 10225
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.211335150960628e-06,
+      "loss": 0.8153,
+      "step": 10226
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.210659905869431e-06,
+      "loss": 1.014,
+      "step": 10227
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.209984637321977e-06,
+      "loss": 1.051,
+      "step": 10228
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.209309345331343e-06,
+      "loss": 0.9735,
+      "step": 10229
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2086340299106144e-06,
+      "loss": 0.974,
+      "step": 10230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.207958691072876e-06,
+      "loss": 1.1423,
+      "step": 10231
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2072833288312116e-06,
+      "loss": 1.1452,
+      "step": 10232
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.206607943198707e-06,
+      "loss": 0.953,
+      "step": 10233
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.205932534188448e-06,
+      "loss": 0.8796,
+      "step": 10234
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.205257101813519e-06,
+      "loss": 1.0505,
+      "step": 10235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.204581646087007e-06,
+      "loss": 0.8249,
+      "step": 10236
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.203906167021999e-06,
+      "loss": 0.9068,
+      "step": 10237
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.203230664631581e-06,
+      "loss": 0.9514,
+      "step": 10238
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.202555138928843e-06,
+      "loss": 1.1107,
+      "step": 10239
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.201879589926873e-06,
+      "loss": 0.7848,
+      "step": 10240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2012040176387565e-06,
+      "loss": 0.9439,
+      "step": 10241
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.200528422077587e-06,
+      "loss": 0.8205,
+      "step": 10242
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.19985280325645e-06,
+      "loss": 0.8231,
+      "step": 10243
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.199177161188436e-06,
+      "loss": 0.9733,
+      "step": 10244
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1985014958866386e-06,
+      "loss": 0.8959,
+      "step": 10245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.197825807364145e-06,
+      "loss": 1.0727,
+      "step": 10246
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1971500956340495e-06,
+      "loss": 0.8938,
+      "step": 10247
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.196474360709441e-06,
+      "loss": 0.7972,
+      "step": 10248
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.195798602603414e-06,
+      "loss": 0.8465,
+      "step": 10249
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.19512282132906e-06,
+      "loss": 0.9164,
+      "step": 10250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.194447016899474e-06,
+      "loss": 1.0076,
+      "step": 10251
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.193771189327746e-06,
+      "loss": 0.9404,
+      "step": 10252
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.193095338626973e-06,
+      "loss": 0.9991,
+      "step": 10253
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1924194648102485e-06,
+      "loss": 0.8663,
+      "step": 10254
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1917435678906675e-06,
+      "loss": 1.0649,
+      "step": 10255
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.191067647881323e-06,
+      "loss": 1.0661,
+      "step": 10256
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.190391704795315e-06,
+      "loss": 0.863,
+      "step": 10257
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1897157386457385e-06,
+      "loss": 0.8745,
+      "step": 10258
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.189039749445689e-06,
+      "loss": 0.8949,
+      "step": 10259
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1883637372082625e-06,
+      "loss": 0.9036,
+      "step": 10260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1876877019465596e-06,
+      "loss": 1.0049,
+      "step": 10261
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.187011643673676e-06,
+      "loss": 0.8305,
+      "step": 10262
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.186335562402712e-06,
+      "loss": 0.9318,
+      "step": 10263
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.185659458146765e-06,
+      "loss": 0.886,
+      "step": 10264
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.184983330918935e-06,
+      "loss": 0.8922,
+      "step": 10265
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.184307180732321e-06,
+      "loss": 0.9923,
+      "step": 10266
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.183631007600025e-06,
+      "loss": 0.9405,
+      "step": 10267
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1829548115351465e-06,
+      "loss": 0.9558,
+      "step": 10268
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.182278592550788e-06,
+      "loss": 0.9426,
+      "step": 10269
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.181602350660048e-06,
+      "loss": 1.1759,
+      "step": 10270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.180926085876032e-06,
+      "loss": 0.8099,
+      "step": 10271
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1802497982118405e-06,
+      "loss": 0.9027,
+      "step": 10272
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.179573487680576e-06,
+      "loss": 1.1465,
+      "step": 10273
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.178897154295343e-06,
+      "loss": 1.0462,
+      "step": 10274
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.178220798069246e-06,
+      "loss": 0.8506,
+      "step": 10275
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.177544419015388e-06,
+      "loss": 0.7396,
+      "step": 10276
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.176868017146874e-06,
+      "loss": 1.0157,
+      "step": 10277
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.176191592476809e-06,
+      "loss": 0.9986,
+      "step": 10278
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.175515145018299e-06,
+      "loss": 1.1847,
+      "step": 10279
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.174838674784449e-06,
+      "loss": 0.9999,
+      "step": 10280
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.174162181788367e-06,
+      "loss": 1.0457,
+      "step": 10281
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.173485666043159e-06,
+      "loss": 0.973,
+      "step": 10282
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.172809127561932e-06,
+      "loss": 0.8094,
+      "step": 10283
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.172132566357795e-06,
+      "loss": 0.9468,
+      "step": 10284
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.171455982443853e-06,
+      "loss": 1.1,
+      "step": 10285
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.170779375833219e-06,
+      "loss": 0.9029,
+      "step": 10286
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.170102746538999e-06,
+      "loss": 1.0386,
+      "step": 10287
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.169426094574306e-06,
+      "loss": 0.9824,
+      "step": 10288
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.168749419952246e-06,
+      "loss": 0.8228,
+      "step": 10289
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.168072722685932e-06,
+      "loss": 0.8595,
+      "step": 10290
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.167396002788472e-06,
+      "loss": 0.9653,
+      "step": 10291
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.166719260272981e-06,
+      "loss": 1.0781,
+      "step": 10292
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.166042495152569e-06,
+      "loss": 0.9901,
+      "step": 10293
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1653657074403475e-06,
+      "loss": 0.903,
+      "step": 10294
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1646888971494305e-06,
+      "loss": 0.7626,
+      "step": 10295
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.164012064292928e-06,
+      "loss": 0.9675,
+      "step": 10296
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.163335208883958e-06,
+      "loss": 1.0229,
+      "step": 10297
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.162658330935632e-06,
+      "loss": 0.8481,
+      "step": 10298
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.161981430461065e-06,
+      "loss": 0.9386,
+      "step": 10299
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1613045074733704e-06,
+      "loss": 0.8416,
+      "step": 10300
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1606275619856646e-06,
+      "loss": 0.8478,
+      "step": 10301
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.159950594011063e-06,
+      "loss": 0.9837,
+      "step": 10302
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.159273603562682e-06,
+      "loss": 1.1524,
+      "step": 10303
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.158596590653638e-06,
+      "loss": 1.0369,
+      "step": 10304
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1579195552970475e-06,
+      "loss": 0.9962,
+      "step": 10305
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.157242497506029e-06,
+      "loss": 0.8705,
+      "step": 10306
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.156565417293699e-06,
+      "loss": 0.7791,
+      "step": 10307
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.155888314673177e-06,
+      "loss": 0.77,
+      "step": 10308
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.15521118965758e-06,
+      "loss": 0.8621,
+      "step": 10309
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1545340422600295e-06,
+      "loss": 1.2714,
+      "step": 10310
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.153856872493643e-06,
+      "loss": 0.9597,
+      "step": 10311
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.153179680371542e-06,
+      "loss": 0.8837,
+      "step": 10312
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.152502465906847e-06,
+      "loss": 0.9485,
+      "step": 10313
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.151825229112676e-06,
+      "loss": 1.0408,
+      "step": 10314
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.151147970002153e-06,
+      "loss": 0.9682,
+      "step": 10315
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.150470688588399e-06,
+      "loss": 0.8585,
+      "step": 10316
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.149793384884538e-06,
+      "loss": 0.9252,
+      "step": 10317
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.14911605890369e-06,
+      "loss": 0.9182,
+      "step": 10318
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.148438710658979e-06,
+      "loss": 1.0158,
+      "step": 10319
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.147761340163527e-06,
+      "loss": 0.9659,
+      "step": 10320
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1470839474304624e-06,
+      "loss": 0.7989,
+      "step": 10321
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.146406532472905e-06,
+      "loss": 0.9232,
+      "step": 10322
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.14572909530398e-06,
+      "loss": 1.0086,
+      "step": 10323
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1450516359368166e-06,
+      "loss": 1.0814,
+      "step": 10324
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1443741543845335e-06,
+      "loss": 0.9087,
+      "step": 10325
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1436966506602635e-06,
+      "loss": 0.9882,
+      "step": 10326
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.14301912477713e-06,
+      "loss": 0.901,
+      "step": 10327
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.14234157674826e-06,
+      "loss": 1.0503,
+      "step": 10328
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.141664006586781e-06,
+      "loss": 0.7806,
+      "step": 10329
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1409864143058215e-06,
+      "loss": 0.9629,
+      "step": 10330
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.140308799918507e-06,
+      "loss": 1.0609,
+      "step": 10331
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.13963116343797e-06,
+      "loss": 1.0126,
+      "step": 10332
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.138953504877337e-06,
+      "loss": 1.0651,
+      "step": 10333
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.13827582424974e-06,
+      "loss": 0.9184,
+      "step": 10334
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1375981215683055e-06,
+      "loss": 1.1479,
+      "step": 10335
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.136920396846165e-06,
+      "loss": 1.0009,
+      "step": 10336
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.136242650096451e-06,
+      "loss": 0.9911,
+      "step": 10337
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.135564881332292e-06,
+      "loss": 0.8349,
+      "step": 10338
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1348870905668225e-06,
+      "loss": 0.8778,
+      "step": 10339
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.134209277813172e-06,
+      "loss": 1.1089,
+      "step": 10340
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.133531443084475e-06,
+      "loss": 1.0497,
+      "step": 10341
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.132853586393863e-06,
+      "loss": 0.9919,
+      "step": 10342
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.132175707754469e-06,
+      "loss": 0.9645,
+      "step": 10343
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.131497807179428e-06,
+      "loss": 1.0774,
+      "step": 10344
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.130819884681874e-06,
+      "loss": 0.875,
+      "step": 10345
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.130141940274939e-06,
+      "loss": 0.8816,
+      "step": 10346
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.129463973971763e-06,
+      "loss": 0.926,
+      "step": 10347
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.128785985785478e-06,
+      "loss": 0.7446,
+      "step": 10348
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.12810797572922e-06,
+      "loss": 0.9352,
+      "step": 10349
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1274299438161254e-06,
+      "loss": 0.9774,
+      "step": 10350
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.126751890059332e-06,
+      "loss": 0.8854,
+      "step": 10351
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1260738144719775e-06,
+      "loss": 1.071,
+      "step": 10352
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.125395717067197e-06,
+      "loss": 0.8601,
+      "step": 10353
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.124717597858131e-06,
+      "loss": 1.001,
+      "step": 10354
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.124039456857916e-06,
+      "loss": 0.9184,
+      "step": 10355
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.123361294079691e-06,
+      "loss": 0.9993,
+      "step": 10356
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.122683109536596e-06,
+      "loss": 0.8585,
+      "step": 10357
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.122004903241769e-06,
+      "loss": 0.9477,
+      "step": 10358
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1213266752083545e-06,
+      "loss": 1.017,
+      "step": 10359
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.120648425449487e-06,
+      "loss": 1.046,
+      "step": 10360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1199701539783115e-06,
+      "loss": 1.0403,
+      "step": 10361
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.11929186080797e-06,
+      "loss": 0.8366,
+      "step": 10362
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.118613545951601e-06,
+      "loss": 0.9988,
+      "step": 10363
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.117935209422349e-06,
+      "loss": 0.9364,
+      "step": 10364
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.117256851233355e-06,
+      "loss": 0.9229,
+      "step": 10365
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.116578471397763e-06,
+      "loss": 0.9628,
+      "step": 10366
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.115900069928718e-06,
+      "loss": 0.8671,
+      "step": 10367
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.115221646839361e-06,
+      "loss": 0.9652,
+      "step": 10368
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.114543202142837e-06,
+      "loss": 0.9143,
+      "step": 10369
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.113864735852293e-06,
+      "loss": 0.8876,
+      "step": 10370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.113186247980871e-06,
+      "loss": 0.8988,
+      "step": 10371
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1125077385417185e-06,
+      "loss": 1.0667,
+      "step": 10372
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1118292075479814e-06,
+      "loss": 0.9018,
+      "step": 10373
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.111150655012805e-06,
+      "loss": 0.9269,
+      "step": 10374
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1104720809493365e-06,
+      "loss": 0.9373,
+      "step": 10375
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.109793485370724e-06,
+      "loss": 0.9219,
+      "step": 10376
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1091148682901145e-06,
+      "loss": 0.9209,
+      "step": 10377
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.108436229720655e-06,
+      "loss": 0.8794,
+      "step": 10378
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.107757569675495e-06,
+      "loss": 0.9885,
+      "step": 10379
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1070788881677845e-06,
+      "loss": 0.9815,
+      "step": 10380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1064001852106705e-06,
+      "loss": 1.1109,
+      "step": 10381
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.105721460817304e-06,
+      "loss": 0.8819,
+      "step": 10382
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.105042715000836e-06,
+      "loss": 0.994,
+      "step": 10383
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.104363947774415e-06,
+      "loss": 0.9363,
+      "step": 10384
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1036851591511934e-06,
+      "loss": 0.9787,
+      "step": 10385
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1030063491443225e-06,
+      "loss": 0.9576,
+      "step": 10386
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.102327517766954e-06,
+      "loss": 1.031,
+      "step": 10387
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.101648665032239e-06,
+      "loss": 0.9906,
+      "step": 10388
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.100969790953332e-06,
+      "loss": 1.0183,
+      "step": 10389
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.100290895543384e-06,
+      "loss": 0.9387,
+      "step": 10390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0996119788155505e-06,
+      "loss": 1.089,
+      "step": 10391
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0989330407829834e-06,
+      "loss": 1.0771,
+      "step": 10392
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.098254081458839e-06,
+      "loss": 0.9187,
+      "step": 10393
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0975751008562704e-06,
+      "loss": 1.0102,
+      "step": 10394
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.096896098988432e-06,
+      "loss": 0.9366,
+      "step": 10395
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.096217075868483e-06,
+      "loss": 0.9499,
+      "step": 10396
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.095538031509575e-06,
+      "loss": 0.9825,
+      "step": 10397
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.094858965924866e-06,
+      "loss": 0.8725,
+      "step": 10398
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0941798791275144e-06,
+      "loss": 0.7528,
+      "step": 10399
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.093500771130675e-06,
+      "loss": 0.8637,
+      "step": 10400
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.092821641947506e-06,
+      "loss": 0.9035,
+      "step": 10401
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.092142491591165e-06,
+      "loss": 0.9376,
+      "step": 10402
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.091463320074813e-06,
+      "loss": 1.087,
+      "step": 10403
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.090784127411605e-06,
+      "loss": 0.9666,
+      "step": 10404
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.090104913614703e-06,
+      "loss": 0.8869,
+      "step": 10405
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.089425678697265e-06,
+      "loss": 1.0352,
+      "step": 10406
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.08874642267245e-06,
+      "loss": 1.009,
+      "step": 10407
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0880671455534225e-06,
+      "loss": 0.9823,
+      "step": 10408
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.08738784735334e-06,
+      "loss": 1.2263,
+      "step": 10409
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.086708528085364e-06,
+      "loss": 1.0322,
+      "step": 10410
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.086029187762657e-06,
+      "loss": 0.97,
+      "step": 10411
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0853498263983826e-06,
+      "loss": 0.7898,
+      "step": 10412
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.084670444005699e-06,
+      "loss": 1.0405,
+      "step": 10413
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.083991040597773e-06,
+      "loss": 1.0228,
+      "step": 10414
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.083311616187766e-06,
+      "loss": 0.8332,
+      "step": 10415
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.082632170788842e-06,
+      "loss": 0.9254,
+      "step": 10416
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.081952704414167e-06,
+      "loss": 0.7813,
+      "step": 10417
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.081273217076901e-06,
+      "loss": 0.9753,
+      "step": 10418
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0805937087902135e-06,
+      "loss": 1.0,
+      "step": 10419
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.079914179567268e-06,
+      "loss": 1.1411,
+      "step": 10420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.079234629421229e-06,
+      "loss": 0.929,
+      "step": 10421
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.078555058365265e-06,
+      "loss": 0.9456,
+      "step": 10422
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.077875466412542e-06,
+      "loss": 1.0144,
+      "step": 10423
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.077195853576224e-06,
+      "loss": 0.8435,
+      "step": 10424
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0765162198694825e-06,
+      "loss": 1.0495,
+      "step": 10425
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.075836565305483e-06,
+      "loss": 1.0465,
+      "step": 10426
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.075156889897395e-06,
+      "loss": 0.8918,
+      "step": 10427
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.074477193658385e-06,
+      "loss": 1.0636,
+      "step": 10428
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.073797476601624e-06,
+      "loss": 0.8864,
+      "step": 10429
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.07311773874028e-06,
+      "loss": 0.7618,
+      "step": 10430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.072437980087522e-06,
+      "loss": 0.9008,
+      "step": 10431
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.071758200656523e-06,
+      "loss": 0.9139,
+      "step": 10432
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.071078400460452e-06,
+      "loss": 0.8926,
+      "step": 10433
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.07039857951248e-06,
+      "loss": 1.0064,
+      "step": 10434
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.069718737825777e-06,
+      "loss": 1.0685,
+      "step": 10435
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.069038875413517e-06,
+      "loss": 0.8625,
+      "step": 10436
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.068358992288871e-06,
+      "loss": 1.114,
+      "step": 10437
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.067679088465012e-06,
+      "loss": 0.8856,
+      "step": 10438
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.066999163955113e-06,
+      "loss": 1.0939,
+      "step": 10439
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.066319218772347e-06,
+      "loss": 1.0413,
+      "step": 10440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.065639252929889e-06,
+      "loss": 0.9033,
+      "step": 10441
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.064959266440909e-06,
+      "loss": 0.9049,
+      "step": 10442
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0642792593185885e-06,
+      "loss": 1.0233,
+      "step": 10443
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.063599231576097e-06,
+      "loss": 1.0151,
+      "step": 10444
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0629191832266125e-06,
+      "loss": 0.9273,
+      "step": 10445
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.062239114283309e-06,
+      "loss": 1.0662,
+      "step": 10446
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0615590247593635e-06,
+      "loss": 1.0007,
+      "step": 10447
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0608789146679535e-06,
+      "loss": 0.8082,
+      "step": 10448
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.060198784022256e-06,
+      "loss": 0.9304,
+      "step": 10449
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.059518632835446e-06,
+      "loss": 0.9468,
+      "step": 10450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.058838461120704e-06,
+      "loss": 0.9534,
+      "step": 10451
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.058158268891205e-06,
+      "loss": 1.0399,
+      "step": 10452
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.057478056160131e-06,
+      "loss": 0.9395,
+      "step": 10453
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.05679782294066e-06,
+      "loss": 0.9877,
+      "step": 10454
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.056117569245971e-06,
+      "loss": 0.9745,
+      "step": 10455
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0554372950892425e-06,
+      "loss": 0.8866,
+      "step": 10456
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0547570004836566e-06,
+      "loss": 0.79,
+      "step": 10457
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.054076685442393e-06,
+      "loss": 1.022,
+      "step": 10458
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.053396349978632e-06,
+      "loss": 0.8347,
+      "step": 10459
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0527159941055554e-06,
+      "loss": 0.8254,
+      "step": 10460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.052035617836347e-06,
+      "loss": 0.9613,
+      "step": 10461
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.051355221184187e-06,
+      "loss": 1.0659,
+      "step": 10462
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.050674804162257e-06,
+      "loss": 1.0408,
+      "step": 10463
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.04999436678374e-06,
+      "loss": 1.054,
+      "step": 10464
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.049313909061823e-06,
+      "loss": 0.9523,
+      "step": 10465
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.048633431009685e-06,
+      "loss": 1.0656,
+      "step": 10466
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.047952932640513e-06,
+      "loss": 0.7765,
+      "step": 10467
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.047272413967491e-06,
+      "loss": 1.0214,
+      "step": 10468
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.046591875003803e-06,
+      "loss": 1.0167,
+      "step": 10469
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.045911315762635e-06,
+      "loss": 1.0708,
+      "step": 10470
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.045230736257171e-06,
+      "loss": 0.9944,
+      "step": 10471
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0445501365006e-06,
+      "loss": 0.929,
+      "step": 10472
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.043869516506107e-06,
+      "loss": 1.0204,
+      "step": 10473
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.04318887628688e-06,
+      "loss": 0.8292,
+      "step": 10474
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.042508215856103e-06,
+      "loss": 0.9763,
+      "step": 10475
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0418275352269675e-06,
+      "loss": 0.9637,
+      "step": 10476
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.041146834412657e-06,
+      "loss": 0.9284,
+      "step": 10477
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.040466113426366e-06,
+      "loss": 0.9079,
+      "step": 10478
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.039785372281279e-06,
+      "loss": 0.9491,
+      "step": 10479
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.039104610990586e-06,
+      "loss": 0.9162,
+      "step": 10480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.038423829567478e-06,
+      "loss": 0.7937,
+      "step": 10481
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.037743028025141e-06,
+      "loss": 0.9228,
+      "step": 10482
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.037062206376772e-06,
+      "loss": 0.9495,
+      "step": 10483
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.036381364635555e-06,
+      "loss": 0.8773,
+      "step": 10484
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.035700502814686e-06,
+      "loss": 0.9462,
+      "step": 10485
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.035019620927354e-06,
+      "loss": 0.8524,
+      "step": 10486
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.034338718986752e-06,
+      "loss": 0.9206,
+      "step": 10487
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.03365779700607e-06,
+      "loss": 0.9425,
+      "step": 10488
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.032976854998504e-06,
+      "loss": 1.0053,
+      "step": 10489
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.032295892977247e-06,
+      "loss": 0.9029,
+      "step": 10490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.03161491095549e-06,
+      "loss": 0.8881,
+      "step": 10491
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.030933908946429e-06,
+      "loss": 1.0503,
+      "step": 10492
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.030252886963257e-06,
+      "loss": 0.9205,
+      "step": 10493
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0295718450191684e-06,
+      "loss": 0.9949,
+      "step": 10494
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.028890783127359e-06,
+      "loss": 0.9732,
+      "step": 10495
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.028209701301024e-06,
+      "loss": 1.0516,
+      "step": 10496
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.02752859955336e-06,
+      "loss": 0.8987,
+      "step": 10497
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0268474778975625e-06,
+      "loss": 1.0805,
+      "step": 10498
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0261663363468285e-06,
+      "loss": 0.8461,
+      "step": 10499
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.025485174914352e-06,
+      "loss": 0.9353,
+      "step": 10500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.024803993613334e-06,
+      "loss": 0.7673,
+      "step": 10501
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.024122792456972e-06,
+      "loss": 0.9059,
+      "step": 10502
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0234415714584624e-06,
+      "loss": 1.2484,
+      "step": 10503
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.022760330631006e-06,
+      "loss": 0.936,
+      "step": 10504
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0220790699877985e-06,
+      "loss": 0.8603,
+      "step": 10505
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.021397789542041e-06,
+      "loss": 0.9736,
+      "step": 10506
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.020716489306932e-06,
+      "loss": 1.0428,
+      "step": 10507
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.020035169295675e-06,
+      "loss": 1.0585,
+      "step": 10508
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.019353829521467e-06,
+      "loss": 0.9019,
+      "step": 10509
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.01867246999751e-06,
+      "loss": 0.8958,
+      "step": 10510
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.017991090737003e-06,
+      "loss": 1.0124,
+      "step": 10511
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.017309691753152e-06,
+      "loss": 0.9341,
+      "step": 10512
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.016628273059156e-06,
+      "loss": 0.7732,
+      "step": 10513
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.015946834668217e-06,
+      "loss": 1.1308,
+      "step": 10514
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.01526537659354e-06,
+      "loss": 1.0482,
+      "step": 10515
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.014583898848325e-06,
+      "loss": 0.9222,
+      "step": 10516
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.013902401445777e-06,
+      "loss": 1.0582,
+      "step": 10517
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.013220884399101e-06,
+      "loss": 1.0491,
+      "step": 10518
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0125393477215e-06,
+      "loss": 0.7738,
+      "step": 10519
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0118577914261786e-06,
+      "loss": 0.9052,
+      "step": 10520
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.011176215526343e-06,
+      "loss": 0.8803,
+      "step": 10521
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.010494620035197e-06,
+      "loss": 0.8233,
+      "step": 10522
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.009813004965947e-06,
+      "loss": 0.8037,
+      "step": 10523
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.009131370331799e-06,
+      "loss": 1.0567,
+      "step": 10524
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0084497161459585e-06,
+      "loss": 0.923,
+      "step": 10525
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0077680424216354e-06,
+      "loss": 1.0575,
+      "step": 10526
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.007086349172034e-06,
+      "loss": 0.8231,
+      "step": 10527
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.006404636410364e-06,
+      "loss": 1.173,
+      "step": 10528
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.00572290414983e-06,
+      "loss": 1.0733,
+      "step": 10529
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0050411524036445e-06,
+      "loss": 1.0211,
+      "step": 10530
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0043593811850144e-06,
+      "loss": 0.8095,
+      "step": 10531
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.00367759050715e-06,
+      "loss": 0.9291,
+      "step": 10532
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.00299578038326e-06,
+      "loss": 0.7524,
+      "step": 10533
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.002313950826553e-06,
+      "loss": 0.9084,
+      "step": 10534
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0016321018502395e-06,
+      "loss": 0.8175,
+      "step": 10535
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.000950233467534e-06,
+      "loss": 0.8721,
+      "step": 10536
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.000268345691644e-06,
+      "loss": 1.0656,
+      "step": 10537
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.999586438535781e-06,
+      "loss": 1.0065,
+      "step": 10538
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.998904512013158e-06,
+      "loss": 0.8127,
+      "step": 10539
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.998222566136985e-06,
+      "loss": 1.1043,
+      "step": 10540
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.997540600920479e-06,
+      "loss": 1.0411,
+      "step": 10541
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.996858616376848e-06,
+      "loss": 0.9054,
+      "step": 10542
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.996176612519309e-06,
+      "loss": 0.7732,
+      "step": 10543
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.995494589361074e-06,
+      "loss": 1.005,
+      "step": 10544
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9948125469153565e-06,
+      "loss": 0.9999,
+      "step": 10545
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.994130485195371e-06,
+      "loss": 0.7326,
+      "step": 10546
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.993448404214332e-06,
+      "loss": 0.8657,
+      "step": 10547
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9927663039854575e-06,
+      "loss": 0.9593,
+      "step": 10548
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.992084184521961e-06,
+      "loss": 0.9656,
+      "step": 10549
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.991402045837058e-06,
+      "loss": 0.9571,
+      "step": 10550
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.990719887943966e-06,
+      "loss": 1.0106,
+      "step": 10551
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9900377108559006e-06,
+      "loss": 0.9993,
+      "step": 10552
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.989355514586078e-06,
+      "loss": 0.902,
+      "step": 10553
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.988673299147718e-06,
+      "loss": 1.0399,
+      "step": 10554
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9879910645540365e-06,
+      "loss": 0.9951,
+      "step": 10555
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.987308810818253e-06,
+      "loss": 0.9241,
+      "step": 10556
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9866265379535825e-06,
+      "loss": 0.8673,
+      "step": 10557
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.985944245973249e-06,
+      "loss": 0.9828,
+      "step": 10558
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.985261934890469e-06,
+      "loss": 1.0403,
+      "step": 10559
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.984579604718463e-06,
+      "loss": 1.0154,
+      "step": 10560
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.98389725547045e-06,
+      "loss": 0.9515,
+      "step": 10561
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.983214887159652e-06,
+      "loss": 0.9896,
+      "step": 10562
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9825324997992875e-06,
+      "loss": 0.8602,
+      "step": 10563
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9818500934025765e-06,
+      "loss": 0.9623,
+      "step": 10564
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9811676679827456e-06,
+      "loss": 1.0019,
+      "step": 10565
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9804852235530134e-06,
+      "loss": 0.9202,
+      "step": 10566
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9798027601266015e-06,
+      "loss": 0.9566,
+      "step": 10567
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9791202777167345e-06,
+      "loss": 0.8562,
+      "step": 10568
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.978437776336634e-06,
+      "loss": 0.8129,
+      "step": 10569
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.977755255999522e-06,
+      "loss": 1.0076,
+      "step": 10570
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.977072716718625e-06,
+      "loss": 1.0428,
+      "step": 10571
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9763901585071645e-06,
+      "loss": 0.8434,
+      "step": 10572
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9757075813783684e-06,
+      "loss": 0.985,
+      "step": 10573
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.975024985345458e-06,
+      "loss": 0.992,
+      "step": 10574
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.974342370421659e-06,
+      "loss": 0.8941,
+      "step": 10575
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.973659736620199e-06,
+      "loss": 0.9997,
+      "step": 10576
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9729770839543014e-06,
+      "loss": 0.784,
+      "step": 10577
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.972294412437194e-06,
+      "loss": 0.9691,
+      "step": 10578
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9716117220821026e-06,
+      "loss": 0.8863,
+      "step": 10579
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.970929012902255e-06,
+      "loss": 1.0471,
+      "step": 10580
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.970246284910877e-06,
+      "loss": 0.8953,
+      "step": 10581
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.969563538121198e-06,
+      "loss": 0.89,
+      "step": 10582
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9688807725464445e-06,
+      "loss": 0.7976,
+      "step": 10583
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.968197988199848e-06,
+      "loss": 1.0692,
+      "step": 10584
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9675151850946334e-06,
+      "loss": 0.9852,
+      "step": 10585
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.966832363244033e-06,
+      "loss": 0.961,
+      "step": 10586
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.966149522661273e-06,
+      "loss": 0.8016,
+      "step": 10587
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.965466663359586e-06,
+      "loss": 0.9976,
+      "step": 10588
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.964783785352201e-06,
+      "loss": 0.9324,
+      "step": 10589
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.964100888652351e-06,
+      "loss": 0.9127,
+      "step": 10590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.963417973273263e-06,
+      "loss": 1.0087,
+      "step": 10591
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9627350392281715e-06,
+      "loss": 0.8668,
+      "step": 10592
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9620520865303046e-06,
+      "loss": 1.1366,
+      "step": 10593
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.961369115192899e-06,
+      "loss": 0.9394,
+      "step": 10594
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.960686125229183e-06,
+      "loss": 1.1639,
+      "step": 10595
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.960003116652393e-06,
+      "loss": 0.9492,
+      "step": 10596
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.959320089475761e-06,
+      "loss": 0.9674,
+      "step": 10597
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.958637043712519e-06,
+      "loss": 0.9254,
+      "step": 10598
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9579539793759e-06,
+      "loss": 0.9048,
+      "step": 10599
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.957270896479141e-06,
+      "loss": 1.0,
+      "step": 10600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.956587795035477e-06,
+      "loss": 1.0105,
+      "step": 10601
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.95590467505814e-06,
+      "loss": 1.1793,
+      "step": 10602
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.955221536560367e-06,
+      "loss": 0.9533,
+      "step": 10603
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.954538379555393e-06,
+      "loss": 1.124,
+      "step": 10604
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.953855204056454e-06,
+      "loss": 0.9638,
+      "step": 10605
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9531720100767885e-06,
+      "loss": 0.9255,
+      "step": 10606
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.952488797629631e-06,
+      "loss": 0.9756,
+      "step": 10607
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.951805566728219e-06,
+      "loss": 1.0313,
+      "step": 10608
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9511223173857905e-06,
+      "loss": 0.9696,
+      "step": 10609
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9504390496155815e-06,
+      "loss": 1.1063,
+      "step": 10610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.949755763430832e-06,
+      "loss": 0.9815,
+      "step": 10611
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.94907245884478e-06,
+      "loss": 0.9382,
+      "step": 10612
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.948389135870666e-06,
+      "loss": 0.9613,
+      "step": 10613
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.947705794521725e-06,
+      "loss": 1.0503,
+      "step": 10614
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.947022434811202e-06,
+      "loss": 0.8538,
+      "step": 10615
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9463390567523325e-06,
+      "loss": 0.9336,
+      "step": 10616
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.945655660358358e-06,
+      "loss": 1.1693,
+      "step": 10617
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.944972245642521e-06,
+      "loss": 1.0738,
+      "step": 10618
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.944288812618061e-06,
+      "loss": 0.8318,
+      "step": 10619
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.943605361298218e-06,
+      "loss": 0.9744,
+      "step": 10620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.942921891696236e-06,
+      "loss": 1.0172,
+      "step": 10621
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.942238403825356e-06,
+      "loss": 1.0379,
+      "step": 10622
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.94155489769882e-06,
+      "loss": 0.9989,
+      "step": 10623
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.940871373329873e-06,
+      "loss": 0.8362,
+      "step": 10624
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9401878307317564e-06,
+      "loss": 0.85,
+      "step": 10625
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.939504269917713e-06,
+      "loss": 0.9384,
+      "step": 10626
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.938820690900988e-06,
+      "loss": 0.9938,
+      "step": 10627
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.938137093694824e-06,
+      "loss": 1.0123,
+      "step": 10628
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.937453478312467e-06,
+      "loss": 0.98,
+      "step": 10629
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.936769844767163e-06,
+      "loss": 0.9508,
+      "step": 10630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9360861930721545e-06,
+      "loss": 0.8492,
+      "step": 10631
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9354025232406885e-06,
+      "loss": 1.0055,
+      "step": 10632
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.934718835286011e-06,
+      "loss": 0.9573,
+      "step": 10633
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.934035129221367e-06,
+      "loss": 1.1073,
+      "step": 10634
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.933351405060006e-06,
+      "loss": 0.907,
+      "step": 10635
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.932667662815172e-06,
+      "loss": 1.0404,
+      "step": 10636
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.931983902500115e-06,
+      "loss": 0.8189,
+      "step": 10637
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.93130012412808e-06,
+      "loss": 1.0207,
+      "step": 10638
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.930616327712317e-06,
+      "loss": 1.0488,
+      "step": 10639
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.929932513266072e-06,
+      "loss": 0.8441,
+      "step": 10640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.929248680802597e-06,
+      "loss": 0.9507,
+      "step": 10641
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9285648303351404e-06,
+      "loss": 0.9377,
+      "step": 10642
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9278809618769505e-06,
+      "loss": 0.8245,
+      "step": 10643
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9271970754412765e-06,
+      "loss": 0.8129,
+      "step": 10644
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.92651317104137e-06,
+      "loss": 0.8875,
+      "step": 10645
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.925829248690481e-06,
+      "loss": 1.0466,
+      "step": 10646
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.92514530840186e-06,
+      "loss": 0.9647,
+      "step": 10647
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.924461350188759e-06,
+      "loss": 0.9304,
+      "step": 10648
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.92377737406443e-06,
+      "loss": 1.0782,
+      "step": 10649
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.923093380042122e-06,
+      "loss": 1.0725,
+      "step": 10650
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.922409368135089e-06,
+      "loss": 0.9669,
+      "step": 10651
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.921725338356585e-06,
+      "loss": 0.8569,
+      "step": 10652
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.921041290719861e-06,
+      "loss": 0.828,
+      "step": 10653
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9203572252381715e-06,
+      "loss": 0.8339,
+      "step": 10654
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9196731419247695e-06,
+      "loss": 0.7573,
+      "step": 10655
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.918989040792909e-06,
+      "loss": 1.0614,
+      "step": 10656
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.918304921855843e-06,
+      "loss": 1.0088,
+      "step": 10657
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.917620785126829e-06,
+      "loss": 0.9122,
+      "step": 10658
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.91693663061912e-06,
+      "loss": 1.0449,
+      "step": 10659
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.916252458345971e-06,
+      "loss": 1.0614,
+      "step": 10660
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9155682683206395e-06,
+      "loss": 0.8581,
+      "step": 10661
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.91488406055638e-06,
+      "loss": 0.8361,
+      "step": 10662
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9141998350664485e-06,
+      "loss": 0.8416,
+      "step": 10663
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.913515591864104e-06,
+      "loss": 0.9214,
+      "step": 10664
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.912831330962602e-06,
+      "loss": 0.8816,
+      "step": 10665
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9121470523752e-06,
+      "loss": 0.923,
+      "step": 10666
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.911462756115155e-06,
+      "loss": 0.9588,
+      "step": 10667
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.910778442195726e-06,
+      "loss": 0.7264,
+      "step": 10668
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.910094110630171e-06,
+      "loss": 1.0021,
+      "step": 10669
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.90940976143175e-06,
+      "loss": 0.9235,
+      "step": 10670
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.908725394613721e-06,
+      "loss": 0.978,
+      "step": 10671
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9080410101893435e-06,
+      "loss": 1.0214,
+      "step": 10672
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.907356608171878e-06,
+      "loss": 0.8912,
+      "step": 10673
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.906672188574582e-06,
+      "loss": 0.9196,
+      "step": 10674
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.90598775141072e-06,
+      "loss": 0.9423,
+      "step": 10675
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.90530329669355e-06,
+      "loss": 1.0427,
+      "step": 10676
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9046188244363345e-06,
+      "loss": 0.8818,
+      "step": 10677
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.903934334652334e-06,
+      "loss": 0.9973,
+      "step": 10678
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.903249827354811e-06,
+      "loss": 1.0265,
+      "step": 10679
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.902565302557028e-06,
+      "loss": 0.8712,
+      "step": 10680
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.901880760272246e-06,
+      "loss": 0.944,
+      "step": 10681
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.90119620051373e-06,
+      "loss": 0.9284,
+      "step": 10682
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.900511623294741e-06,
+      "loss": 1.0846,
+      "step": 10683
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.899827028628544e-06,
+      "loss": 0.8242,
+      "step": 10684
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8991424165284026e-06,
+      "loss": 1.0243,
+      "step": 10685
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.89845778700758e-06,
+      "loss": 1.0153,
+      "step": 10686
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8977731400793425e-06,
+      "loss": 0.9551,
+      "step": 10687
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.897088475756953e-06,
+      "loss": 0.9953,
+      "step": 10688
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.896403794053679e-06,
+      "loss": 0.8647,
+      "step": 10689
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8957190949827845e-06,
+      "loss": 1.21,
+      "step": 10690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.895034378557536e-06,
+      "loss": 1.0807,
+      "step": 10691
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.894349644791199e-06,
+      "loss": 0.8801,
+      "step": 10692
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.89366489369704e-06,
+      "loss": 0.9786,
+      "step": 10693
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.892980125288327e-06,
+      "loss": 1.014,
+      "step": 10694
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.892295339578327e-06,
+      "loss": 0.8742,
+      "step": 10695
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.891610536580307e-06,
+      "loss": 0.892,
+      "step": 10696
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.890925716307533e-06,
+      "loss": 0.9409,
+      "step": 10697
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.890240878773278e-06,
+      "loss": 0.7703,
+      "step": 10698
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.889556023990806e-06,
+      "loss": 0.6992,
+      "step": 10699
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.88887115197339e-06,
+      "loss": 0.8813,
+      "step": 10700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8881862627342956e-06,
+      "loss": 0.8154,
+      "step": 10701
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8875013562867946e-06,
+      "loss": 0.7419,
+      "step": 10702
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.886816432644155e-06,
+      "loss": 0.8929,
+      "step": 10703
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.886131491819649e-06,
+      "loss": 1.0467,
+      "step": 10704
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.885446533826546e-06,
+      "loss": 1.1504,
+      "step": 10705
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.884761558678118e-06,
+      "loss": 0.7993,
+      "step": 10706
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.884076566387635e-06,
+      "loss": 1.0162,
+      "step": 10707
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.88339155696837e-06,
+      "loss": 0.8757,
+      "step": 10708
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.882706530433593e-06,
+      "loss": 0.9742,
+      "step": 10709
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.882021486796578e-06,
+      "loss": 0.7998,
+      "step": 10710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.881336426070595e-06,
+      "loss": 0.9379,
+      "step": 10711
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.880651348268921e-06,
+      "loss": 0.9603,
+      "step": 10712
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.879966253404825e-06,
+      "loss": 0.972,
+      "step": 10713
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.879281141491584e-06,
+      "loss": 0.8235,
+      "step": 10714
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.878596012542468e-06,
+      "loss": 0.8997,
+      "step": 10715
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.877910866570755e-06,
+      "loss": 0.8232,
+      "step": 10716
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.877225703589719e-06,
+      "loss": 0.927,
+      "step": 10717
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.876540523612633e-06,
+      "loss": 0.8506,
+      "step": 10718
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.875855326652773e-06,
+      "loss": 0.8957,
+      "step": 10719
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.875170112723415e-06,
+      "loss": 1.0512,
+      "step": 10720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.874484881837834e-06,
+      "loss": 0.9948,
+      "step": 10721
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.873799634009307e-06,
+      "loss": 1.0622,
+      "step": 10722
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.87311436925111e-06,
+      "loss": 1.0257,
+      "step": 10723
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.872429087576522e-06,
+      "loss": 0.9838,
+      "step": 10724
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.871743788998817e-06,
+      "loss": 0.9777,
+      "step": 10725
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.871058473531273e-06,
+      "loss": 0.8757,
+      "step": 10726
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.87037314118717e-06,
+      "loss": 1.0528,
+      "step": 10727
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.869687791979784e-06,
+      "loss": 1.0317,
+      "step": 10728
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.869002425922396e-06,
+      "loss": 1.0062,
+      "step": 10729
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.868317043028281e-06,
+      "loss": 1.0775,
+      "step": 10730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.867631643310721e-06,
+      "loss": 0.9838,
+      "step": 10731
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.866946226782995e-06,
+      "loss": 0.9562,
+      "step": 10732
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.866260793458382e-06,
+      "loss": 0.8815,
+      "step": 10733
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.865575343350163e-06,
+      "loss": 1.0673,
+      "step": 10734
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8648898764716174e-06,
+      "loss": 0.9758,
+      "step": 10735
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.864204392836028e-06,
+      "loss": 0.9948,
+      "step": 10736
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.863518892456675e-06,
+      "loss": 0.9723,
+      "step": 10737
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.862833375346837e-06,
+      "loss": 0.8959,
+      "step": 10738
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8621478415198e-06,
+      "loss": 1.0659,
+      "step": 10739
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.861462290988843e-06,
+      "loss": 0.8274,
+      "step": 10740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.860776723767251e-06,
+      "loss": 0.952,
+      "step": 10741
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.860091139868304e-06,
+      "loss": 0.9804,
+      "step": 10742
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.859405539305286e-06,
+      "loss": 1.0039,
+      "step": 10743
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.85871992209148e-06,
+      "loss": 0.9302,
+      "step": 10744
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.858034288240172e-06,
+      "loss": 0.9465,
+      "step": 10745
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.857348637764644e-06,
+      "loss": 0.8304,
+      "step": 10746
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.85666297067818e-06,
+      "loss": 0.9493,
+      "step": 10747
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8559772869940655e-06,
+      "loss": 0.8673,
+      "step": 10748
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.855291586725586e-06,
+      "loss": 1.0635,
+      "step": 10749
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.854605869886023e-06,
+      "loss": 0.9303,
+      "step": 10750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.853920136488668e-06,
+      "loss": 1.0512,
+      "step": 10751
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.853234386546803e-06,
+      "loss": 0.8813,
+      "step": 10752
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.852548620073717e-06,
+      "loss": 0.9411,
+      "step": 10753
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.851862837082694e-06,
+      "loss": 0.9224,
+      "step": 10754
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.851177037587019e-06,
+      "loss": 1.0495,
+      "step": 10755
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.850491221599985e-06,
+      "loss": 1.0027,
+      "step": 10756
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.849805389134876e-06,
+      "loss": 1.1092,
+      "step": 10757
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.849119540204981e-06,
+      "loss": 1.0902,
+      "step": 10758
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.848433674823586e-06,
+      "loss": 1.0273,
+      "step": 10759
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.847747793003983e-06,
+      "loss": 0.9229,
+      "step": 10760
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.847061894759456e-06,
+      "loss": 0.8859,
+      "step": 10761
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.846375980103298e-06,
+      "loss": 0.9663,
+      "step": 10762
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.845690049048799e-06,
+      "loss": 1.0415,
+      "step": 10763
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8450041016092465e-06,
+      "loss": 0.929,
+      "step": 10764
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.844318137797931e-06,
+      "loss": 1.0047,
+      "step": 10765
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.843632157628142e-06,
+      "loss": 1.1391,
+      "step": 10766
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8429461611131725e-06,
+      "loss": 0.7325,
+      "step": 10767
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.842260148266312e-06,
+      "loss": 0.929,
+      "step": 10768
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.841574119100853e-06,
+      "loss": 0.9212,
+      "step": 10769
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.840888073630085e-06,
+      "loss": 0.9779,
+      "step": 10770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.840202011867301e-06,
+      "loss": 1.125,
+      "step": 10771
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.839515933825794e-06,
+      "loss": 0.8164,
+      "step": 10772
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.838829839518857e-06,
+      "loss": 0.9493,
+      "step": 10773
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.838143728959782e-06,
+      "loss": 0.923,
+      "step": 10774
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.837457602161863e-06,
+      "loss": 0.9875,
+      "step": 10775
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.836771459138392e-06,
+      "loss": 0.8753,
+      "step": 10776
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8360852999026644e-06,
+      "loss": 1.037,
+      "step": 10777
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.835399124467973e-06,
+      "loss": 0.9747,
+      "step": 10778
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.834712932847612e-06,
+      "loss": 1.1162,
+      "step": 10779
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.834026725054879e-06,
+      "loss": 0.7946,
+      "step": 10780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.833340501103066e-06,
+      "loss": 0.8829,
+      "step": 10781
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.832654261005472e-06,
+      "loss": 0.9694,
+      "step": 10782
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.831968004775389e-06,
+      "loss": 0.9337,
+      "step": 10783
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.831281732426115e-06,
+      "loss": 0.8797,
+      "step": 10784
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.830595443970946e-06,
+      "loss": 0.8487,
+      "step": 10785
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.829909139423178e-06,
+      "loss": 0.7782,
+      "step": 10786
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.82922281879611e-06,
+      "loss": 0.9799,
+      "step": 10787
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8285364821030375e-06,
+      "loss": 0.9665,
+      "step": 10788
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8278501293572586e-06,
+      "loss": 0.8643,
+      "step": 10789
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.82716376057207e-06,
+      "loss": 0.9604,
+      "step": 10790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.826477375760772e-06,
+      "loss": 1.0267,
+      "step": 10791
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8257909749366624e-06,
+      "loss": 0.8299,
+      "step": 10792
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.825104558113041e-06,
+      "loss": 0.9256,
+      "step": 10793
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.824418125303205e-06,
+      "loss": 1.0236,
+      "step": 10794
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.823731676520455e-06,
+      "loss": 1.0155,
+      "step": 10795
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.82304521177809e-06,
+      "loss": 0.9509,
+      "step": 10796
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.822358731089412e-06,
+      "loss": 0.8803,
+      "step": 10797
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.821672234467719e-06,
+      "loss": 0.8734,
+      "step": 10798
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8209857219263136e-06,
+      "loss": 0.8156,
+      "step": 10799
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.820299193478496e-06,
+      "loss": 0.9846,
+      "step": 10800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.819612649137567e-06,
+      "loss": 0.9332,
+      "step": 10801
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8189260889168285e-06,
+      "loss": 0.8899,
+      "step": 10802
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.818239512829585e-06,
+      "loss": 1.04,
+      "step": 10803
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.817552920889134e-06,
+      "loss": 0.9274,
+      "step": 10804
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.816866313108782e-06,
+      "loss": 1.0578,
+      "step": 10805
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.81617968950183e-06,
+      "loss": 0.9858,
+      "step": 10806
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8154930500815824e-06,
+      "loss": 0.8565,
+      "step": 10807
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.81480639486134e-06,
+      "loss": 1.0571,
+      "step": 10808
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.81411972385441e-06,
+      "loss": 0.8974,
+      "step": 10809
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.813433037074095e-06,
+      "loss": 0.851,
+      "step": 10810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8127463345336985e-06,
+      "loss": 0.94,
+      "step": 10811
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.812059616246528e-06,
+      "loss": 0.96,
+      "step": 10812
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8113728822258855e-06,
+      "loss": 0.9714,
+      "step": 10813
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.810686132485076e-06,
+      "loss": 0.8257,
+      "step": 10814
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8099993670374086e-06,
+      "loss": 0.9175,
+      "step": 10815
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.809312585896188e-06,
+      "loss": 0.7683,
+      "step": 10816
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.808625789074718e-06,
+      "loss": 0.8739,
+      "step": 10817
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.807938976586307e-06,
+      "loss": 0.9729,
+      "step": 10818
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.80725214844426e-06,
+      "loss": 0.9174,
+      "step": 10819
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.806565304661888e-06,
+      "loss": 1.009,
+      "step": 10820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.805878445252495e-06,
+      "loss": 0.978,
+      "step": 10821
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.805191570229391e-06,
+      "loss": 0.9723,
+      "step": 10822
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.8045046796058825e-06,
+      "loss": 1.0287,
+      "step": 10823
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.803817773395278e-06,
+      "loss": 0.8566,
+      "step": 10824
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.803130851610887e-06,
+      "loss": 0.9924,
+      "step": 10825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.802443914266017e-06,
+      "loss": 0.9992,
+      "step": 10826
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.80175696137398e-06,
+      "loss": 0.8781,
+      "step": 10827
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.801069992948082e-06,
+      "loss": 0.72,
+      "step": 10828
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.800383009001636e-06,
+      "loss": 1.0436,
+      "step": 10829
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.799696009547949e-06,
+      "loss": 0.8554,
+      "step": 10830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.799008994600335e-06,
+      "loss": 1.0279,
+      "step": 10831
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.798321964172102e-06,
+      "loss": 0.9452,
+      "step": 10832
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.797634918276562e-06,
+      "loss": 0.7755,
+      "step": 10833
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.796947856927028e-06,
+      "loss": 0.6866,
+      "step": 10834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.796260780136809e-06,
+      "loss": 0.935,
+      "step": 10835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.795573687919217e-06,
+      "loss": 0.9973,
+      "step": 10836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.794886580287565e-06,
+      "loss": 0.8442,
+      "step": 10837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.794199457255167e-06,
+      "loss": 0.9568,
+      "step": 10838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.793512318835335e-06,
+      "loss": 1.004,
+      "step": 10839
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.792825165041381e-06,
+      "loss": 0.8189,
+      "step": 10840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.79213799588662e-06,
+      "loss": 0.8786,
+      "step": 10841
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.791450811384365e-06,
+      "loss": 1.0443,
+      "step": 10842
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.790763611547929e-06,
+      "loss": 0.9623,
+      "step": 10843
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.790076396390627e-06,
+      "loss": 0.8803,
+      "step": 10844
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.789389165925775e-06,
+      "loss": 0.8928,
+      "step": 10845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.788701920166687e-06,
+      "loss": 0.9661,
+      "step": 10846
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.788014659126678e-06,
+      "loss": 0.817,
+      "step": 10847
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.787327382819063e-06,
+      "loss": 0.959,
+      "step": 10848
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7866400912571605e-06,
+      "loss": 1.0607,
+      "step": 10849
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.785952784454284e-06,
+      "loss": 0.9364,
+      "step": 10850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.78526546242375e-06,
+      "loss": 1.0312,
+      "step": 10851
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.784578125178877e-06,
+      "loss": 0.8449,
+      "step": 10852
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.78389077273298e-06,
+      "loss": 1.1539,
+      "step": 10853
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.783203405099376e-06,
+      "loss": 0.8917,
+      "step": 10854
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.782516022291384e-06,
+      "loss": 0.8017,
+      "step": 10855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.781828624322322e-06,
+      "loss": 0.8285,
+      "step": 10856
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.781141211205509e-06,
+      "loss": 1.0904,
+      "step": 10857
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.780453782954261e-06,
+      "loss": 1.1815,
+      "step": 10858
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.779766339581898e-06,
+      "loss": 0.844,
+      "step": 10859
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.77907888110174e-06,
+      "loss": 0.8999,
+      "step": 10860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.778391407527104e-06,
+      "loss": 0.9532,
+      "step": 10861
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.777703918871311e-06,
+      "loss": 0.9906,
+      "step": 10862
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.777016415147683e-06,
+      "loss": 0.9596,
+      "step": 10863
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.776328896369537e-06,
+      "loss": 1.0008,
+      "step": 10864
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.775641362550195e-06,
+      "loss": 0.8809,
+      "step": 10865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.774953813702976e-06,
+      "loss": 0.9456,
+      "step": 10866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.774266249841204e-06,
+      "loss": 0.9161,
+      "step": 10867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7735786709781985e-06,
+      "loss": 0.8079,
+      "step": 10868
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.772891077127283e-06,
+      "loss": 1.1585,
+      "step": 10869
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7722034683017765e-06,
+      "loss": 0.9252,
+      "step": 10870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7715158445150045e-06,
+      "loss": 0.8594,
+      "step": 10871
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7708282057802855e-06,
+      "loss": 0.9463,
+      "step": 10872
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.770140552110947e-06,
+      "loss": 0.7561,
+      "step": 10873
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.76945288352031e-06,
+      "loss": 1.0038,
+      "step": 10874
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.768765200021697e-06,
+      "loss": 0.9452,
+      "step": 10875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.768077501628434e-06,
+      "loss": 1.1047,
+      "step": 10876
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.767389788353844e-06,
+      "loss": 1.0564,
+      "step": 10877
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.766702060211248e-06,
+      "loss": 1.0908,
+      "step": 10878
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.766014317213975e-06,
+      "loss": 0.8725,
+      "step": 10879
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.765326559375348e-06,
+      "loss": 1.0586,
+      "step": 10880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7646387867086935e-06,
+      "loss": 0.8759,
+      "step": 10881
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.763950999227336e-06,
+      "loss": 0.9441,
+      "step": 10882
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7632631969445996e-06,
+      "loss": 1.0625,
+      "step": 10883
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.762575379873814e-06,
+      "loss": 0.9954,
+      "step": 10884
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.761887548028302e-06,
+      "loss": 0.9526,
+      "step": 10885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.761199701421392e-06,
+      "loss": 0.9093,
+      "step": 10886
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.760511840066411e-06,
+      "loss": 0.7809,
+      "step": 10887
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.759823963976685e-06,
+      "loss": 0.886,
+      "step": 10888
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.759136073165541e-06,
+      "loss": 0.9308,
+      "step": 10889
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.758448167646309e-06,
+      "loss": 0.986,
+      "step": 10890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.757760247432316e-06,
+      "loss": 1.1617,
+      "step": 10891
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.75707231253689e-06,
+      "loss": 1.0002,
+      "step": 10892
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.75638436297336e-06,
+      "loss": 0.9215,
+      "step": 10893
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.755696398755054e-06,
+      "loss": 1.1706,
+      "step": 10894
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.755008419895301e-06,
+      "loss": 1.0468,
+      "step": 10895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7543204264074325e-06,
+      "loss": 1.0156,
+      "step": 10896
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.753632418304777e-06,
+      "loss": 0.9783,
+      "step": 10897
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.752944395600664e-06,
+      "loss": 1.067,
+      "step": 10898
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.752256358308424e-06,
+      "loss": 0.9591,
+      "step": 10899
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7515683064413865e-06,
+      "loss": 0.9343,
+      "step": 10900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.750880240012884e-06,
+      "loss": 1.0879,
+      "step": 10901
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7501921590362476e-06,
+      "loss": 0.9679,
+      "step": 10902
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.749504063524809e-06,
+      "loss": 0.797,
+      "step": 10903
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.748815953491897e-06,
+      "loss": 0.9518,
+      "step": 10904
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.748127828950848e-06,
+      "loss": 0.955,
+      "step": 10905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.747439689914992e-06,
+      "loss": 0.9753,
+      "step": 10906
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.74675153639766e-06,
+      "loss": 0.9416,
+      "step": 10907
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.746063368412187e-06,
+      "loss": 0.9772,
+      "step": 10908
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.745375185971906e-06,
+      "loss": 0.9509,
+      "step": 10909
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.744686989090149e-06,
+      "loss": 1.0922,
+      "step": 10910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.743998777780252e-06,
+      "loss": 0.8066,
+      "step": 10911
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.743310552055546e-06,
+      "loss": 0.9469,
+      "step": 10912
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.742622311929367e-06,
+      "loss": 0.8554,
+      "step": 10913
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.741934057415048e-06,
+      "loss": 1.0493,
+      "step": 10914
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.741245788525928e-06,
+      "loss": 0.99,
+      "step": 10915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7405575052753374e-06,
+      "loss": 0.9994,
+      "step": 10916
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.739869207676613e-06,
+      "loss": 0.9301,
+      "step": 10917
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.739180895743089e-06,
+      "loss": 1.0174,
+      "step": 10918
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.738492569488104e-06,
+      "loss": 0.9214,
+      "step": 10919
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.737804228924994e-06,
+      "loss": 1.0028,
+      "step": 10920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.737115874067094e-06,
+      "loss": 0.877,
+      "step": 10921
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7364275049277404e-06,
+      "loss": 0.9469,
+      "step": 10922
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.735739121520272e-06,
+      "loss": 1.1079,
+      "step": 10923
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7350507238580245e-06,
+      "loss": 0.9189,
+      "step": 10924
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.734362311954336e-06,
+      "loss": 0.8976,
+      "step": 10925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.733673885822543e-06,
+      "loss": 0.9583,
+      "step": 10926
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.732985445475987e-06,
+      "loss": 0.851,
+      "step": 10927
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.732296990928004e-06,
+      "loss": 0.9886,
+      "step": 10928
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.731608522191932e-06,
+      "loss": 1.0151,
+      "step": 10929
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.730920039281111e-06,
+      "loss": 1.062,
+      "step": 10930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.730231542208879e-06,
+      "loss": 0.9581,
+      "step": 10931
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.729543030988578e-06,
+      "loss": 0.7974,
+      "step": 10932
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.728854505633547e-06,
+      "loss": 0.9732,
+      "step": 10933
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.728165966157125e-06,
+      "loss": 0.8515,
+      "step": 10934
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.727477412572651e-06,
+      "loss": 1.1411,
+      "step": 10935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.726788844893467e-06,
+      "loss": 0.8231,
+      "step": 10936
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.726100263132915e-06,
+      "loss": 0.9229,
+      "step": 10937
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.725411667304335e-06,
+      "loss": 0.875,
+      "step": 10938
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.724723057421069e-06,
+      "loss": 0.9175,
+      "step": 10939
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.724034433496457e-06,
+      "loss": 0.8016,
+      "step": 10940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.723345795543842e-06,
+      "loss": 0.896,
+      "step": 10941
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.722657143576568e-06,
+      "loss": 0.9499,
+      "step": 10942
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.721968477607974e-06,
+      "loss": 0.7938,
+      "step": 10943
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.721279797651406e-06,
+      "loss": 0.9196,
+      "step": 10944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7205911037202045e-06,
+      "loss": 0.9748,
+      "step": 10945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.719902395827714e-06,
+      "loss": 1.112,
+      "step": 10946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.719213673987277e-06,
+      "loss": 0.965,
+      "step": 10947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.71852493821224e-06,
+      "loss": 0.9065,
+      "step": 10948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.717836188515945e-06,
+      "loss": 1.0727,
+      "step": 10949
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.717147424911736e-06,
+      "loss": 0.9617,
+      "step": 10950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.716458647412959e-06,
+      "loss": 0.9345,
+      "step": 10951
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7157698560329575e-06,
+      "loss": 1.0197,
+      "step": 10952
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.715081050785079e-06,
+      "loss": 0.9294,
+      "step": 10953
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.714392231682665e-06,
+      "loss": 0.906,
+      "step": 10954
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.713703398739066e-06,
+      "loss": 0.8997,
+      "step": 10955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7130145519676235e-06,
+      "loss": 0.9334,
+      "step": 10956
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.712325691381686e-06,
+      "loss": 1.1528,
+      "step": 10957
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7116368169946016e-06,
+      "loss": 0.7305,
+      "step": 10958
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.710947928819712e-06,
+      "loss": 0.9444,
+      "step": 10959
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.710259026870371e-06,
+      "loss": 0.8474,
+      "step": 10960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.70957011115992e-06,
+      "loss": 0.9016,
+      "step": 10961
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.708881181701711e-06,
+      "loss": 0.918,
+      "step": 10962
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.708192238509088e-06,
+      "loss": 0.9789,
+      "step": 10963
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.707503281595403e-06,
+      "loss": 0.9526,
+      "step": 10964
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.706814310974e-06,
+      "loss": 0.9914,
+      "step": 10965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.706125326658232e-06,
+      "loss": 0.9662,
+      "step": 10966
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.705436328661446e-06,
+      "loss": 0.93,
+      "step": 10967
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.70474731699699e-06,
+      "loss": 0.8857,
+      "step": 10968
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.704058291678215e-06,
+      "loss": 0.8161,
+      "step": 10969
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.70336925271847e-06,
+      "loss": 0.8785,
+      "step": 10970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.702680200131105e-06,
+      "loss": 0.9253,
+      "step": 10971
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.701991133929472e-06,
+      "loss": 1.0114,
+      "step": 10972
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.701302054126919e-06,
+      "loss": 1.0446,
+      "step": 10973
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.700612960736799e-06,
+      "loss": 1.0692,
+      "step": 10974
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.699923853772461e-06,
+      "loss": 1.0009,
+      "step": 10975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.699234733247255e-06,
+      "loss": 0.9139,
+      "step": 10976
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6985455991745376e-06,
+      "loss": 0.8339,
+      "step": 10977
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.697856451567657e-06,
+      "loss": 0.9766,
+      "step": 10978
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.697167290439966e-06,
+      "loss": 0.8345,
+      "step": 10979
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.696478115804817e-06,
+      "loss": 1.0748,
+      "step": 10980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.695788927675562e-06,
+      "loss": 0.8532,
+      "step": 10981
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6950997260655535e-06,
+      "loss": 1.1533,
+      "step": 10982
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.694410510988146e-06,
+      "loss": 0.9094,
+      "step": 10983
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.693721282456692e-06,
+      "loss": 0.9218,
+      "step": 10984
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6930320404845475e-06,
+      "loss": 0.8623,
+      "step": 10985
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6923427850850635e-06,
+      "loss": 0.8382,
+      "step": 10986
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.691653516271595e-06,
+      "loss": 1.0115,
+      "step": 10987
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.690964234057496e-06,
+      "loss": 1.1525,
+      "step": 10988
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.690274938456122e-06,
+      "loss": 0.9536,
+      "step": 10989
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.689585629480828e-06,
+      "loss": 1.0071,
+      "step": 10990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.688896307144969e-06,
+      "loss": 0.7206,
+      "step": 10991
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.688206971461899e-06,
+      "loss": 0.9171,
+      "step": 10992
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.687517622444977e-06,
+      "loss": 1.148,
+      "step": 10993
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6868282601075545e-06,
+      "loss": 0.9493,
+      "step": 10994
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.686138884462991e-06,
+      "loss": 0.8881,
+      "step": 10995
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.685449495524644e-06,
+      "loss": 1.0439,
+      "step": 10996
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.684760093305868e-06,
+      "loss": 0.9425,
+      "step": 10997
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.684070677820019e-06,
+      "loss": 1.0086,
+      "step": 10998
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.683381249080458e-06,
+      "loss": 0.9388,
+      "step": 10999
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.682691807100538e-06,
+      "loss": 1.0768,
+      "step": 11000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6820023518936195e-06,
+      "loss": 1.0571,
+      "step": 11001
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6813128834730615e-06,
+      "loss": 0.8442,
+      "step": 11002
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.68062340185222e-06,
+      "loss": 0.9668,
+      "step": 11003
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.679933907044455e-06,
+      "loss": 1.2747,
+      "step": 11004
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.679244399063123e-06,
+      "loss": 1.0133,
+      "step": 11005
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.678554877921586e-06,
+      "loss": 1.1044,
+      "step": 11006
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.677865343633203e-06,
+      "loss": 0.8664,
+      "step": 11007
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6771757962113325e-06,
+      "loss": 1.0292,
+      "step": 11008
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.676486235669334e-06,
+      "loss": 0.872,
+      "step": 11009
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.675796662020568e-06,
+      "loss": 0.9951,
+      "step": 11010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.675107075278396e-06,
+      "loss": 0.9946,
+      "step": 11011
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.674417475456177e-06,
+      "loss": 0.9841,
+      "step": 11012
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.673727862567272e-06,
+      "loss": 1.132,
+      "step": 11013
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6730382366250425e-06,
+      "loss": 1.0957,
+      "step": 11014
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6723485976428505e-06,
+      "loss": 1.0408,
+      "step": 11015
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6716589456340566e-06,
+      "loss": 1.0395,
+      "step": 11016
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.670969280612023e-06,
+      "loss": 0.7902,
+      "step": 11017
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.670279602590112e-06,
+      "loss": 0.8405,
+      "step": 11018
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.669589911581686e-06,
+      "loss": 0.8729,
+      "step": 11019
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.668900207600106e-06,
+      "loss": 1.0053,
+      "step": 11020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6682104906587375e-06,
+      "loss": 0.9463,
+      "step": 11021
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6675207607709426e-06,
+      "loss": 0.801,
+      "step": 11022
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.666831017950083e-06,
+      "loss": 0.8916,
+      "step": 11023
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6661412622095236e-06,
+      "loss": 1.0531,
+      "step": 11024
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.665451493562629e-06,
+      "loss": 1.0612,
+      "step": 11025
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6647617120227636e-06,
+      "loss": 0.7361,
+      "step": 11026
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.66407191760329e-06,
+      "loss": 1.0291,
+      "step": 11027
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.663382110317573e-06,
+      "loss": 0.9782,
+      "step": 11028
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.662692290178978e-06,
+      "loss": 0.9375,
+      "step": 11029
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.662002457200871e-06,
+      "loss": 0.9476,
+      "step": 11030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.661312611396618e-06,
+      "loss": 1.0017,
+      "step": 11031
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.660622752779581e-06,
+      "loss": 1.0108,
+      "step": 11032
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.659932881363128e-06,
+      "loss": 0.9629,
+      "step": 11033
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.659242997160625e-06,
+      "loss": 0.9313,
+      "step": 11034
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.65855310018544e-06,
+      "loss": 1.0196,
+      "step": 11035
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.657863190450938e-06,
+      "loss": 0.9275,
+      "step": 11036
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.657173267970485e-06,
+      "loss": 0.9007,
+      "step": 11037
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.65648333275745e-06,
+      "loss": 0.9799,
+      "step": 11038
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.655793384825198e-06,
+      "loss": 1.0119,
+      "step": 11039
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.655103424187097e-06,
+      "loss": 0.9567,
+      "step": 11040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.654413450856517e-06,
+      "loss": 0.9159,
+      "step": 11041
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6537234648468264e-06,
+      "loss": 0.942,
+      "step": 11042
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.65303346617139e-06,
+      "loss": 1.0116,
+      "step": 11043
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.65234345484358e-06,
+      "loss": 1.0251,
+      "step": 11044
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.651653430876762e-06,
+      "loss": 0.8841,
+      "step": 11045
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.650963394284308e-06,
+      "loss": 0.9469,
+      "step": 11046
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.650273345079584e-06,
+      "loss": 0.9407,
+      "step": 11047
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.649583283275963e-06,
+      "loss": 0.9182,
+      "step": 11048
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.648893208886813e-06,
+      "loss": 0.8805,
+      "step": 11049
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.648203121925504e-06,
+      "loss": 0.7516,
+      "step": 11050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.647513022405407e-06,
+      "loss": 0.889,
+      "step": 11051
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.646822910339891e-06,
+      "loss": 0.8831,
+      "step": 11052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6461327857423285e-06,
+      "loss": 0.9768,
+      "step": 11053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.645442648626089e-06,
+      "loss": 1.0998,
+      "step": 11054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.644752499004546e-06,
+      "loss": 0.97,
+      "step": 11055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.64406233689107e-06,
+      "loss": 0.9978,
+      "step": 11056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.64337216229903e-06,
+      "loss": 0.8793,
+      "step": 11057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.642681975241801e-06,
+      "loss": 0.9754,
+      "step": 11058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.641991775732756e-06,
+      "loss": 0.9365,
+      "step": 11059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.641301563785264e-06,
+      "loss": 0.8597,
+      "step": 11060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.640611339412702e-06,
+      "loss": 1.0171,
+      "step": 11061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.639921102628438e-06,
+      "loss": 0.9881,
+      "step": 11062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6392308534458495e-06,
+      "loss": 1.012,
+      "step": 11063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.638540591878308e-06,
+      "loss": 0.9192,
+      "step": 11064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.637850317939187e-06,
+      "loss": 0.6244,
+      "step": 11065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.637160031641861e-06,
+      "loss": 0.913,
+      "step": 11066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.636469732999704e-06,
+      "loss": 0.9391,
+      "step": 11067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.635779422026091e-06,
+      "loss": 0.9676,
+      "step": 11068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.635089098734394e-06,
+      "loss": 0.9829,
+      "step": 11069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.634398763137991e-06,
+      "loss": 1.0154,
+      "step": 11070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6337084152502555e-06,
+      "loss": 0.799,
+      "step": 11071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.633018055084564e-06,
+      "loss": 0.878,
+      "step": 11072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.63232768265429e-06,
+      "loss": 0.9817,
+      "step": 11073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.631637297972812e-06,
+      "loss": 0.9795,
+      "step": 11074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.630946901053505e-06,
+      "loss": 1.0131,
+      "step": 11075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.630256491909744e-06,
+      "loss": 0.8288,
+      "step": 11076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.629566070554906e-06,
+      "loss": 0.8858,
+      "step": 11077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.628875637002369e-06,
+      "loss": 0.8578,
+      "step": 11078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.628185191265508e-06,
+      "loss": 0.9413,
+      "step": 11079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.627494733357702e-06,
+      "loss": 0.8499,
+      "step": 11080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.626804263292328e-06,
+      "loss": 1.0595,
+      "step": 11081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.626113781082764e-06,
+      "loss": 0.9926,
+      "step": 11082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.625423286742387e-06,
+      "loss": 0.9447,
+      "step": 11083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6247327802845756e-06,
+      "loss": 0.9258,
+      "step": 11084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.624042261722709e-06,
+      "loss": 0.7723,
+      "step": 11085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6233517310701646e-06,
+      "loss": 0.9483,
+      "step": 11086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.622661188340321e-06,
+      "loss": 1.054,
+      "step": 11087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.621970633546558e-06,
+      "loss": 1.0904,
+      "step": 11088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.621280066702257e-06,
+      "loss": 0.8456,
+      "step": 11089
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.620589487820794e-06,
+      "loss": 0.9951,
+      "step": 11090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.619898896915552e-06,
+      "loss": 1.095,
+      "step": 11091
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.619208293999908e-06,
+      "loss": 1.072,
+      "step": 11092
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.618517679087244e-06,
+      "loss": 0.8481,
+      "step": 11093
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.617827052190941e-06,
+      "loss": 0.8664,
+      "step": 11094
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.617136413324379e-06,
+      "loss": 0.8937,
+      "step": 11095
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6164457625009386e-06,
+      "loss": 0.9275,
+      "step": 11096
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.615755099734002e-06,
+      "loss": 0.9162,
+      "step": 11097
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.615064425036948e-06,
+      "loss": 0.9936,
+      "step": 11098
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.614373738423161e-06,
+      "loss": 1.1614,
+      "step": 11099
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.613683039906024e-06,
+      "loss": 0.9315,
+      "step": 11100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.612992329498917e-06,
+      "loss": 0.9689,
+      "step": 11101
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.612301607215222e-06,
+      "loss": 0.9179,
+      "step": 11102
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.611610873068321e-06,
+      "loss": 0.9258,
+      "step": 11103
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.610920127071599e-06,
+      "loss": 0.9986,
+      "step": 11104
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.610229369238438e-06,
+      "loss": 1.0164,
+      "step": 11105
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.609538599582222e-06,
+      "loss": 0.8941,
+      "step": 11106
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.608847818116333e-06,
+      "loss": 0.8486,
+      "step": 11107
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.608157024854157e-06,
+      "loss": 1.0124,
+      "step": 11108
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.607466219809075e-06,
+      "loss": 1.0375,
+      "step": 11109
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.606775402994474e-06,
+      "loss": 0.9852,
+      "step": 11110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.606084574423738e-06,
+      "loss": 0.8936,
+      "step": 11111
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.605393734110249e-06,
+      "loss": 1.0647,
+      "step": 11112
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.604702882067394e-06,
+      "loss": 0.9554,
+      "step": 11113
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.604012018308559e-06,
+      "loss": 0.9344,
+      "step": 11114
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.603321142847128e-06,
+      "loss": 0.844,
+      "step": 11115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.602630255696485e-06,
+      "loss": 0.9324,
+      "step": 11116
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.601939356870019e-06,
+      "loss": 0.949,
+      "step": 11117
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.601248446381114e-06,
+      "loss": 1.1185,
+      "step": 11118
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.600557524243157e-06,
+      "loss": 0.9342,
+      "step": 11119
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.599866590469534e-06,
+      "loss": 1.1665,
+      "step": 11120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.599175645073632e-06,
+      "loss": 0.8263,
+      "step": 11121
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.598484688068837e-06,
+      "loss": 0.937,
+      "step": 11122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.597793719468537e-06,
+      "loss": 0.9158,
+      "step": 11123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.597102739286121e-06,
+      "loss": 0.9429,
+      "step": 11124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.596411747534973e-06,
+      "loss": 1.0061,
+      "step": 11125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.595720744228484e-06,
+      "loss": 1.0586,
+      "step": 11126
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5950297293800395e-06,
+      "loss": 1.1327,
+      "step": 11127
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.594338703003029e-06,
+      "loss": 1.1805,
+      "step": 11128
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.593647665110843e-06,
+      "loss": 0.81,
+      "step": 11129
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.592956615716867e-06,
+      "loss": 0.9416,
+      "step": 11130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.592265554834491e-06,
+      "loss": 0.8235,
+      "step": 11131
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5915744824771035e-06,
+      "loss": 0.8838,
+      "step": 11132
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.590883398658095e-06,
+      "loss": 0.9218,
+      "step": 11133
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.590192303390855e-06,
+      "loss": 0.9244,
+      "step": 11134
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.589501196688773e-06,
+      "loss": 1.0026,
+      "step": 11135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.588810078565239e-06,
+      "loss": 0.9555,
+      "step": 11136
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.588118949033643e-06,
+      "loss": 0.8302,
+      "step": 11137
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.587427808107375e-06,
+      "loss": 1.1644,
+      "step": 11138
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5867366557998285e-06,
+      "loss": 1.1142,
+      "step": 11139
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.586045492124391e-06,
+      "loss": 1.0797,
+      "step": 11140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.585354317094456e-06,
+      "loss": 1.0377,
+      "step": 11141
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5846631307234135e-06,
+      "loss": 0.8675,
+      "step": 11142
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5839719330246546e-06,
+      "loss": 1.0566,
+      "step": 11143
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5832807240115725e-06,
+      "loss": 0.9165,
+      "step": 11144
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.582589503697558e-06,
+      "loss": 1.0806,
+      "step": 11145
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.581898272096005e-06,
+      "loss": 1.0223,
+      "step": 11146
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.581207029220305e-06,
+      "loss": 0.9868,
+      "step": 11147
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.58051577508385e-06,
+      "loss": 0.8095,
+      "step": 11148
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.579824509700035e-06,
+      "loss": 0.8302,
+      "step": 11149
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.57913323308225e-06,
+      "loss": 1.0877,
+      "step": 11150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.578441945243889e-06,
+      "loss": 0.9258,
+      "step": 11151
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.577750646198349e-06,
+      "loss": 0.8669,
+      "step": 11152
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.577059335959021e-06,
+      "loss": 1.0063,
+      "step": 11153
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.576368014539298e-06,
+      "loss": 0.9865,
+      "step": 11154
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.575676681952577e-06,
+      "loss": 1.0159,
+      "step": 11155
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.574985338212249e-06,
+      "loss": 1.0414,
+      "step": 11156
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.574293983331711e-06,
+      "loss": 0.8258,
+      "step": 11157
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.573602617324358e-06,
+      "loss": 0.9524,
+      "step": 11158
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.572911240203585e-06,
+      "loss": 0.9102,
+      "step": 11159
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.572219851982785e-06,
+      "loss": 0.9806,
+      "step": 11160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5715284526753565e-06,
+      "loss": 1.0431,
+      "step": 11161
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5708370422946925e-06,
+      "loss": 1.1174,
+      "step": 11162
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.570145620854191e-06,
+      "loss": 1.016,
+      "step": 11163
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.569454188367248e-06,
+      "loss": 0.8506,
+      "step": 11164
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5687627448472594e-06,
+      "loss": 0.8887,
+      "step": 11165
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.56807129030762e-06,
+      "loss": 0.9725,
+      "step": 11166
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.567379824761731e-06,
+      "loss": 0.8895,
+      "step": 11167
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.566688348222984e-06,
+      "loss": 0.8969,
+      "step": 11168
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5659968607047795e-06,
+      "loss": 0.9128,
+      "step": 11169
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.565305362220515e-06,
+      "loss": 0.9729,
+      "step": 11170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.564613852783587e-06,
+      "loss": 1.0163,
+      "step": 11171
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.563922332407394e-06,
+      "loss": 0.8909,
+      "step": 11172
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5632308011053325e-06,
+      "loss": 1.0364,
+      "step": 11173
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.562539258890802e-06,
+      "loss": 0.8443,
+      "step": 11174
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.561847705777202e-06,
+      "loss": 0.8493,
+      "step": 11175
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5611561417779306e-06,
+      "loss": 0.9422,
+      "step": 11176
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.560464566906385e-06,
+      "loss": 1.0064,
+      "step": 11177
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.559772981175966e-06,
+      "loss": 0.9858,
+      "step": 11178
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.559081384600074e-06,
+      "loss": 1.0329,
+      "step": 11179
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.558389777192103e-06,
+      "loss": 0.9074,
+      "step": 11180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.55769815896546e-06,
+      "loss": 0.7951,
+      "step": 11181
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.557006529933541e-06,
+      "loss": 1.046,
+      "step": 11182
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.556314890109747e-06,
+      "loss": 0.9617,
+      "step": 11183
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.555623239507477e-06,
+      "loss": 0.943,
+      "step": 11184
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.554931578140134e-06,
+      "loss": 0.7754,
+      "step": 11185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.554239906021117e-06,
+      "loss": 0.9108,
+      "step": 11186
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.553548223163827e-06,
+      "loss": 0.9241,
+      "step": 11187
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.552856529581667e-06,
+      "loss": 0.8586,
+      "step": 11188
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.552164825288037e-06,
+      "loss": 0.9632,
+      "step": 11189
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.551473110296339e-06,
+      "loss": 0.7881,
+      "step": 11190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.550781384619974e-06,
+      "loss": 1.1457,
+      "step": 11191
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.550089648272343e-06,
+      "loss": 0.8382,
+      "step": 11192
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.549397901266852e-06,
+      "loss": 1.0225,
+      "step": 11193
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.548706143616901e-06,
+      "loss": 0.9101,
+      "step": 11194
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.548014375335893e-06,
+      "loss": 1.013,
+      "step": 11195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.54732259643723e-06,
+      "loss": 0.8385,
+      "step": 11196
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.546630806934315e-06,
+      "loss": 0.8704,
+      "step": 11197
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.545939006840555e-06,
+      "loss": 0.9138,
+      "step": 11198
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5452471961693485e-06,
+      "loss": 0.7885,
+      "step": 11199
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5445553749341015e-06,
+      "loss": 1.0274,
+      "step": 11200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.543863543148218e-06,
+      "loss": 0.8529,
+      "step": 11201
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5431717008251015e-06,
+      "loss": 0.8857,
+      "step": 11202
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.542479847978157e-06,
+      "loss": 0.9587,
+      "step": 11203
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.541787984620789e-06,
+      "loss": 0.9135,
+      "step": 11204
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5410961107664e-06,
+      "loss": 0.8874,
+      "step": 11205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.540404226428399e-06,
+      "loss": 0.9134,
+      "step": 11206
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.539712331620186e-06,
+      "loss": 1.031,
+      "step": 11207
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.53902042635517e-06,
+      "loss": 0.8169,
+      "step": 11208
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.538328510646756e-06,
+      "loss": 0.9925,
+      "step": 11209
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.537636584508349e-06,
+      "loss": 0.8985,
+      "step": 11210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.536944647953355e-06,
+      "loss": 0.9742,
+      "step": 11211
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5362527009951805e-06,
+      "loss": 0.9244,
+      "step": 11212
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.535560743647231e-06,
+      "loss": 0.8321,
+      "step": 11213
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.534868775922914e-06,
+      "loss": 1.0697,
+      "step": 11214
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.534176797835635e-06,
+      "loss": 1.0061,
+      "step": 11215
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5334848093988025e-06,
+      "loss": 0.806,
+      "step": 11216
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.532792810625822e-06,
+      "loss": 1.0201,
+      "step": 11217
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5321008015301035e-06,
+      "loss": 0.9818,
+      "step": 11218
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.531408782125052e-06,
+      "loss": 0.9733,
+      "step": 11219
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.530716752424074e-06,
+      "loss": 0.9023,
+      "step": 11220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.53002471244058e-06,
+      "loss": 0.8839,
+      "step": 11221
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.529332662187979e-06,
+      "loss": 1.0654,
+      "step": 11222
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.528640601679677e-06,
+      "loss": 1.0842,
+      "step": 11223
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.527948530929084e-06,
+      "loss": 0.8482,
+      "step": 11224
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.527256449949606e-06,
+      "loss": 1.0519,
+      "step": 11225
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.526564358754655e-06,
+      "loss": 0.9248,
+      "step": 11226
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.525872257357638e-06,
+      "loss": 0.8948,
+      "step": 11227
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.525180145771967e-06,
+      "loss": 0.9386,
+      "step": 11228
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5244880240110495e-06,
+      "loss": 1.1302,
+      "step": 11229
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.523795892088295e-06,
+      "loss": 0.9048,
+      "step": 11230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.523103750017115e-06,
+      "loss": 0.9577,
+      "step": 11231
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5224115978109175e-06,
+      "loss": 1.1941,
+      "step": 11232
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.521719435483114e-06,
+      "loss": 1.0567,
+      "step": 11233
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.521027263047114e-06,
+      "loss": 0.8178,
+      "step": 11234
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.520335080516331e-06,
+      "loss": 1.0946,
+      "step": 11235
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.519642887904172e-06,
+      "loss": 0.8297,
+      "step": 11236
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.51895068522405e-06,
+      "loss": 0.8012,
+      "step": 11237
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.518258472489376e-06,
+      "loss": 0.9363,
+      "step": 11238
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.517566249713562e-06,
+      "loss": 0.9665,
+      "step": 11239
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.51687401691002e-06,
+      "loss": 1.0177,
+      "step": 11240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5161817740921605e-06,
+      "loss": 0.9724,
+      "step": 11241
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.515489521273396e-06,
+      "loss": 0.8573,
+      "step": 11242
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5147972584671405e-06,
+      "loss": 0.7504,
+      "step": 11243
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.514104985686802e-06,
+      "loss": 0.8196,
+      "step": 11244
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.513412702945799e-06,
+      "loss": 0.8893,
+      "step": 11245
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.512720410257539e-06,
+      "loss": 0.8381,
+      "step": 11246
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5120281076354385e-06,
+      "loss": 1.0922,
+      "step": 11247
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.511335795092911e-06,
+      "loss": 0.9166,
+      "step": 11248
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.510643472643365e-06,
+      "loss": 1.0953,
+      "step": 11249
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.509951140300221e-06,
+      "loss": 0.9978,
+      "step": 11250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5092587980768875e-06,
+      "loss": 0.8734,
+      "step": 11251
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.50856644598678e-06,
+      "loss": 0.9237,
+      "step": 11252
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5078740840433145e-06,
+      "loss": 0.958,
+      "step": 11253
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.507181712259903e-06,
+      "loss": 0.9076,
+      "step": 11254
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.50648933064996e-06,
+      "loss": 1.0618,
+      "step": 11255
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.505796939226901e-06,
+      "loss": 0.8899,
+      "step": 11256
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.505104538004144e-06,
+      "loss": 1.0354,
+      "step": 11257
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.504412126995098e-06,
+      "loss": 0.8505,
+      "step": 11258
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.503719706213183e-06,
+      "loss": 0.9632,
+      "step": 11259
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.503027275671813e-06,
+      "loss": 1.1913,
+      "step": 11260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.502334835384403e-06,
+      "loss": 0.9715,
+      "step": 11261
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.50164238536437e-06,
+      "loss": 0.883,
+      "step": 11262
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.50094992562513e-06,
+      "loss": 0.8611,
+      "step": 11263
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.500257456180098e-06,
+      "loss": 1.1013,
+      "step": 11264
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.499564977042691e-06,
+      "loss": 0.9213,
+      "step": 11265
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4988724882263265e-06,
+      "loss": 0.9144,
+      "step": 11266
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.49817998974442e-06,
+      "loss": 0.7186,
+      "step": 11267
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.49748748161039e-06,
+      "loss": 0.9988,
+      "step": 11268
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.496794963837653e-06,
+      "loss": 0.9493,
+      "step": 11269
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.496102436439625e-06,
+      "loss": 1.1014,
+      "step": 11270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.495409899429726e-06,
+      "loss": 0.8509,
+      "step": 11271
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4947173528213724e-06,
+      "loss": 0.8541,
+      "step": 11272
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.494024796627982e-06,
+      "loss": 0.8782,
+      "step": 11273
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.493332230862973e-06,
+      "loss": 1.0521,
+      "step": 11274
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.492639655539764e-06,
+      "loss": 0.8105,
+      "step": 11275
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.491947070671774e-06,
+      "loss": 0.7564,
+      "step": 11276
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.491254476272421e-06,
+      "loss": 0.851,
+      "step": 11277
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.490561872355122e-06,
+      "loss": 1.077,
+      "step": 11278
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4898692589333005e-06,
+      "loss": 0.9435,
+      "step": 11279
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.489176636020372e-06,
+      "loss": 0.8418,
+      "step": 11280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.488484003629759e-06,
+      "loss": 0.9476,
+      "step": 11281
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.487791361774878e-06,
+      "loss": 1.0273,
+      "step": 11282
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.48709871046915e-06,
+      "loss": 0.9572,
+      "step": 11283
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.486406049725994e-06,
+      "loss": 1.0354,
+      "step": 11284
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.485713379558833e-06,
+      "loss": 0.9304,
+      "step": 11285
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.485020699981084e-06,
+      "loss": 0.7821,
+      "step": 11286
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.484328011006169e-06,
+      "loss": 0.8985,
+      "step": 11287
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.48363531264751e-06,
+      "loss": 1.0158,
+      "step": 11288
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4829426049185255e-06,
+      "loss": 1.0429,
+      "step": 11289
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.482249887832638e-06,
+      "loss": 0.7992,
+      "step": 11290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.481557161403268e-06,
+      "loss": 1.0156,
+      "step": 11291
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.480864425643837e-06,
+      "loss": 0.9071,
+      "step": 11292
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.480171680567767e-06,
+      "loss": 1.0426,
+      "step": 11293
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.47947892618848e-06,
+      "loss": 1.0167,
+      "step": 11294
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.478786162519398e-06,
+      "loss": 0.8818,
+      "step": 11295
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.478093389573941e-06,
+      "loss": 0.9461,
+      "step": 11296
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.477400607365534e-06,
+      "loss": 0.885,
+      "step": 11297
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.476707815907598e-06,
+      "loss": 0.9275,
+      "step": 11298
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.476015015213557e-06,
+      "loss": 0.9303,
+      "step": 11299
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.475322205296832e-06,
+      "loss": 1.0425,
+      "step": 11300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.474629386170849e-06,
+      "loss": 0.9139,
+      "step": 11301
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.473936557849027e-06,
+      "loss": 0.9119,
+      "step": 11302
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.473243720344793e-06,
+      "loss": 0.8803,
+      "step": 11303
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.472550873671569e-06,
+      "loss": 0.986,
+      "step": 11304
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.47185801784278e-06,
+      "loss": 0.8969,
+      "step": 11305
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.471165152871848e-06,
+      "loss": 0.9512,
+      "step": 11306
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4704722787721984e-06,
+      "loss": 0.9464,
+      "step": 11307
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.469779395557253e-06,
+      "loss": 1.022,
+      "step": 11308
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.46908650324044e-06,
+      "loss": 1.0087,
+      "step": 11309
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.468393601835184e-06,
+      "loss": 1.2135,
+      "step": 11310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.467700691354907e-06,
+      "loss": 0.8725,
+      "step": 11311
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4670077718130354e-06,
+      "loss": 0.8306,
+      "step": 11312
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.466314843222993e-06,
+      "loss": 0.9044,
+      "step": 11313
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.465621905598206e-06,
+      "loss": 0.8603,
+      "step": 11314
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4649289589521025e-06,
+      "loss": 0.9351,
+      "step": 11315
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.464236003298104e-06,
+      "loss": 1.074,
+      "step": 11316
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4635430386496395e-06,
+      "loss": 0.9642,
+      "step": 11317
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.462850065020133e-06,
+      "loss": 0.7816,
+      "step": 11318
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.462157082423011e-06,
+      "loss": 0.8301,
+      "step": 11319
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.461464090871701e-06,
+      "loss": 0.9147,
+      "step": 11320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.460771090379629e-06,
+      "loss": 0.8493,
+      "step": 11321
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.460078080960222e-06,
+      "loss": 0.9792,
+      "step": 11322
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.459385062626906e-06,
+      "loss": 0.9205,
+      "step": 11323
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.45869203539311e-06,
+      "loss": 0.9529,
+      "step": 11324
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.457998999272258e-06,
+      "loss": 0.8511,
+      "step": 11325
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.45730595427778e-06,
+      "loss": 1.0681,
+      "step": 11326
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4566129004231025e-06,
+      "loss": 1.0229,
+      "step": 11327
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.455919837721655e-06,
+      "loss": 0.945,
+      "step": 11328
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.455226766186863e-06,
+      "loss": 0.8953,
+      "step": 11329
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.454533685832155e-06,
+      "loss": 1.0294,
+      "step": 11330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.45384059667096e-06,
+      "loss": 0.9183,
+      "step": 11331
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.453147498716706e-06,
+      "loss": 0.8382,
+      "step": 11332
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.452454391982824e-06,
+      "loss": 0.8668,
+      "step": 11333
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.45176127648274e-06,
+      "loss": 0.9966,
+      "step": 11334
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.451068152229884e-06,
+      "loss": 0.9136,
+      "step": 11335
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.450375019237684e-06,
+      "loss": 1.0203,
+      "step": 11336
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.449681877519569e-06,
+      "loss": 0.9392,
+      "step": 11337
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.448988727088972e-06,
+      "loss": 0.7802,
+      "step": 11338
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4482955679593196e-06,
+      "loss": 0.9353,
+      "step": 11339
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.447602400144042e-06,
+      "loss": 1.0847,
+      "step": 11340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.44690922365657e-06,
+      "loss": 1.0102,
+      "step": 11341
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.446216038510331e-06,
+      "loss": 1.0111,
+      "step": 11342
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.44552284471876e-06,
+      "loss": 0.925,
+      "step": 11343
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.444829642295284e-06,
+      "loss": 0.8856,
+      "step": 11344
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4441364312533355e-06,
+      "loss": 1.0194,
+      "step": 11345
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.443443211606343e-06,
+      "loss": 1.1044,
+      "step": 11346
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.44274998336774e-06,
+      "loss": 0.9565,
+      "step": 11347
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.442056746550955e-06,
+      "loss": 1.0144,
+      "step": 11348
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.441363501169421e-06,
+      "loss": 0.79,
+      "step": 11349
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.44067024723657e-06,
+      "loss": 0.9092,
+      "step": 11350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.439976984765834e-06,
+      "loss": 0.7895,
+      "step": 11351
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.439283713770642e-06,
+      "loss": 1.0672,
+      "step": 11352
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.438590434264429e-06,
+      "loss": 0.9789,
+      "step": 11353
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.437897146260626e-06,
+      "loss": 0.9635,
+      "step": 11354
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.437203849772664e-06,
+      "loss": 1.0204,
+      "step": 11355
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.436510544813978e-06,
+      "loss": 0.9249,
+      "step": 11356
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.435817231397997e-06,
+      "loss": 0.947,
+      "step": 11357
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.435123909538158e-06,
+      "loss": 0.8628,
+      "step": 11358
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.434430579247891e-06,
+      "loss": 1.0189,
+      "step": 11359
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.433737240540629e-06,
+      "loss": 0.9369,
+      "step": 11360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.433043893429808e-06,
+      "loss": 0.8311,
+      "step": 11361
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.432350537928859e-06,
+      "loss": 0.9776,
+      "step": 11362
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.431657174051217e-06,
+      "loss": 0.7774,
+      "step": 11363
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.430963801810315e-06,
+      "loss": 0.9732,
+      "step": 11364
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.430270421219587e-06,
+      "loss": 0.9893,
+      "step": 11365
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.429577032292465e-06,
+      "loss": 1.0498,
+      "step": 11366
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.428883635042389e-06,
+      "loss": 1.0314,
+      "step": 11367
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4281902294827884e-06,
+      "loss": 0.9202,
+      "step": 11368
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4274968156271e-06,
+      "loss": 0.8891,
+      "step": 11369
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.426803393488757e-06,
+      "loss": 0.9266,
+      "step": 11370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.426109963081194e-06,
+      "loss": 0.9998,
+      "step": 11371
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.425416524417848e-06,
+      "loss": 0.9496,
+      "step": 11372
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.424723077512154e-06,
+      "loss": 0.9216,
+      "step": 11373
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4240296223775465e-06,
+      "loss": 0.9443,
+      "step": 11374
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4233361590274614e-06,
+      "loss": 0.9415,
+      "step": 11375
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.422642687475334e-06,
+      "loss": 1.0245,
+      "step": 11376
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4219492077345995e-06,
+      "loss": 1.0254,
+      "step": 11377
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4212557198186955e-06,
+      "loss": 0.845,
+      "step": 11378
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.420562223741058e-06,
+      "loss": 0.8678,
+      "step": 11379
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.419868719515122e-06,
+      "loss": 0.8927,
+      "step": 11380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.419175207154326e-06,
+      "loss": 0.8463,
+      "step": 11381
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.418481686672104e-06,
+      "loss": 1.1055,
+      "step": 11382
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.417788158081896e-06,
+      "loss": 0.8708,
+      "step": 11383
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.417094621397136e-06,
+      "loss": 1.056,
+      "step": 11384
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.416401076631263e-06,
+      "loss": 0.9604,
+      "step": 11385
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.415707523797713e-06,
+      "loss": 1.0074,
+      "step": 11386
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.415013962909924e-06,
+      "loss": 0.8791,
+      "step": 11387
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.414320393981334e-06,
+      "loss": 1.0669,
+      "step": 11388
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.413626817025379e-06,
+      "loss": 0.9066,
+      "step": 11389
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4129332320555e-06,
+      "loss": 1.1169,
+      "step": 11390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.412239639085132e-06,
+      "loss": 0.995,
+      "step": 11391
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.411546038127715e-06,
+      "loss": 1.0199,
+      "step": 11392
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.410852429196687e-06,
+      "loss": 0.8731,
+      "step": 11393
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.410158812305486e-06,
+      "loss": 0.8081,
+      "step": 11394
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.409465187467551e-06,
+      "loss": 0.9305,
+      "step": 11395
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.408771554696322e-06,
+      "loss": 1.0274,
+      "step": 11396
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.408077914005236e-06,
+      "loss": 0.8771,
+      "step": 11397
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.407384265407732e-06,
+      "loss": 0.9539,
+      "step": 11398
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.406690608917252e-06,
+      "loss": 1.0916,
+      "step": 11399
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.405996944547233e-06,
+      "loss": 0.8381,
+      "step": 11400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.405303272311116e-06,
+      "loss": 0.8871,
+      "step": 11401
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4046095922223385e-06,
+      "loss": 1.0424,
+      "step": 11402
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.403915904294343e-06,
+      "loss": 0.9246,
+      "step": 11403
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.403222208540568e-06,
+      "loss": 0.961,
+      "step": 11404
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.402528504974455e-06,
+      "loss": 1.108,
+      "step": 11405
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.401834793609442e-06,
+      "loss": 1.0565,
+      "step": 11406
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.401141074458972e-06,
+      "loss": 0.939,
+      "step": 11407
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.400447347536485e-06,
+      "loss": 0.8501,
+      "step": 11408
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.399753612855421e-06,
+      "loss": 0.9609,
+      "step": 11409
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.399059870429222e-06,
+      "loss": 1.0655,
+      "step": 11410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.398366120271329e-06,
+      "loss": 0.983,
+      "step": 11411
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.39767236239518e-06,
+      "loss": 0.951,
+      "step": 11412
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3969785968142215e-06,
+      "loss": 0.9701,
+      "step": 11413
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.396284823541893e-06,
+      "loss": 0.862,
+      "step": 11414
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.395591042591635e-06,
+      "loss": 0.971,
+      "step": 11415
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.394897253976891e-06,
+      "loss": 0.8863,
+      "step": 11416
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.394203457711101e-06,
+      "loss": 0.7792,
+      "step": 11417
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.393509653807711e-06,
+      "loss": 1.0048,
+      "step": 11418
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.392815842280158e-06,
+      "loss": 0.9123,
+      "step": 11419
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.392122023141889e-06,
+      "loss": 0.9709,
+      "step": 11420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3914281964063445e-06,
+      "loss": 0.8737,
+      "step": 11421
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.390734362086966e-06,
+      "loss": 0.9969,
+      "step": 11422
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.390040520197199e-06,
+      "loss": 0.9022,
+      "step": 11423
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.389346670750483e-06,
+      "loss": 1.0103,
+      "step": 11424
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.388652813760267e-06,
+      "loss": 0.974,
+      "step": 11425
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.38795894923999e-06,
+      "loss": 0.8321,
+      "step": 11426
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.387265077203097e-06,
+      "loss": 0.8749,
+      "step": 11427
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.38657119766303e-06,
+      "loss": 0.9334,
+      "step": 11428
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.385877310633233e-06,
+      "loss": 0.9262,
+      "step": 11429
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.385183416127151e-06,
+      "loss": 0.9891,
+      "step": 11430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.384489514158228e-06,
+      "loss": 1.0818,
+      "step": 11431
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.383795604739908e-06,
+      "loss": 0.9859,
+      "step": 11432
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.383101687885635e-06,
+      "loss": 0.8615,
+      "step": 11433
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.382407763608853e-06,
+      "loss": 0.8934,
+      "step": 11434
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.381713831923008e-06,
+      "loss": 1.0035,
+      "step": 11435
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3810198928415446e-06,
+      "loss": 0.9283,
+      "step": 11436
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.380325946377907e-06,
+      "loss": 1.0814,
+      "step": 11437
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.37963199254554e-06,
+      "loss": 0.9569,
+      "step": 11438
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.37893803135789e-06,
+      "loss": 0.8914,
+      "step": 11439
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.378244062828402e-06,
+      "loss": 0.898,
+      "step": 11440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.377550086970519e-06,
+      "loss": 0.9617,
+      "step": 11441
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.37685610379769e-06,
+      "loss": 0.9524,
+      "step": 11442
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.376162113323359e-06,
+      "loss": 0.8006,
+      "step": 11443
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.375468115560973e-06,
+      "loss": 0.9428,
+      "step": 11444
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.374774110523978e-06,
+      "loss": 0.9346,
+      "step": 11445
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.374080098225819e-06,
+      "loss": 0.8836,
+      "step": 11446
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.373386078679943e-06,
+      "loss": 1.0494,
+      "step": 11447
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.372692051899796e-06,
+      "loss": 0.9105,
+      "step": 11448
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.371998017898825e-06,
+      "loss": 0.8427,
+      "step": 11449
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.371303976690477e-06,
+      "loss": 0.9308,
+      "step": 11450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3706099282882e-06,
+      "loss": 0.9532,
+      "step": 11451
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.369915872705438e-06,
+      "loss": 0.9932,
+      "step": 11452
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3692218099556395e-06,
+      "loss": 0.9954,
+      "step": 11453
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.368527740052253e-06,
+      "loss": 0.9865,
+      "step": 11454
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.367833663008726e-06,
+      "loss": 0.684,
+      "step": 11455
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.367139578838504e-06,
+      "loss": 0.8711,
+      "step": 11456
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3664454875550365e-06,
+      "loss": 0.838,
+      "step": 11457
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.365751389171771e-06,
+      "loss": 1.0792,
+      "step": 11458
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.365057283702153e-06,
+      "loss": 1.1434,
+      "step": 11459
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.364363171159635e-06,
+      "loss": 0.9579,
+      "step": 11460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.363669051557664e-06,
+      "loss": 0.9287,
+      "step": 11461
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3629749249096865e-06,
+      "loss": 0.9653,
+      "step": 11462
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.362280791229153e-06,
+      "loss": 0.9877,
+      "step": 11463
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.361586650529509e-06,
+      "loss": 1.0871,
+      "step": 11464
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.360892502824208e-06,
+      "loss": 0.9368,
+      "step": 11465
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.360198348126696e-06,
+      "loss": 0.9315,
+      "step": 11466
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.359504186450422e-06,
+      "loss": 1.1602,
+      "step": 11467
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.358810017808837e-06,
+      "loss": 1.093,
+      "step": 11468
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.35811584221539e-06,
+      "loss": 0.9,
+      "step": 11469
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.357421659683527e-06,
+      "loss": 0.9106,
+      "step": 11470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.356727470226702e-06,
+      "loss": 1.0303,
+      "step": 11471
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3560332738583645e-06,
+      "loss": 0.8251,
+      "step": 11472
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3553390705919615e-06,
+      "loss": 0.8998,
+      "step": 11473
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.354644860440945e-06,
+      "loss": 0.9788,
+      "step": 11474
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.353950643418765e-06,
+      "loss": 1.0148,
+      "step": 11475
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.353256419538872e-06,
+      "loss": 0.9168,
+      "step": 11476
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.352562188814715e-06,
+      "loss": 0.8545,
+      "step": 11477
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.351867951259746e-06,
+      "loss": 0.9703,
+      "step": 11478
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.351173706887416e-06,
+      "loss": 1.0811,
+      "step": 11479
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3504794557111735e-06,
+      "loss": 0.8437,
+      "step": 11480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.349785197744473e-06,
+      "loss": 1.004,
+      "step": 11481
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.349090933000761e-06,
+      "loss": 1.0498,
+      "step": 11482
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.348396661493493e-06,
+      "loss": 0.9382,
+      "step": 11483
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.347702383236118e-06,
+      "loss": 1.0214,
+      "step": 11484
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.347008098242089e-06,
+      "loss": 0.9013,
+      "step": 11485
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3463138065248575e-06,
+      "loss": 0.6729,
+      "step": 11486
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.345619508097873e-06,
+      "loss": 0.8544,
+      "step": 11487
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.344925202974589e-06,
+      "loss": 0.8288,
+      "step": 11488
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.344230891168458e-06,
+      "loss": 0.9113,
+      "step": 11489
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.343536572692932e-06,
+      "loss": 0.9285,
+      "step": 11490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.342842247561463e-06,
+      "loss": 0.9859,
+      "step": 11491
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.342147915787502e-06,
+      "loss": 0.9117,
+      "step": 11492
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.341453577384502e-06,
+      "loss": 0.9194,
+      "step": 11493
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.340759232365918e-06,
+      "loss": 0.8455,
+      "step": 11494
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.340064880745201e-06,
+      "loss": 0.8184,
+      "step": 11495
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.339370522535805e-06,
+      "loss": 0.8817,
+      "step": 11496
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.33867615775118e-06,
+      "loss": 0.841,
+      "step": 11497
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.337981786404784e-06,
+      "loss": 0.924,
+      "step": 11498
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.337287408510064e-06,
+      "loss": 0.8677,
+      "step": 11499
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.336593024080478e-06,
+      "loss": 0.953,
+      "step": 11500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3358986331294805e-06,
+      "loss": 0.9457,
+      "step": 11501
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.335204235670523e-06,
+      "loss": 0.8429,
+      "step": 11502
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.334509831717058e-06,
+      "loss": 0.9631,
+      "step": 11503
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.333815421282542e-06,
+      "loss": 1.1916,
+      "step": 11504
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.333121004380427e-06,
+      "loss": 0.8958,
+      "step": 11505
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.332426581024168e-06,
+      "loss": 1.0052,
+      "step": 11506
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.33173215122722e-06,
+      "loss": 0.8919,
+      "step": 11507
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.331037715003038e-06,
+      "loss": 1.1451,
+      "step": 11508
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.330343272365073e-06,
+      "loss": 0.8103,
+      "step": 11509
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.329648823326783e-06,
+      "loss": 0.9699,
+      "step": 11510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.328954367901622e-06,
+      "loss": 0.9314,
+      "step": 11511
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.328259906103045e-06,
+      "loss": 1.0,
+      "step": 11512
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3275654379445064e-06,
+      "loss": 1.0359,
+      "step": 11513
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.326870963439462e-06,
+      "loss": 1.2523,
+      "step": 11514
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.326176482601366e-06,
+      "loss": 0.7752,
+      "step": 11515
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.325481995443675e-06,
+      "loss": 0.8458,
+      "step": 11516
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.324787501979845e-06,
+      "loss": 1.0222,
+      "step": 11517
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.32409300222333e-06,
+      "loss": 0.9611,
+      "step": 11518
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.323398496187586e-06,
+      "loss": 0.8016,
+      "step": 11519
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3227039838860715e-06,
+      "loss": 1.0316,
+      "step": 11520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.322009465332239e-06,
+      "loss": 0.9841,
+      "step": 11521
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.321314940539547e-06,
+      "loss": 0.7334,
+      "step": 11522
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.32062040952145e-06,
+      "loss": 0.8488,
+      "step": 11523
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.319925872291405e-06,
+      "loss": 0.9096,
+      "step": 11524
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.319231328862869e-06,
+      "loss": 1.041,
+      "step": 11525
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.318536779249299e-06,
+      "loss": 0.9042,
+      "step": 11526
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3178422234641526e-06,
+      "loss": 0.8784,
+      "step": 11527
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3171476615208815e-06,
+      "loss": 1.0535,
+      "step": 11528
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.31645309343295e-06,
+      "loss": 0.916,
+      "step": 11529
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.31575851921381e-06,
+      "loss": 1.0565,
+      "step": 11530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.31506393887692e-06,
+      "loss": 0.8525,
+      "step": 11531
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.314369352435738e-06,
+      "loss": 0.973,
+      "step": 11532
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3136747599037215e-06,
+      "loss": 0.7427,
+      "step": 11533
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.312980161294326e-06,
+      "loss": 0.8063,
+      "step": 11534
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.312285556621013e-06,
+      "loss": 1.0193,
+      "step": 11535
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.311590945897238e-06,
+      "loss": 0.9224,
+      "step": 11536
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.310896329136459e-06,
+      "loss": 0.9752,
+      "step": 11537
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.310201706352134e-06,
+      "loss": 0.8134,
+      "step": 11538
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.30950707755772e-06,
+      "loss": 0.8816,
+      "step": 11539
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.308812442766679e-06,
+      "loss": 0.9431,
+      "step": 11540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.308117801992465e-06,
+      "loss": 0.8865,
+      "step": 11541
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.30742315524854e-06,
+      "loss": 1.0551,
+      "step": 11542
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.30672850254836e-06,
+      "loss": 0.8213,
+      "step": 11543
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.306033843905385e-06,
+      "loss": 0.8241,
+      "step": 11544
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.305339179333074e-06,
+      "loss": 0.9855,
+      "step": 11545
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.304644508844885e-06,
+      "loss": 0.9515,
+      "step": 11546
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.303949832454278e-06,
+      "loss": 1.0093,
+      "step": 11547
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.303255150174712e-06,
+      "loss": 1.0516,
+      "step": 11548
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.302560462019647e-06,
+      "loss": 0.9192,
+      "step": 11549
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.301865768002541e-06,
+      "loss": 1.0103,
+      "step": 11550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.301171068136854e-06,
+      "loss": 0.9626,
+      "step": 11551
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.300476362436046e-06,
+      "loss": 0.8808,
+      "step": 11552
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.299781650913576e-06,
+      "loss": 0.9121,
+      "step": 11553
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.299086933582906e-06,
+      "loss": 0.9768,
+      "step": 11554
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.298392210457494e-06,
+      "loss": 0.9301,
+      "step": 11555
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2976974815508e-06,
+      "loss": 0.9487,
+      "step": 11556
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.297002746876284e-06,
+      "loss": 0.9027,
+      "step": 11557
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.296308006447408e-06,
+      "loss": 0.9331,
+      "step": 11558
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.295613260277632e-06,
+      "loss": 0.7874,
+      "step": 11559
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.294918508380417e-06,
+      "loss": 1.0292,
+      "step": 11560
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.294223750769222e-06,
+      "loss": 0.9153,
+      "step": 11561
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.293528987457508e-06,
+      "loss": 0.8514,
+      "step": 11562
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.292834218458737e-06,
+      "loss": 0.9043,
+      "step": 11563
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.292139443786368e-06,
+      "loss": 0.9145,
+      "step": 11564
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2914446634538654e-06,
+      "loss": 1.0561,
+      "step": 11565
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.290749877474688e-06,
+      "loss": 0.8254,
+      "step": 11566
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.290055085862298e-06,
+      "loss": 0.8915,
+      "step": 11567
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.289360288630157e-06,
+      "loss": 0.8483,
+      "step": 11568
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.288665485791725e-06,
+      "loss": 1.0397,
+      "step": 11569
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.287970677360465e-06,
+      "loss": 1.0196,
+      "step": 11570
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2872758633498386e-06,
+      "loss": 0.8519,
+      "step": 11571
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2865810437733075e-06,
+      "loss": 1.0173,
+      "step": 11572
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.285886218644334e-06,
+      "loss": 0.8853,
+      "step": 11573
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.285191387976379e-06,
+      "loss": 1.006,
+      "step": 11574
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.284496551782905e-06,
+      "loss": 0.9551,
+      "step": 11575
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.283801710077376e-06,
+      "loss": 0.9044,
+      "step": 11576
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.283106862873253e-06,
+      "loss": 0.8128,
+      "step": 11577
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.282412010183999e-06,
+      "loss": 0.9696,
+      "step": 11578
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.281717152023076e-06,
+      "loss": 0.9113,
+      "step": 11579
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.281022288403948e-06,
+      "loss": 0.882,
+      "step": 11580
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.280327419340074e-06,
+      "loss": 0.993,
+      "step": 11581
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.279632544844923e-06,
+      "loss": 0.9518,
+      "step": 11582
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.278937664931954e-06,
+      "loss": 1.0101,
+      "step": 11583
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.278242779614631e-06,
+      "loss": 0.9918,
+      "step": 11584
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.277547888906417e-06,
+      "loss": 0.8119,
+      "step": 11585
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.276852992820775e-06,
+      "loss": 1.0101,
+      "step": 11586
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.27615809137117e-06,
+      "loss": 0.9597,
+      "step": 11587
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2754631845710645e-06,
+      "loss": 0.8959,
+      "step": 11588
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.274768272433922e-06,
+      "loss": 0.7978,
+      "step": 11589
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.274073354973207e-06,
+      "loss": 0.8739,
+      "step": 11590
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.273378432202383e-06,
+      "loss": 0.8495,
+      "step": 11591
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.272683504134913e-06,
+      "loss": 0.9561,
+      "step": 11592
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.271988570784262e-06,
+      "loss": 0.959,
+      "step": 11593
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.271293632163895e-06,
+      "loss": 0.9301,
+      "step": 11594
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.270598688287275e-06,
+      "loss": 1.2779,
+      "step": 11595
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.269903739167868e-06,
+      "loss": 1.1626,
+      "step": 11596
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.269208784819135e-06,
+      "loss": 0.8993,
+      "step": 11597
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2685138252545445e-06,
+      "loss": 1.103,
+      "step": 11598
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.267818860487559e-06,
+      "loss": 1.1411,
+      "step": 11599
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2671238905316445e-06,
+      "loss": 1.1754,
+      "step": 11600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.266428915400265e-06,
+      "loss": 1.12,
+      "step": 11601
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.265733935106885e-06,
+      "loss": 0.9143,
+      "step": 11602
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.26503894966497e-06,
+      "loss": 1.0759,
+      "step": 11603
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.264343959087985e-06,
+      "loss": 0.7463,
+      "step": 11604
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.263648963389397e-06,
+      "loss": 0.93,
+      "step": 11605
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2629539625826705e-06,
+      "loss": 0.9517,
+      "step": 11606
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.26225895668127e-06,
+      "loss": 0.76,
+      "step": 11607
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.261563945698662e-06,
+      "loss": 0.8141,
+      "step": 11608
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2608689296483116e-06,
+      "loss": 0.914,
+      "step": 11609
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.260173908543684e-06,
+      "loss": 1.0042,
+      "step": 11610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.259478882398248e-06,
+      "loss": 0.8942,
+      "step": 11611
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.258783851225466e-06,
+      "loss": 0.9351,
+      "step": 11612
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.258088815038805e-06,
+      "loss": 1.1222,
+      "step": 11613
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.257393773851733e-06,
+      "loss": 0.994,
+      "step": 11614
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.256698727677716e-06,
+      "loss": 0.9976,
+      "step": 11615
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.256003676530217e-06,
+      "loss": 0.9658,
+      "step": 11616
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.255308620422706e-06,
+      "loss": 0.9171,
+      "step": 11617
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2546135593686484e-06,
+      "loss": 1.1727,
+      "step": 11618
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.253918493381511e-06,
+      "loss": 0.9773,
+      "step": 11619
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.25322342247476e-06,
+      "loss": 0.9471,
+      "step": 11620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.252528346661862e-06,
+      "loss": 1.1334,
+      "step": 11621
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.251833265956285e-06,
+      "loss": 0.9604,
+      "step": 11622
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.251138180371496e-06,
+      "loss": 0.9579,
+      "step": 11623
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2504430899209625e-06,
+      "loss": 0.9115,
+      "step": 11624
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.24974799461815e-06,
+      "loss": 0.9355,
+      "step": 11625
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.249052894476527e-06,
+      "loss": 0.9042,
+      "step": 11626
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2483577895095595e-06,
+      "loss": 0.9253,
+      "step": 11627
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.247662679730717e-06,
+      "loss": 0.8723,
+      "step": 11628
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.246967565153467e-06,
+      "loss": 0.9293,
+      "step": 11629
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.246272445791276e-06,
+      "loss": 0.8705,
+      "step": 11630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.245577321657612e-06,
+      "loss": 0.9468,
+      "step": 11631
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.244882192765944e-06,
+      "loss": 0.8869,
+      "step": 11632
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.244187059129739e-06,
+      "loss": 1.005,
+      "step": 11633
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.243491920762464e-06,
+      "loss": 0.9692,
+      "step": 11634
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.242796777677589e-06,
+      "loss": 0.8723,
+      "step": 11635
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.242101629888581e-06,
+      "loss": 0.9806,
+      "step": 11636
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.241406477408909e-06,
+      "loss": 0.9766,
+      "step": 11637
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.240711320252042e-06,
+      "loss": 0.8845,
+      "step": 11638
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.240016158431446e-06,
+      "loss": 0.9608,
+      "step": 11639
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2393209919605925e-06,
+      "loss": 1.0826,
+      "step": 11640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.23862582085295e-06,
+      "loss": 0.7766,
+      "step": 11641
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.237930645121985e-06,
+      "loss": 0.9484,
+      "step": 11642
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.237235464781167e-06,
+      "loss": 0.984,
+      "step": 11643
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.236540279843968e-06,
+      "loss": 1.059,
+      "step": 11644
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.235845090323852e-06,
+      "loss": 0.9086,
+      "step": 11645
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2351498962342925e-06,
+      "loss": 0.8387,
+      "step": 11646
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.234454697588758e-06,
+      "loss": 0.8853,
+      "step": 11647
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.233759494400716e-06,
+      "loss": 0.9846,
+      "step": 11648
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.233064286683637e-06,
+      "loss": 0.7858,
+      "step": 11649
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.232369074450988e-06,
+      "loss": 0.754,
+      "step": 11650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.231673857716244e-06,
+      "loss": 0.8782,
+      "step": 11651
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.23097863649287e-06,
+      "loss": 0.917,
+      "step": 11652
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.230283410794338e-06,
+      "loss": 0.9192,
+      "step": 11653
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2295881806341174e-06,
+      "loss": 0.7096,
+      "step": 11654
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2288929460256774e-06,
+      "loss": 0.9008,
+      "step": 11655
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.228197706982487e-06,
+      "loss": 1.0818,
+      "step": 11656
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.22750246351802e-06,
+      "loss": 0.9374,
+      "step": 11657
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.226807215645744e-06,
+      "loss": 0.9009,
+      "step": 11658
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.226111963379129e-06,
+      "loss": 0.977,
+      "step": 11659
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.225416706731646e-06,
+      "loss": 0.8998,
+      "step": 11660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.224721445716767e-06,
+      "loss": 0.861,
+      "step": 11661
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.224026180347959e-06,
+      "loss": 1.058,
+      "step": 11662
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.223330910638697e-06,
+      "loss": 0.828,
+      "step": 11663
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.222635636602447e-06,
+      "loss": 0.8289,
+      "step": 11664
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.221940358252683e-06,
+      "loss": 0.8882,
+      "step": 11665
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.221245075602874e-06,
+      "loss": 1.0042,
+      "step": 11666
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.220549788666492e-06,
+      "loss": 1.0589,
+      "step": 11667
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2198544974570075e-06,
+      "loss": 0.9131,
+      "step": 11668
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.219159201987893e-06,
+      "loss": 1.0026,
+      "step": 11669
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.218463902272619e-06,
+      "loss": 0.7898,
+      "step": 11670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.217768598324656e-06,
+      "loss": 1.1177,
+      "step": 11671
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.217073290157476e-06,
+      "loss": 1.1228,
+      "step": 11672
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2163779777845505e-06,
+      "loss": 0.8582,
+      "step": 11673
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.215682661219348e-06,
+      "loss": 1.0652,
+      "step": 11674
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.214987340475346e-06,
+      "loss": 0.89,
+      "step": 11675
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.214292015566012e-06,
+      "loss": 1.0959,
+      "step": 11676
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.213596686504819e-06,
+      "loss": 0.9525,
+      "step": 11677
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.212901353305238e-06,
+      "loss": 0.7875,
+      "step": 11678
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.212206015980742e-06,
+      "loss": 1.0599,
+      "step": 11679
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.211510674544803e-06,
+      "loss": 0.8692,
+      "step": 11680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.210815329010892e-06,
+      "loss": 1.0986,
+      "step": 11681
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.210119979392483e-06,
+      "loss": 0.9153,
+      "step": 11682
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.209424625703045e-06,
+      "loss": 1.1539,
+      "step": 11683
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.208729267956054e-06,
+      "loss": 0.8729,
+      "step": 11684
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.20803390616498e-06,
+      "loss": 0.9857,
+      "step": 11685
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.207338540343296e-06,
+      "loss": 0.8406,
+      "step": 11686
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.206643170504476e-06,
+      "loss": 0.8452,
+      "step": 11687
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.205947796661991e-06,
+      "loss": 0.9146,
+      "step": 11688
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.205252418829314e-06,
+      "loss": 0.9863,
+      "step": 11689
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.204557037019918e-06,
+      "loss": 0.9515,
+      "step": 11690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.203861651247276e-06,
+      "loss": 0.9397,
+      "step": 11691
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.203166261524861e-06,
+      "loss": 0.7759,
+      "step": 11692
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.202470867866146e-06,
+      "loss": 0.9803,
+      "step": 11693
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.201775470284603e-06,
+      "loss": 0.9062,
+      "step": 11694
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2010800687937065e-06,
+      "loss": 1.0522,
+      "step": 11695
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.200384663406929e-06,
+      "loss": 1.0003,
+      "step": 11696
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.199689254137743e-06,
+      "loss": 0.8228,
+      "step": 11697
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1989938409996245e-06,
+      "loss": 0.7974,
+      "step": 11698
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.198298424006045e-06,
+      "loss": 0.9656,
+      "step": 11699
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.19760300317048e-06,
+      "loss": 1.1482,
+      "step": 11700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.196907578506399e-06,
+      "loss": 1.0868,
+      "step": 11701
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.19621215002728e-06,
+      "loss": 0.9796,
+      "step": 11702
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1955167177465935e-06,
+      "loss": 0.8558,
+      "step": 11703
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.194821281677816e-06,
+      "loss": 1.0137,
+      "step": 11704
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.194125841834421e-06,
+      "loss": 0.8916,
+      "step": 11705
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.193430398229882e-06,
+      "loss": 0.8645,
+      "step": 11706
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.192734950877673e-06,
+      "loss": 0.9497,
+      "step": 11707
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.192039499791265e-06,
+      "loss": 0.8955,
+      "step": 11708
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.191344044984138e-06,
+      "loss": 0.9591,
+      "step": 11709
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.190648586469763e-06,
+      "loss": 0.9713,
+      "step": 11710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1899531242616155e-06,
+      "loss": 0.7757,
+      "step": 11711
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.189257658373169e-06,
+      "loss": 0.7157,
+      "step": 11712
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.188562188817899e-06,
+      "loss": 1.013,
+      "step": 11713
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.187866715609276e-06,
+      "loss": 0.9805,
+      "step": 11714
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1871712387607806e-06,
+      "loss": 0.7727,
+      "step": 11715
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.186475758285885e-06,
+      "loss": 0.8846,
+      "step": 11716
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.185780274198065e-06,
+      "loss": 1.0344,
+      "step": 11717
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.185084786510792e-06,
+      "loss": 0.8722,
+      "step": 11718
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.184389295237544e-06,
+      "loss": 1.0322,
+      "step": 11719
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1836938003917955e-06,
+      "loss": 1.0151,
+      "step": 11720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.182998301987022e-06,
+      "loss": 0.9651,
+      "step": 11721
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.182302800036696e-06,
+      "loss": 1.0487,
+      "step": 11722
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.181607294554296e-06,
+      "loss": 0.9681,
+      "step": 11723
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.180911785553295e-06,
+      "loss": 1.0166,
+      "step": 11724
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1802162730471704e-06,
+      "loss": 1.0788,
+      "step": 11725
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1795207570493955e-06,
+      "loss": 0.8484,
+      "step": 11726
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1788252375734475e-06,
+      "loss": 0.8871,
+      "step": 11727
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1781297146328e-06,
+      "loss": 0.9449,
+      "step": 11728
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.177434188240931e-06,
+      "loss": 1.035,
+      "step": 11729
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.176738658411314e-06,
+      "loss": 1.01,
+      "step": 11730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.176043125157426e-06,
+      "loss": 0.8301,
+      "step": 11731
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.175347588492741e-06,
+      "loss": 0.9713,
+      "step": 11732
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.174652048430737e-06,
+      "loss": 0.9391,
+      "step": 11733
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.17395650498489e-06,
+      "loss": 0.9183,
+      "step": 11734
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.173260958168674e-06,
+      "loss": 0.964,
+      "step": 11735
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.172565407995568e-06,
+      "loss": 0.9085,
+      "step": 11736
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.171869854479044e-06,
+      "loss": 0.87,
+      "step": 11737
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.171174297632582e-06,
+      "loss": 1.0721,
+      "step": 11738
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.170478737469656e-06,
+      "loss": 1.1188,
+      "step": 11739
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.169783174003744e-06,
+      "loss": 0.8407,
+      "step": 11740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.169087607248322e-06,
+      "loss": 0.8999,
+      "step": 11741
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.168392037216865e-06,
+      "loss": 0.8866,
+      "step": 11742
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.16769646392285e-06,
+      "loss": 0.8214,
+      "step": 11743
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.167000887379755e-06,
+      "loss": 0.8646,
+      "step": 11744
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.166305307601055e-06,
+      "loss": 1.0537,
+      "step": 11745
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.165609724600228e-06,
+      "loss": 1.0336,
+      "step": 11746
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.164914138390751e-06,
+      "loss": 0.9634,
+      "step": 11747
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.164218548986098e-06,
+      "loss": 0.7881,
+      "step": 11748
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.163522956399748e-06,
+      "loss": 0.8554,
+      "step": 11749
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.162827360645179e-06,
+      "loss": 0.9442,
+      "step": 11750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.162131761735867e-06,
+      "loss": 0.9014,
+      "step": 11751
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.161436159685288e-06,
+      "loss": 0.9366,
+      "step": 11752
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.160740554506921e-06,
+      "loss": 0.8947,
+      "step": 11753
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.160044946214241e-06,
+      "loss": 1.1585,
+      "step": 11754
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.159349334820728e-06,
+      "loss": 1.0447,
+      "step": 11755
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.158653720339857e-06,
+      "loss": 0.9489,
+      "step": 11756
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.157958102785106e-06,
+      "loss": 0.9303,
+      "step": 11757
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.157262482169953e-06,
+      "loss": 0.9664,
+      "step": 11758
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.156566858507874e-06,
+      "loss": 0.8926,
+      "step": 11759
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.155871231812348e-06,
+      "loss": 1.0577,
+      "step": 11760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.155175602096853e-06,
+      "loss": 0.871,
+      "step": 11761
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.154479969374865e-06,
+      "loss": 0.8997,
+      "step": 11762
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.153784333659863e-06,
+      "loss": 0.963,
+      "step": 11763
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.153088694965325e-06,
+      "loss": 0.9587,
+      "step": 11764
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.152393053304727e-06,
+      "loss": 1.167,
+      "step": 11765
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.151697408691549e-06,
+      "loss": 1.142,
+      "step": 11766
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.151001761139267e-06,
+      "loss": 0.9334,
+      "step": 11767
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.150306110661362e-06,
+      "loss": 1.0675,
+      "step": 11768
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.149610457271309e-06,
+      "loss": 1.0114,
+      "step": 11769
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.148914800982588e-06,
+      "loss": 0.8395,
+      "step": 11770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.148219141808677e-06,
+      "loss": 0.8133,
+      "step": 11771
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.147523479763052e-06,
+      "loss": 0.9069,
+      "step": 11772
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.146827814859193e-06,
+      "loss": 0.8434,
+      "step": 11773
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1461321471105806e-06,
+      "loss": 0.9966,
+      "step": 11774
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1454364765306895e-06,
+      "loss": 0.8647,
+      "step": 11775
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.144740803133e-06,
+      "loss": 0.8559,
+      "step": 11776
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.144045126930991e-06,
+      "loss": 0.9239,
+      "step": 11777
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.143349447938139e-06,
+      "loss": 0.9785,
+      "step": 11778
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1426537661679236e-06,
+      "loss": 1.038,
+      "step": 11779
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.141958081633825e-06,
+      "loss": 1.0964,
+      "step": 11780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1412623943493214e-06,
+      "loss": 0.9835,
+      "step": 11781
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.14056670432789e-06,
+      "loss": 0.8816,
+      "step": 11782
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.139871011583011e-06,
+      "loss": 1.0282,
+      "step": 11783
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1391753161281625e-06,
+      "loss": 0.8964,
+      "step": 11784
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.138479617976824e-06,
+      "loss": 0.806,
+      "step": 11785
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.137783917142474e-06,
+      "loss": 0.8273,
+      "step": 11786
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.137088213638592e-06,
+      "loss": 0.9787,
+      "step": 11787
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.136392507478657e-06,
+      "loss": 0.9642,
+      "step": 11788
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1356967986761485e-06,
+      "loss": 0.839,
+      "step": 11789
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1350010872445445e-06,
+      "loss": 0.8494,
+      "step": 11790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.134305373197325e-06,
+      "loss": 0.9002,
+      "step": 11791
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.13360965654797e-06,
+      "loss": 0.8677,
+      "step": 11792
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.132913937309958e-06,
+      "loss": 1.0458,
+      "step": 11793
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.132218215496769e-06,
+      "loss": 0.9184,
+      "step": 11794
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1315224911218816e-06,
+      "loss": 0.9682,
+      "step": 11795
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.130826764198775e-06,
+      "loss": 1.0585,
+      "step": 11796
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.13013103474093e-06,
+      "loss": 0.9583,
+      "step": 11797
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.129435302761827e-06,
+      "loss": 0.9443,
+      "step": 11798
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1287395682749444e-06,
+      "loss": 0.8287,
+      "step": 11799
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1280438312937605e-06,
+      "loss": 0.9866,
+      "step": 11800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.127348091831755e-06,
+      "loss": 1.0305,
+      "step": 11801
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.126652349902412e-06,
+      "loss": 0.7781,
+      "step": 11802
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.125956605519208e-06,
+      "loss": 1.0359,
+      "step": 11803
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.125260858695622e-06,
+      "loss": 1.0913,
+      "step": 11804
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1245651094451366e-06,
+      "loss": 1.018,
+      "step": 11805
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.12386935778123e-06,
+      "loss": 1.0841,
+      "step": 11806
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.123173603717381e-06,
+      "loss": 1.0457,
+      "step": 11807
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.122477847267074e-06,
+      "loss": 0.8491,
+      "step": 11808
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.121782088443785e-06,
+      "loss": 1.003,
+      "step": 11809
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.121086327260997e-06,
+      "loss": 0.975,
+      "step": 11810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.120390563732188e-06,
+      "loss": 0.9204,
+      "step": 11811
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.119694797870839e-06,
+      "loss": 1.0172,
+      "step": 11812
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.118999029690431e-06,
+      "loss": 0.978,
+      "step": 11813
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.118303259204443e-06,
+      "loss": 0.8563,
+      "step": 11814
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.117607486426357e-06,
+      "loss": 1.0499,
+      "step": 11815
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1169117113696535e-06,
+      "loss": 1.0451,
+      "step": 11816
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.11621593404781e-06,
+      "loss": 1.022,
+      "step": 11817
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.11552015447431e-06,
+      "loss": 1.0621,
+      "step": 11818
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.114824372662633e-06,
+      "loss": 0.9123,
+      "step": 11819
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1141285886262596e-06,
+      "loss": 0.9167,
+      "step": 11820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.113432802378671e-06,
+      "loss": 0.9724,
+      "step": 11821
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.112737013933348e-06,
+      "loss": 1.0963,
+      "step": 11822
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.112041223303771e-06,
+      "loss": 1.0728,
+      "step": 11823
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.111345430503421e-06,
+      "loss": 0.9641,
+      "step": 11824
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.110649635545776e-06,
+      "loss": 0.9552,
+      "step": 11825
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.109953838444321e-06,
+      "loss": 0.9426,
+      "step": 11826
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.109258039212536e-06,
+      "loss": 1.0674,
+      "step": 11827
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.108562237863901e-06,
+      "loss": 1.0218,
+      "step": 11828
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1078664344118976e-06,
+      "loss": 0.9757,
+      "step": 11829
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.107170628870006e-06,
+      "loss": 0.8762,
+      "step": 11830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.106474821251707e-06,
+      "loss": 1.0173,
+      "step": 11831
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.105779011570483e-06,
+      "loss": 0.9716,
+      "step": 11832
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1050831998398156e-06,
+      "loss": 0.8725,
+      "step": 11833
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.104387386073184e-06,
+      "loss": 0.8394,
+      "step": 11834
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.103691570284072e-06,
+      "loss": 0.7562,
+      "step": 11835
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.102995752485956e-06,
+      "loss": 0.9595,
+      "step": 11836
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.102299932692324e-06,
+      "loss": 0.8862,
+      "step": 11837
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1016041109166526e-06,
+      "loss": 0.982,
+      "step": 11838
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.100908287172424e-06,
+      "loss": 0.9819,
+      "step": 11839
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.100212461473122e-06,
+      "loss": 0.8065,
+      "step": 11840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.099516633832226e-06,
+      "loss": 1.0441,
+      "step": 11841
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.098820804263216e-06,
+      "loss": 1.0404,
+      "step": 11842
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.098124972779577e-06,
+      "loss": 0.9328,
+      "step": 11843
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.097429139394788e-06,
+      "loss": 1.0739,
+      "step": 11844
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.096733304122332e-06,
+      "loss": 1.1227,
+      "step": 11845
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.096037466975691e-06,
+      "loss": 0.9449,
+      "step": 11846
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.095341627968345e-06,
+      "loss": 0.9623,
+      "step": 11847
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0946457871137765e-06,
+      "loss": 1.0658,
+      "step": 11848
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.093949944425467e-06,
+      "loss": 1.0285,
+      "step": 11849
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.093254099916899e-06,
+      "loss": 1.0066,
+      "step": 11850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0925582536015546e-06,
+      "loss": 0.933,
+      "step": 11851
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.091862405492914e-06,
+      "loss": 0.9617,
+      "step": 11852
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.091166555604461e-06,
+      "loss": 0.99,
+      "step": 11853
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.090470703949676e-06,
+      "loss": 1.017,
+      "step": 11854
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.089774850542042e-06,
+      "loss": 0.9141,
+      "step": 11855
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.08907899539504e-06,
+      "loss": 0.9096,
+      "step": 11856
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.088383138522154e-06,
+      "loss": 0.8853,
+      "step": 11857
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.087687279936864e-06,
+      "loss": 0.9919,
+      "step": 11858
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0869914196526525e-06,
+      "loss": 1.0667,
+      "step": 11859
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0862955576830006e-06,
+      "loss": 0.9112,
+      "step": 11860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0855996940413945e-06,
+      "loss": 0.8041,
+      "step": 11861
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.084903828741312e-06,
+      "loss": 0.9382,
+      "step": 11862
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.084207961796238e-06,
+      "loss": 0.9994,
+      "step": 11863
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0835120932196545e-06,
+      "loss": 1.1331,
+      "step": 11864
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.082816223025041e-06,
+      "loss": 0.9445,
+      "step": 11865
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.082120351225883e-06,
+      "loss": 0.8936,
+      "step": 11866
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.081424477835662e-06,
+      "loss": 0.9298,
+      "step": 11867
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.080728602867861e-06,
+      "loss": 0.8882,
+      "step": 11868
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.080032726335962e-06,
+      "loss": 0.8017,
+      "step": 11869
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0793368482534454e-06,
+      "loss": 0.8952,
+      "step": 11870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0786409686337944e-06,
+      "loss": 0.9593,
+      "step": 11871
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.077945087490495e-06,
+      "loss": 0.9151,
+      "step": 11872
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.077249204837026e-06,
+      "loss": 0.8602,
+      "step": 11873
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0765533206868725e-06,
+      "loss": 0.9577,
+      "step": 11874
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.075857435053514e-06,
+      "loss": 1.0068,
+      "step": 11875
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.075161547950436e-06,
+      "loss": 0.8773,
+      "step": 11876
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.07446565939112e-06,
+      "loss": 0.9277,
+      "step": 11877
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0737697693890485e-06,
+      "loss": 0.8276,
+      "step": 11878
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0730738779577035e-06,
+      "loss": 0.9525,
+      "step": 11879
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.07237798511057e-06,
+      "loss": 0.9604,
+      "step": 11880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.071682090861129e-06,
+      "loss": 0.8336,
+      "step": 11881
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.070986195222865e-06,
+      "loss": 1.0068,
+      "step": 11882
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0702902982092575e-06,
+      "loss": 0.8543,
+      "step": 11883
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0695943998337915e-06,
+      "loss": 0.8602,
+      "step": 11884
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.068898500109951e-06,
+      "loss": 0.7842,
+      "step": 11885
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.068202599051217e-06,
+      "loss": 0.9801,
+      "step": 11886
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.067506696671074e-06,
+      "loss": 0.936,
+      "step": 11887
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0668107929830034e-06,
+      "loss": 0.8473,
+      "step": 11888
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.066114888000487e-06,
+      "loss": 0.9321,
+      "step": 11889
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.065418981737013e-06,
+      "loss": 0.8256,
+      "step": 11890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.064723074206059e-06,
+      "loss": 0.9374,
+      "step": 11891
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.06402716542111e-06,
+      "loss": 0.9283,
+      "step": 11892
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.063331255395651e-06,
+      "loss": 1.0164,
+      "step": 11893
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.062635344143161e-06,
+      "loss": 0.9422,
+      "step": 11894
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.061939431677127e-06,
+      "loss": 0.9194,
+      "step": 11895
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.06124351801103e-06,
+      "loss": 0.8697,
+      "step": 11896
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.060547603158355e-06,
+      "loss": 0.81,
+      "step": 11897
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.059851687132582e-06,
+      "loss": 1.0488,
+      "step": 11898
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.059155769947197e-06,
+      "loss": 0.8694,
+      "step": 11899
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.058459851615681e-06,
+      "loss": 0.8966,
+      "step": 11900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.057763932151521e-06,
+      "loss": 0.885,
+      "step": 11901
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.057068011568196e-06,
+      "loss": 0.9354,
+      "step": 11902
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0563720898791925e-06,
+      "loss": 0.9746,
+      "step": 11903
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.055676167097993e-06,
+      "loss": 1.0275,
+      "step": 11904
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.054980243238079e-06,
+      "loss": 0.9929,
+      "step": 11905
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0542843183129364e-06,
+      "loss": 0.944,
+      "step": 11906
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.053588392336046e-06,
+      "loss": 0.8307,
+      "step": 11907
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.052892465320895e-06,
+      "loss": 0.8424,
+      "step": 11908
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.052196537280963e-06,
+      "loss": 0.9589,
+      "step": 11909
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.051500608229734e-06,
+      "loss": 0.917,
+      "step": 11910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.050804678180694e-06,
+      "loss": 0.9331,
+      "step": 11911
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.050108747147323e-06,
+      "loss": 1.0375,
+      "step": 11912
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.049412815143108e-06,
+      "loss": 0.8179,
+      "step": 11913
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.04871688218153e-06,
+      "loss": 0.918,
+      "step": 11914
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.048020948276074e-06,
+      "loss": 0.9598,
+      "step": 11915
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.047325013440223e-06,
+      "loss": 1.0434,
+      "step": 11916
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.046629077687461e-06,
+      "loss": 0.7903,
+      "step": 11917
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.045933141031269e-06,
+      "loss": 0.8269,
+      "step": 11918
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.045237203485136e-06,
+      "loss": 0.8022,
+      "step": 11919
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0445412650625405e-06,
+      "loss": 0.9767,
+      "step": 11920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.04384532577697e-06,
+      "loss": 1.0203,
+      "step": 11921
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0431493856419035e-06,
+      "loss": 0.9808,
+      "step": 11922
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.042453444670829e-06,
+      "loss": 1.0838,
+      "step": 11923
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.041757502877228e-06,
+      "loss": 0.8656,
+      "step": 11924
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.041061560274585e-06,
+      "loss": 0.8751,
+      "step": 11925
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.040365616876383e-06,
+      "loss": 0.9008,
+      "step": 11926
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.039669672696108e-06,
+      "loss": 0.9181,
+      "step": 11927
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.038973727747242e-06,
+      "loss": 0.9914,
+      "step": 11928
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.038277782043267e-06,
+      "loss": 1.0842,
+      "step": 11929
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.03758183559767e-06,
+      "loss": 0.8884,
+      "step": 11930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.036885888423934e-06,
+      "loss": 0.9684,
+      "step": 11931
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0361899405355405e-06,
+      "loss": 1.0068,
+      "step": 11932
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.035493991945977e-06,
+      "loss": 0.9058,
+      "step": 11933
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.034798042668725e-06,
+      "loss": 0.911,
+      "step": 11934
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.034102092717267e-06,
+      "loss": 1.0688,
+      "step": 11935
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.033406142105091e-06,
+      "loss": 0.8637,
+      "step": 11936
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.032710190845678e-06,
+      "loss": 0.8005,
+      "step": 11937
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.032014238952512e-06,
+      "loss": 1.0124,
+      "step": 11938
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.031318286439077e-06,
+      "loss": 0.8834,
+      "step": 11939
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.030622333318859e-06,
+      "loss": 0.8189,
+      "step": 11940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.029926379605339e-06,
+      "loss": 0.9685,
+      "step": 11941
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.029230425312002e-06,
+      "loss": 0.9037,
+      "step": 11942
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0285344704523335e-06,
+      "loss": 0.9734,
+      "step": 11943
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.027838515039814e-06,
+      "loss": 1.1043,
+      "step": 11944
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.027142559087931e-06,
+      "loss": 0.8614,
+      "step": 11945
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.026446602610166e-06,
+      "loss": 0.9111,
+      "step": 11946
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.025750645620004e-06,
+      "loss": 1.0595,
+      "step": 11947
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0250546881309305e-06,
+      "loss": 0.999,
+      "step": 11948
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.024358730156428e-06,
+      "loss": 0.8294,
+      "step": 11949
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.023662771709979e-06,
+      "loss": 0.8468,
+      "step": 11950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.022966812805071e-06,
+      "loss": 0.9335,
+      "step": 11951
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0222708534551856e-06,
+      "loss": 1.1281,
+      "step": 11952
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.021574893673806e-06,
+      "loss": 0.872,
+      "step": 11953
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.02087893347442e-06,
+      "loss": 0.9005,
+      "step": 11954
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.020182972870509e-06,
+      "loss": 1.1176,
+      "step": 11955
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.019487011875557e-06,
+      "loss": 0.958,
+      "step": 11956
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.018791050503048e-06,
+      "loss": 0.9706,
+      "step": 11957
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.018095088766467e-06,
+      "loss": 1.012,
+      "step": 11958
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.017399126679298e-06,
+      "loss": 0.8842,
+      "step": 11959
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0167031642550245e-06,
+      "loss": 0.9284,
+      "step": 11960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0160072015071314e-06,
+      "loss": 0.8981,
+      "step": 11961
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.015311238449102e-06,
+      "loss": 0.7738,
+      "step": 11962
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.014615275094422e-06,
+      "loss": 0.9199,
+      "step": 11963
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.013919311456572e-06,
+      "loss": 0.9382,
+      "step": 11964
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.01322334754904e-06,
+      "loss": 0.8205,
+      "step": 11965
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.012527383385309e-06,
+      "loss": 1.0598,
+      "step": 11966
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.011831418978862e-06,
+      "loss": 0.9499,
+      "step": 11967
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.011135454343184e-06,
+      "loss": 0.9902,
+      "step": 11968
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.010439489491761e-06,
+      "loss": 0.8936,
+      "step": 11969
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.009743524438073e-06,
+      "loss": 0.9808,
+      "step": 11970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.009047559195608e-06,
+      "loss": 1.0469,
+      "step": 11971
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.008351593777847e-06,
+      "loss": 0.8776,
+      "step": 11972
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.007655628198277e-06,
+      "loss": 0.8515,
+      "step": 11973
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.006959662470381e-06,
+      "loss": 1.0846,
+      "step": 11974
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.006263696607641e-06,
+      "loss": 1.1064,
+      "step": 11975
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0055677306235454e-06,
+      "loss": 0.9226,
+      "step": 11976
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.004871764531576e-06,
+      "loss": 0.723,
+      "step": 11977
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.004175798345217e-06,
+      "loss": 0.8639,
+      "step": 11978
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.003479832077952e-06,
+      "loss": 0.9375,
+      "step": 11979
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.002783865743267e-06,
+      "loss": 1.1142,
+      "step": 11980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.002087899354646e-06,
+      "loss": 0.7992,
+      "step": 11981
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.00139193292557e-06,
+      "loss": 0.8456,
+      "step": 11982
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0006959664695275e-06,
+      "loss": 1.076,
+      "step": 11983
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5e-06,
+      "loss": 0.9831,
+      "step": 11984
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.999304033530474e-06,
+      "loss": 0.9343,
+      "step": 11985
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.998608067074432e-06,
+      "loss": 0.9537,
+      "step": 11986
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.997912100645356e-06,
+      "loss": 1.1025,
+      "step": 11987
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.997216134256735e-06,
+      "loss": 1.1001,
+      "step": 11988
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.99652016792205e-06,
+      "loss": 1.0032,
+      "step": 11989
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.995824201654783e-06,
+      "loss": 0.9403,
+      "step": 11990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.995128235468425e-06,
+      "loss": 0.8108,
+      "step": 11991
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.994432269376455e-06,
+      "loss": 0.9098,
+      "step": 11992
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9937363033923594e-06,
+      "loss": 0.8986,
+      "step": 11993
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.993040337529621e-06,
+      "loss": 0.9758,
+      "step": 11994
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.992344371801725e-06,
+      "loss": 0.9459,
+      "step": 11995
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.991648406222154e-06,
+      "loss": 0.8846,
+      "step": 11996
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.990952440804394e-06,
+      "loss": 1.0163,
+      "step": 11997
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9902564755619285e-06,
+      "loss": 1.0264,
+      "step": 11998
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9895605105082415e-06,
+      "loss": 0.979,
+      "step": 11999
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.988864545656817e-06,
+      "loss": 0.8296,
+      "step": 12000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.988168581021139e-06,
+      "loss": 1.0119,
+      "step": 12001
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.987472616614692e-06,
+      "loss": 0.8588,
+      "step": 12002
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.98677665245096e-06,
+      "loss": 0.8612,
+      "step": 12003
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.986080688543428e-06,
+      "loss": 0.8836,
+      "step": 12004
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.98538472490558e-06,
+      "loss": 0.9676,
+      "step": 12005
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.984688761550899e-06,
+      "loss": 0.8452,
+      "step": 12006
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.98399279849287e-06,
+      "loss": 0.9838,
+      "step": 12007
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.983296835744977e-06,
+      "loss": 0.8501,
+      "step": 12008
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.982600873320704e-06,
+      "loss": 1.0098,
+      "step": 12009
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.981904911233535e-06,
+      "loss": 0.8555,
+      "step": 12010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.981208949496954e-06,
+      "loss": 0.9865,
+      "step": 12011
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.980512988124445e-06,
+      "loss": 0.9792,
+      "step": 12012
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.979817027129494e-06,
+      "loss": 0.9528,
+      "step": 12013
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.97912106652558e-06,
+      "loss": 1.14,
+      "step": 12014
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.978425106326193e-06,
+      "loss": 0.81,
+      "step": 12015
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.977729146544815e-06,
+      "loss": 0.9134,
+      "step": 12016
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.97703318719493e-06,
+      "loss": 0.8579,
+      "step": 12017
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.976337228290022e-06,
+      "loss": 1.0002,
+      "step": 12018
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.975641269843574e-06,
+      "loss": 0.9575,
+      "step": 12019
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.97494531186907e-06,
+      "loss": 1.0575,
+      "step": 12020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.974249354379997e-06,
+      "loss": 0.9558,
+      "step": 12021
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9735533973898354e-06,
+      "loss": 0.8692,
+      "step": 12022
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.972857440912072e-06,
+      "loss": 1.0171,
+      "step": 12023
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.972161484960188e-06,
+      "loss": 1.0053,
+      "step": 12024
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.971465529547669e-06,
+      "loss": 0.8748,
+      "step": 12025
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.970769574687999e-06,
+      "loss": 0.8801,
+      "step": 12026
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.970073620394663e-06,
+      "loss": 0.9732,
+      "step": 12027
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.969377666681143e-06,
+      "loss": 0.9454,
+      "step": 12028
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.968681713560924e-06,
+      "loss": 0.8306,
+      "step": 12029
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.96798576104749e-06,
+      "loss": 0.9301,
+      "step": 12030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.967289809154324e-06,
+      "loss": 1.0462,
+      "step": 12031
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.966593857894911e-06,
+      "loss": 1.0754,
+      "step": 12032
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.965897907282734e-06,
+      "loss": 0.9716,
+      "step": 12033
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.965201957331277e-06,
+      "loss": 0.8206,
+      "step": 12034
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.964506008054024e-06,
+      "loss": 0.8858,
+      "step": 12035
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.96381005946446e-06,
+      "loss": 0.9447,
+      "step": 12036
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.963114111576067e-06,
+      "loss": 0.8881,
+      "step": 12037
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.96241816440233e-06,
+      "loss": 0.9263,
+      "step": 12038
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.961722217956733e-06,
+      "loss": 0.9103,
+      "step": 12039
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.96102627225276e-06,
+      "loss": 1.0242,
+      "step": 12040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.960330327303893e-06,
+      "loss": 0.9283,
+      "step": 12041
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9596343831236175e-06,
+      "loss": 0.976,
+      "step": 12042
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.958938439725416e-06,
+      "loss": 0.7406,
+      "step": 12043
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.958242497122774e-06,
+      "loss": 0.9135,
+      "step": 12044
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.957546555329174e-06,
+      "loss": 0.968,
+      "step": 12045
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.956850614358098e-06,
+      "loss": 0.9541,
+      "step": 12046
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.956154674223034e-06,
+      "loss": 1.0764,
+      "step": 12047
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9554587349374595e-06,
+      "loss": 0.9619,
+      "step": 12048
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9547627965148644e-06,
+      "loss": 0.8422,
+      "step": 12049
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.954066858968731e-06,
+      "loss": 1.001,
+      "step": 12050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.95337092231254e-06,
+      "loss": 1.0712,
+      "step": 12051
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9526749865597775e-06,
+      "loss": 0.917,
+      "step": 12052
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.951979051723927e-06,
+      "loss": 1.0233,
+      "step": 12053
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.951283117818471e-06,
+      "loss": 0.7954,
+      "step": 12054
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.950587184856894e-06,
+      "loss": 0.8572,
+      "step": 12055
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.949891252852679e-06,
+      "loss": 1.0089,
+      "step": 12056
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.949195321819309e-06,
+      "loss": 0.9738,
+      "step": 12057
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.948499391770268e-06,
+      "loss": 1.0061,
+      "step": 12058
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.94780346271904e-06,
+      "loss": 0.9637,
+      "step": 12059
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.947107534679108e-06,
+      "loss": 1.1836,
+      "step": 12060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.946411607663955e-06,
+      "loss": 1.016,
+      "step": 12061
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.945715681687065e-06,
+      "loss": 1.137,
+      "step": 12062
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.945019756761922e-06,
+      "loss": 0.9164,
+      "step": 12063
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.944323832902009e-06,
+      "loss": 0.9734,
+      "step": 12064
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.943627910120809e-06,
+      "loss": 0.9978,
+      "step": 12065
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.942931988431805e-06,
+      "loss": 0.9261,
+      "step": 12066
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.942236067848481e-06,
+      "loss": 0.874,
+      "step": 12067
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9415401483843204e-06,
+      "loss": 0.8971,
+      "step": 12068
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.940844230052806e-06,
+      "loss": 1.0909,
+      "step": 12069
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9401483128674205e-06,
+      "loss": 1.1055,
+      "step": 12070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.939452396841649e-06,
+      "loss": 0.808,
+      "step": 12071
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.93875648198897e-06,
+      "loss": 0.9618,
+      "step": 12072
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.938060568322874e-06,
+      "loss": 0.925,
+      "step": 12073
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.93736465585684e-06,
+      "loss": 0.8624,
+      "step": 12074
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.93666874460435e-06,
+      "loss": 1.0165,
+      "step": 12075
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9359728345788905e-06,
+      "loss": 0.9015,
+      "step": 12076
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.935276925793943e-06,
+      "loss": 1.0649,
+      "step": 12077
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.934581018262989e-06,
+      "loss": 0.9514,
+      "step": 12078
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.933885111999514e-06,
+      "loss": 0.9057,
+      "step": 12079
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.933189207017e-06,
+      "loss": 0.8499,
+      "step": 12080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.932493303328929e-06,
+      "loss": 1.0279,
+      "step": 12081
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.931797400948785e-06,
+      "loss": 0.9715,
+      "step": 12082
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.93110149989005e-06,
+      "loss": 0.693,
+      "step": 12083
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9304056001662084e-06,
+      "loss": 1.1574,
+      "step": 12084
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.929709701790743e-06,
+      "loss": 0.788,
+      "step": 12085
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.929013804777137e-06,
+      "loss": 0.8181,
+      "step": 12086
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9283179091388714e-06,
+      "loss": 0.8942,
+      "step": 12087
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.927622014889431e-06,
+      "loss": 0.8868,
+      "step": 12088
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.926926122042297e-06,
+      "loss": 0.9605,
+      "step": 12089
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.926230230610953e-06,
+      "loss": 1.0392,
+      "step": 12090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.925534340608882e-06,
+      "loss": 0.9457,
+      "step": 12091
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.924838452049566e-06,
+      "loss": 0.9989,
+      "step": 12092
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.924142564946487e-06,
+      "loss": 0.937,
+      "step": 12093
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.92344667931313e-06,
+      "loss": 0.9565,
+      "step": 12094
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.922750795162974e-06,
+      "loss": 0.8721,
+      "step": 12095
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.922054912509505e-06,
+      "loss": 0.8922,
+      "step": 12096
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9213590313662055e-06,
+      "loss": 0.7084,
+      "step": 12097
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.920663151746556e-06,
+      "loss": 0.8577,
+      "step": 12098
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9199672736640405e-06,
+      "loss": 0.8174,
+      "step": 12099
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9192713971321405e-06,
+      "loss": 0.8482,
+      "step": 12100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9185755221643385e-06,
+      "loss": 0.9842,
+      "step": 12101
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.917879648774118e-06,
+      "loss": 0.987,
+      "step": 12102
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.91718377697496e-06,
+      "loss": 1.0473,
+      "step": 12103
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.916487906780348e-06,
+      "loss": 0.9001,
+      "step": 12104
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.915792038203764e-06,
+      "loss": 0.9427,
+      "step": 12105
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.91509617125869e-06,
+      "loss": 0.9563,
+      "step": 12106
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.914400305958606e-06,
+      "loss": 0.8397,
+      "step": 12107
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9137044423169986e-06,
+      "loss": 0.9021,
+      "step": 12108
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.913008580347348e-06,
+      "loss": 0.8888,
+      "step": 12109
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.912312720063138e-06,
+      "loss": 1.0877,
+      "step": 12110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.911616861477848e-06,
+      "loss": 1.0511,
+      "step": 12111
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.910921004604961e-06,
+      "loss": 0.7899,
+      "step": 12112
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.91022514945796e-06,
+      "loss": 0.8698,
+      "step": 12113
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.909529296050326e-06,
+      "loss": 0.8844,
+      "step": 12114
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.908833444395542e-06,
+      "loss": 0.8626,
+      "step": 12115
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.908137594507088e-06,
+      "loss": 0.8731,
+      "step": 12116
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.907441746398448e-06,
+      "loss": 0.9799,
+      "step": 12117
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.906745900083102e-06,
+      "loss": 0.8224,
+      "step": 12118
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.906050055574534e-06,
+      "loss": 0.9042,
+      "step": 12119
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.905354212886225e-06,
+      "loss": 0.9134,
+      "step": 12120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.904658372031657e-06,
+      "loss": 0.9235,
+      "step": 12121
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.903962533024311e-06,
+      "loss": 0.8042,
+      "step": 12122
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.903266695877669e-06,
+      "loss": 0.9203,
+      "step": 12123
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9025708606052136e-06,
+      "loss": 0.8082,
+      "step": 12124
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.901875027220425e-06,
+      "loss": 1.0347,
+      "step": 12125
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9011791957367855e-06,
+      "loss": 0.9942,
+      "step": 12126
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.900483366167777e-06,
+      "loss": 0.9947,
+      "step": 12127
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.89978753852688e-06,
+      "loss": 1.0396,
+      "step": 12128
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.899091712827578e-06,
+      "loss": 0.9468,
+      "step": 12129
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.898395889083347e-06,
+      "loss": 0.8356,
+      "step": 12130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.897700067307677e-06,
+      "loss": 1.0432,
+      "step": 12131
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.897004247514044e-06,
+      "loss": 0.6895,
+      "step": 12132
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.89630842971593e-06,
+      "loss": 0.979,
+      "step": 12133
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.895612613926818e-06,
+      "loss": 0.8884,
+      "step": 12134
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.894916800160186e-06,
+      "loss": 0.8789,
+      "step": 12135
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8942209884295185e-06,
+      "loss": 0.8818,
+      "step": 12136
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.893525178748295e-06,
+      "loss": 0.951,
+      "step": 12137
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.892829371129997e-06,
+      "loss": 0.8343,
+      "step": 12138
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.892133565588105e-06,
+      "loss": 0.9427,
+      "step": 12139
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.891437762136101e-06,
+      "loss": 1.0073,
+      "step": 12140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.890741960787464e-06,
+      "loss": 0.9187,
+      "step": 12141
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.890046161555679e-06,
+      "loss": 0.9194,
+      "step": 12142
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.889350364454225e-06,
+      "loss": 0.8217,
+      "step": 12143
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.888654569496582e-06,
+      "loss": 1.0933,
+      "step": 12144
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.887958776696231e-06,
+      "loss": 0.8303,
+      "step": 12145
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.887262986066654e-06,
+      "loss": 0.9065,
+      "step": 12146
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.88656719762133e-06,
+      "loss": 0.7848,
+      "step": 12147
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.885871411373742e-06,
+      "loss": 0.9363,
+      "step": 12148
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.885175627337369e-06,
+      "loss": 1.073,
+      "step": 12149
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.884479845525692e-06,
+      "loss": 1.1238,
+      "step": 12150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.883784065952192e-06,
+      "loss": 0.7991,
+      "step": 12151
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.88308828863035e-06,
+      "loss": 0.9661,
+      "step": 12152
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.882392513573644e-06,
+      "loss": 1.1633,
+      "step": 12153
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.881696740795558e-06,
+      "loss": 0.8596,
+      "step": 12154
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8810009703095705e-06,
+      "loss": 0.8857,
+      "step": 12155
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.880305202129162e-06,
+      "loss": 0.9003,
+      "step": 12156
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.879609436267814e-06,
+      "loss": 1.0609,
+      "step": 12157
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8789136727390045e-06,
+      "loss": 1.0995,
+      "step": 12158
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.878217911556216e-06,
+      "loss": 1.0381,
+      "step": 12159
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.877522152732928e-06,
+      "loss": 0.963,
+      "step": 12160
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.876826396282619e-06,
+      "loss": 1.0573,
+      "step": 12161
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.876130642218772e-06,
+      "loss": 0.991,
+      "step": 12162
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.875434890554866e-06,
+      "loss": 0.9337,
+      "step": 12163
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.87473914130438e-06,
+      "loss": 1.0094,
+      "step": 12164
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.874043394480793e-06,
+      "loss": 0.867,
+      "step": 12165
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.873347650097589e-06,
+      "loss": 1.0227,
+      "step": 12166
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.872651908168244e-06,
+      "loss": 1.0414,
+      "step": 12167
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.871956168706241e-06,
+      "loss": 0.9961,
+      "step": 12168
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.871260431725058e-06,
+      "loss": 1.0148,
+      "step": 12169
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.870564697238174e-06,
+      "loss": 0.9471,
+      "step": 12170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.869868965259071e-06,
+      "loss": 0.8627,
+      "step": 12171
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.869173235801226e-06,
+      "loss": 1.0538,
+      "step": 12172
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.868477508878121e-06,
+      "loss": 1.0058,
+      "step": 12173
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8677817845032335e-06,
+      "loss": 0.9791,
+      "step": 12174
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.867086062690045e-06,
+      "loss": 0.759,
+      "step": 12175
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.86639034345203e-06,
+      "loss": 1.0132,
+      "step": 12176
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.865694626802675e-06,
+      "loss": 0.8883,
+      "step": 12177
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.864998912755456e-06,
+      "loss": 1.1631,
+      "step": 12178
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.864303201323853e-06,
+      "loss": 0.8378,
+      "step": 12179
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8636074925213445e-06,
+      "loss": 1.0539,
+      "step": 12180
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.862911786361409e-06,
+      "loss": 0.7926,
+      "step": 12181
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.862216082857528e-06,
+      "loss": 0.9492,
+      "step": 12182
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.861520382023178e-06,
+      "loss": 0.8875,
+      "step": 12183
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.860824683871839e-06,
+      "loss": 1.0558,
+      "step": 12184
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.860128988416991e-06,
+      "loss": 1.0723,
+      "step": 12185
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.859433295672112e-06,
+      "loss": 0.9503,
+      "step": 12186
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.858737605650681e-06,
+      "loss": 0.8497,
+      "step": 12187
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8580419183661754e-06,
+      "loss": 1.0085,
+      "step": 12188
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.857346233832076e-06,
+      "loss": 1.0523,
+      "step": 12189
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.856650552061862e-06,
+      "loss": 0.7754,
+      "step": 12190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.855954873069011e-06,
+      "loss": 1.1059,
+      "step": 12191
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8552591968670015e-06,
+      "loss": 0.967,
+      "step": 12192
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.854563523469311e-06,
+      "loss": 0.9344,
+      "step": 12193
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.853867852889421e-06,
+      "loss": 1.068,
+      "step": 12194
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.853172185140808e-06,
+      "loss": 1.1765,
+      "step": 12195
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.85247652023695e-06,
+      "loss": 0.7739,
+      "step": 12196
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.851780858191326e-06,
+      "loss": 0.9389,
+      "step": 12197
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.851085199017414e-06,
+      "loss": 0.9214,
+      "step": 12198
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.850389542728693e-06,
+      "loss": 0.8219,
+      "step": 12199
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.849693889338638e-06,
+      "loss": 1.0345,
+      "step": 12200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.848998238860732e-06,
+      "loss": 0.8735,
+      "step": 12201
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.848302591308452e-06,
+      "loss": 0.9718,
+      "step": 12202
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.847606946695274e-06,
+      "loss": 0.9401,
+      "step": 12203
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.846911305034677e-06,
+      "loss": 0.8743,
+      "step": 12204
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.846215666340138e-06,
+      "loss": 0.8272,
+      "step": 12205
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.845520030625136e-06,
+      "loss": 1.0614,
+      "step": 12206
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.84482439790315e-06,
+      "loss": 1.0545,
+      "step": 12207
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.844128768187654e-06,
+      "loss": 0.8454,
+      "step": 12208
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.843433141492129e-06,
+      "loss": 0.9702,
+      "step": 12209
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.84273751783005e-06,
+      "loss": 0.8701,
+      "step": 12210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.842041897214896e-06,
+      "loss": 1.027,
+      "step": 12211
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.841346279660145e-06,
+      "loss": 0.9275,
+      "step": 12212
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.840650665179274e-06,
+      "loss": 0.9543,
+      "step": 12213
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.83995505378576e-06,
+      "loss": 1.2326,
+      "step": 12214
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8392594454930805e-06,
+      "loss": 0.9396,
+      "step": 12215
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.838563840314713e-06,
+      "loss": 0.8415,
+      "step": 12216
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8378682382641354e-06,
+      "loss": 1.0143,
+      "step": 12217
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8371726393548225e-06,
+      "loss": 1.0783,
+      "step": 12218
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.836477043600253e-06,
+      "loss": 0.8904,
+      "step": 12219
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.835781451013904e-06,
+      "loss": 0.9157,
+      "step": 12220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.835085861609252e-06,
+      "loss": 1.1368,
+      "step": 12221
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.834390275399774e-06,
+      "loss": 1.1214,
+      "step": 12222
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.833694692398945e-06,
+      "loss": 0.8928,
+      "step": 12223
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.832999112620246e-06,
+      "loss": 0.8698,
+      "step": 12224
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.83230353607715e-06,
+      "loss": 0.862,
+      "step": 12225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8316079627831365e-06,
+      "loss": 0.9648,
+      "step": 12226
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.83091239275168e-06,
+      "loss": 0.913,
+      "step": 12227
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.830216825996257e-06,
+      "loss": 0.9695,
+      "step": 12228
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.829521262530345e-06,
+      "loss": 0.913,
+      "step": 12229
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.82882570236742e-06,
+      "loss": 0.9679,
+      "step": 12230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.828130145520957e-06,
+      "loss": 0.8275,
+      "step": 12231
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.827434592004435e-06,
+      "loss": 0.825,
+      "step": 12232
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.826739041831328e-06,
+      "loss": 0.9466,
+      "step": 12233
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.82604349501511e-06,
+      "loss": 0.9684,
+      "step": 12234
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.825347951569263e-06,
+      "loss": 0.8878,
+      "step": 12235
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.824652411507259e-06,
+      "loss": 1.0117,
+      "step": 12236
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.823956874842575e-06,
+      "loss": 0.9486,
+      "step": 12237
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.823261341588687e-06,
+      "loss": 0.7365,
+      "step": 12238
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.82256581175907e-06,
+      "loss": 0.9421,
+      "step": 12239
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.821870285367201e-06,
+      "loss": 0.8288,
+      "step": 12240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.821174762426553e-06,
+      "loss": 0.9207,
+      "step": 12241
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.820479242950605e-06,
+      "loss": 0.9484,
+      "step": 12242
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.819783726952831e-06,
+      "loss": 0.7578,
+      "step": 12243
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.819088214446706e-06,
+      "loss": 0.837,
+      "step": 12244
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.818392705445706e-06,
+      "loss": 0.9634,
+      "step": 12245
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8176971999633045e-06,
+      "loss": 1.0293,
+      "step": 12246
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.81700169801298e-06,
+      "loss": 0.857,
+      "step": 12247
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.816306199608205e-06,
+      "loss": 0.6997,
+      "step": 12248
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.815610704762457e-06,
+      "loss": 0.8538,
+      "step": 12249
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.814915213489209e-06,
+      "loss": 1.0004,
+      "step": 12250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.814219725801938e-06,
+      "loss": 0.7173,
+      "step": 12251
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.813524241714116e-06,
+      "loss": 0.828,
+      "step": 12252
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.81282876123922e-06,
+      "loss": 1.2057,
+      "step": 12253
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8121332843907245e-06,
+      "loss": 0.9614,
+      "step": 12254
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.811437811182105e-06,
+      "loss": 1.1039,
+      "step": 12255
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.810742341626834e-06,
+      "loss": 0.8885,
+      "step": 12256
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.810046875738387e-06,
+      "loss": 1.0277,
+      "step": 12257
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8093514135302374e-06,
+      "loss": 1.0304,
+      "step": 12258
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.808655955015862e-06,
+      "loss": 1.0179,
+      "step": 12259
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.807960500208735e-06,
+      "loss": 0.9179,
+      "step": 12260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.80726504912233e-06,
+      "loss": 0.8859,
+      "step": 12261
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.80656960177012e-06,
+      "loss": 0.9511,
+      "step": 12262
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.80587415816558e-06,
+      "loss": 0.9755,
+      "step": 12263
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8051787183221845e-06,
+      "loss": 1.0588,
+      "step": 12264
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.804483282253407e-06,
+      "loss": 0.7511,
+      "step": 12265
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.803787849972722e-06,
+      "loss": 0.7764,
+      "step": 12266
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8030924214936035e-06,
+      "loss": 1.0293,
+      "step": 12267
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.802396996829524e-06,
+      "loss": 0.9382,
+      "step": 12268
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.801701575993955e-06,
+      "loss": 0.8533,
+      "step": 12269
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8010061590003755e-06,
+      "loss": 0.826,
+      "step": 12270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8003107458622574e-06,
+      "loss": 0.92,
+      "step": 12271
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7996153365930725e-06,
+      "loss": 1.0369,
+      "step": 12272
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.798919931206295e-06,
+      "loss": 0.8701,
+      "step": 12273
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7982245297153985e-06,
+      "loss": 0.921,
+      "step": 12274
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.797529132133855e-06,
+      "loss": 1.0464,
+      "step": 12275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.79683373847514e-06,
+      "loss": 0.944,
+      "step": 12276
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.796138348752725e-06,
+      "loss": 1.0496,
+      "step": 12277
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.795442962980083e-06,
+      "loss": 0.941,
+      "step": 12278
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.794747581170687e-06,
+      "loss": 1.0296,
+      "step": 12279
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.79405220333801e-06,
+      "loss": 0.9128,
+      "step": 12280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.793356829495524e-06,
+      "loss": 0.9856,
+      "step": 12281
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.792661459656704e-06,
+      "loss": 1.0135,
+      "step": 12282
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.79196609383502e-06,
+      "loss": 1.0003,
+      "step": 12283
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.791270732043947e-06,
+      "loss": 1.0787,
+      "step": 12284
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7905753742969555e-06,
+      "loss": 1.2016,
+      "step": 12285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.789880020607519e-06,
+      "loss": 0.9301,
+      "step": 12286
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7891846709891095e-06,
+      "loss": 1.0099,
+      "step": 12287
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.788489325455199e-06,
+      "loss": 0.8898,
+      "step": 12288
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.78779398401926e-06,
+      "loss": 0.8393,
+      "step": 12289
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.787098646694764e-06,
+      "loss": 0.9712,
+      "step": 12290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.786403313495184e-06,
+      "loss": 0.8573,
+      "step": 12291
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.785707984433989e-06,
+      "loss": 0.9819,
+      "step": 12292
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.785012659524655e-06,
+      "loss": 1.0114,
+      "step": 12293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.784317338780652e-06,
+      "loss": 0.8322,
+      "step": 12294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.783622022215452e-06,
+      "loss": 0.9697,
+      "step": 12295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.782926709842526e-06,
+      "loss": 1.0118,
+      "step": 12296
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.782231401675346e-06,
+      "loss": 1.0224,
+      "step": 12297
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.781536097727383e-06,
+      "loss": 1.012,
+      "step": 12298
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.780840798012108e-06,
+      "loss": 0.9123,
+      "step": 12299
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.780145502542993e-06,
+      "loss": 0.9967,
+      "step": 12300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.77945021133351e-06,
+      "loss": 0.9093,
+      "step": 12301
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.778754924397128e-06,
+      "loss": 0.9092,
+      "step": 12302
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.77805964174732e-06,
+      "loss": 0.9339,
+      "step": 12303
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.777364363397555e-06,
+      "loss": 1.0579,
+      "step": 12304
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.776669089361306e-06,
+      "loss": 0.9288,
+      "step": 12305
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.775973819652041e-06,
+      "loss": 0.9217,
+      "step": 12306
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.775278554283235e-06,
+      "loss": 1.0924,
+      "step": 12307
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.774583293268355e-06,
+      "loss": 1.0064,
+      "step": 12308
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.773888036620872e-06,
+      "loss": 0.9495,
+      "step": 12309
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7731927843542575e-06,
+      "loss": 1.0463,
+      "step": 12310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.772497536481981e-06,
+      "loss": 0.878,
+      "step": 12311
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.771802293017514e-06,
+      "loss": 1.1603,
+      "step": 12312
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.771107053974325e-06,
+      "loss": 0.7532,
+      "step": 12313
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.770411819365885e-06,
+      "loss": 0.9011,
+      "step": 12314
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.769716589205665e-06,
+      "loss": 0.9044,
+      "step": 12315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7690213635071306e-06,
+      "loss": 0.9964,
+      "step": 12316
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.768326142283757e-06,
+      "loss": 0.895,
+      "step": 12317
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.767630925549011e-06,
+      "loss": 0.9768,
+      "step": 12318
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7669357133163654e-06,
+      "loss": 0.888,
+      "step": 12319
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.766240505599286e-06,
+      "loss": 0.8845,
+      "step": 12320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.765545302411244e-06,
+      "loss": 0.9925,
+      "step": 12321
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.764850103765708e-06,
+      "loss": 0.9919,
+      "step": 12322
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.764154909676149e-06,
+      "loss": 1.1932,
+      "step": 12323
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.763459720156034e-06,
+      "loss": 0.8901,
+      "step": 12324
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7627645352188345e-06,
+      "loss": 1.0489,
+      "step": 12325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7620693548780175e-06,
+      "loss": 0.8443,
+      "step": 12326
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.761374179147051e-06,
+      "loss": 0.9616,
+      "step": 12327
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7606790080394075e-06,
+      "loss": 0.9792,
+      "step": 12328
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.759983841568554e-06,
+      "loss": 0.9415,
+      "step": 12329
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.75928867974796e-06,
+      "loss": 0.9518,
+      "step": 12330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.758593522591092e-06,
+      "loss": 1.1417,
+      "step": 12331
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.75789837011142e-06,
+      "loss": 0.9658,
+      "step": 12332
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.757203222322412e-06,
+      "loss": 1.0743,
+      "step": 12333
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.756508079237537e-06,
+      "loss": 1.0017,
+      "step": 12334
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.755812940870263e-06,
+      "loss": 0.9891,
+      "step": 12335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7551178072340576e-06,
+      "loss": 0.7912,
+      "step": 12336
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7544226783423886e-06,
+      "loss": 0.9625,
+      "step": 12337
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7537275542087256e-06,
+      "loss": 1.1196,
+      "step": 12338
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.753032434846533e-06,
+      "loss": 1.0342,
+      "step": 12339
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.752337320269283e-06,
+      "loss": 0.8509,
+      "step": 12340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7516422104904405e-06,
+      "loss": 1.1111,
+      "step": 12341
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.750947105523474e-06,
+      "loss": 1.1495,
+      "step": 12342
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.750252005381851e-06,
+      "loss": 0.8085,
+      "step": 12343
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.749556910079039e-06,
+      "loss": 0.8111,
+      "step": 12344
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7488618196285044e-06,
+      "loss": 1.037,
+      "step": 12345
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.748166734043716e-06,
+      "loss": 0.8635,
+      "step": 12346
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.747471653338139e-06,
+      "loss": 0.95,
+      "step": 12347
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.746776577525242e-06,
+      "loss": 1.0395,
+      "step": 12348
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.746081506618491e-06,
+      "loss": 0.8612,
+      "step": 12349
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.745386440631354e-06,
+      "loss": 0.9007,
+      "step": 12350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.744691379577294e-06,
+      "loss": 0.8667,
+      "step": 12351
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.743996323469783e-06,
+      "loss": 1.0045,
+      "step": 12352
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.743301272322286e-06,
+      "loss": 0.971,
+      "step": 12353
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.742606226148268e-06,
+      "loss": 0.9208,
+      "step": 12354
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.741911184961195e-06,
+      "loss": 0.9709,
+      "step": 12355
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7412161487745356e-06,
+      "loss": 1.0742,
+      "step": 12356
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.740521117601754e-06,
+      "loss": 0.9558,
+      "step": 12357
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.739826091456317e-06,
+      "loss": 0.9606,
+      "step": 12358
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.739131070351691e-06,
+      "loss": 0.9543,
+      "step": 12359
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.73843605430134e-06,
+      "loss": 1.0233,
+      "step": 12360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.737741043318732e-06,
+      "loss": 0.9395,
+      "step": 12361
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.73704603741733e-06,
+      "loss": 0.7909,
+      "step": 12362
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.736351036610603e-06,
+      "loss": 0.8729,
+      "step": 12363
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.735656040912015e-06,
+      "loss": 0.8606,
+      "step": 12364
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.734961050335031e-06,
+      "loss": 1.0581,
+      "step": 12365
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.734266064893117e-06,
+      "loss": 0.9822,
+      "step": 12366
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.733571084599737e-06,
+      "loss": 1.1641,
+      "step": 12367
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.732876109468357e-06,
+      "loss": 1.1303,
+      "step": 12368
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.732181139512442e-06,
+      "loss": 1.0495,
+      "step": 12369
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.731486174745457e-06,
+      "loss": 0.9714,
+      "step": 12370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.730791215180866e-06,
+      "loss": 0.915,
+      "step": 12371
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.730096260832135e-06,
+      "loss": 0.9202,
+      "step": 12372
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.729401311712726e-06,
+      "loss": 0.8998,
+      "step": 12373
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7287063678361054e-06,
+      "loss": 0.7081,
+      "step": 12374
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.728011429215738e-06,
+      "loss": 0.7962,
+      "step": 12375
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.727316495865088e-06,
+      "loss": 0.9459,
+      "step": 12376
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7266215677976185e-06,
+      "loss": 1.0596,
+      "step": 12377
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.725926645026794e-06,
+      "loss": 0.9905,
+      "step": 12378
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.725231727566079e-06,
+      "loss": 1.1801,
+      "step": 12379
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.724536815428937e-06,
+      "loss": 1.225,
+      "step": 12380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.723841908628832e-06,
+      "loss": 1.0175,
+      "step": 12381
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.723147007179226e-06,
+      "loss": 0.7907,
+      "step": 12382
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.722452111093586e-06,
+      "loss": 1.0834,
+      "step": 12383
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.721757220385372e-06,
+      "loss": 0.9224,
+      "step": 12384
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.721062335068046e-06,
+      "loss": 0.9116,
+      "step": 12385
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.720367455155077e-06,
+      "loss": 1.0042,
+      "step": 12386
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.719672580659925e-06,
+      "loss": 1.0062,
+      "step": 12387
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.718977711596054e-06,
+      "loss": 0.9982,
+      "step": 12388
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.718282847976925e-06,
+      "loss": 0.9301,
+      "step": 12389
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.717587989816002e-06,
+      "loss": 0.9957,
+      "step": 12390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.716893137126748e-06,
+      "loss": 1.0133,
+      "step": 12391
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7161982899226254e-06,
+      "loss": 1.1061,
+      "step": 12392
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.715503448217096e-06,
+      "loss": 0.9654,
+      "step": 12393
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.714808612023623e-06,
+      "loss": 1.0443,
+      "step": 12394
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7141137813556685e-06,
+      "loss": 0.783,
+      "step": 12395
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.713418956226695e-06,
+      "loss": 1.0952,
+      "step": 12396
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.712724136650162e-06,
+      "loss": 0.9488,
+      "step": 12397
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.712029322639536e-06,
+      "loss": 0.9612,
+      "step": 12398
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.711334514208276e-06,
+      "loss": 0.8678,
+      "step": 12399
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.710639711369845e-06,
+      "loss": 0.7706,
+      "step": 12400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.709944914137703e-06,
+      "loss": 0.9649,
+      "step": 12401
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.709250122525313e-06,
+      "loss": 0.8502,
+      "step": 12402
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.708555336546135e-06,
+      "loss": 0.8375,
+      "step": 12403
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.707860556213633e-06,
+      "loss": 0.9806,
+      "step": 12404
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.707165781541265e-06,
+      "loss": 1.029,
+      "step": 12405
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.706471012542494e-06,
+      "loss": 0.9469,
+      "step": 12406
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7057762492307805e-06,
+      "loss": 1.0593,
+      "step": 12407
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.705081491619586e-06,
+      "loss": 1.1251,
+      "step": 12408
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.704386739722368e-06,
+      "loss": 0.8752,
+      "step": 12409
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.703691993552592e-06,
+      "loss": 1.0025,
+      "step": 12410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.702997253123716e-06,
+      "loss": 0.9039,
+      "step": 12411
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.702302518449202e-06,
+      "loss": 0.733,
+      "step": 12412
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.701607789542508e-06,
+      "loss": 1.0195,
+      "step": 12413
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.700913066417096e-06,
+      "loss": 0.9222,
+      "step": 12414
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.700218349086425e-06,
+      "loss": 0.8703,
+      "step": 12415
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.699523637563956e-06,
+      "loss": 0.9491,
+      "step": 12416
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.698828931863148e-06,
+      "loss": 1.0281,
+      "step": 12417
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.698134231997462e-06,
+      "loss": 1.0742,
+      "step": 12418
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.697439537980356e-06,
+      "loss": 1.1683,
+      "step": 12419
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6967448498252886e-06,
+      "loss": 0.8935,
+      "step": 12420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.696050167545723e-06,
+      "loss": 0.8442,
+      "step": 12421
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.695355491155116e-06,
+      "loss": 1.091,
+      "step": 12422
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.694660820666927e-06,
+      "loss": 1.1519,
+      "step": 12423
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.693966156094617e-06,
+      "loss": 0.912,
+      "step": 12424
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.693271497451641e-06,
+      "loss": 0.8954,
+      "step": 12425
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.692576844751462e-06,
+      "loss": 0.9488,
+      "step": 12426
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.691882198007536e-06,
+      "loss": 0.9729,
+      "step": 12427
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.691187557233323e-06,
+      "loss": 0.9124,
+      "step": 12428
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6904929224422806e-06,
+      "loss": 0.8392,
+      "step": 12429
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.689798293647868e-06,
+      "loss": 0.8934,
+      "step": 12430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.689103670863542e-06,
+      "loss": 0.9655,
+      "step": 12431
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.688409054102762e-06,
+      "loss": 0.9082,
+      "step": 12432
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.687714443378987e-06,
+      "loss": 0.8142,
+      "step": 12433
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.687019838705673e-06,
+      "loss": 0.9485,
+      "step": 12434
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.686325240096279e-06,
+      "loss": 0.9285,
+      "step": 12435
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.685630647564262e-06,
+      "loss": 0.8349,
+      "step": 12436
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.684936061123081e-06,
+      "loss": 1.0542,
+      "step": 12437
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.684241480786192e-06,
+      "loss": 0.9554,
+      "step": 12438
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.683546906567052e-06,
+      "loss": 0.8239,
+      "step": 12439
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.682852338479119e-06,
+      "loss": 0.8885,
+      "step": 12440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.682157776535851e-06,
+      "loss": 0.7846,
+      "step": 12441
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.681463220750702e-06,
+      "loss": 0.8393,
+      "step": 12442
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.680768671137133e-06,
+      "loss": 0.8603,
+      "step": 12443
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.680074127708596e-06,
+      "loss": 1.0674,
+      "step": 12444
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.679379590478551e-06,
+      "loss": 1.0122,
+      "step": 12445
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.678685059460455e-06,
+      "loss": 0.9248,
+      "step": 12446
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.677990534667762e-06,
+      "loss": 0.9713,
+      "step": 12447
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.67729601611393e-06,
+      "loss": 0.9362,
+      "step": 12448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.676601503812415e-06,
+      "loss": 1.0443,
+      "step": 12449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.675906997776672e-06,
+      "loss": 1.0496,
+      "step": 12450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.675212498020157e-06,
+      "loss": 1.1186,
+      "step": 12451
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6745180045563266e-06,
+      "loss": 0.926,
+      "step": 12452
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.673823517398635e-06,
+      "loss": 0.9157,
+      "step": 12453
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.673129036560541e-06,
+      "loss": 1.0089,
+      "step": 12454
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.672434562055495e-06,
+      "loss": 0.8583,
+      "step": 12455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.671740093896956e-06,
+      "loss": 0.9708,
+      "step": 12456
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.67104563209838e-06,
+      "loss": 0.8662,
+      "step": 12457
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6703511766732185e-06,
+      "loss": 0.9989,
+      "step": 12458
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.669656727634929e-06,
+      "loss": 0.763,
+      "step": 12459
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.668962284996964e-06,
+      "loss": 0.8773,
+      "step": 12460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6682678487727804e-06,
+      "loss": 0.9875,
+      "step": 12461
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.667573418975833e-06,
+      "loss": 0.9718,
+      "step": 12462
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.666878995619574e-06,
+      "loss": 0.8779,
+      "step": 12463
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.66618457871746e-06,
+      "loss": 0.8715,
+      "step": 12464
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.665490168282943e-06,
+      "loss": 1.0281,
+      "step": 12465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.664795764329479e-06,
+      "loss": 1.0586,
+      "step": 12466
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6641013668705195e-06,
+      "loss": 0.902,
+      "step": 12467
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.663406975919521e-06,
+      "loss": 1.1085,
+      "step": 12468
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6627125914899364e-06,
+      "loss": 0.9566,
+      "step": 12469
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.662018213595218e-06,
+      "loss": 1.0858,
+      "step": 12470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.661323842248821e-06,
+      "loss": 1.0723,
+      "step": 12471
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.660629477464197e-06,
+      "loss": 0.7566,
+      "step": 12472
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6599351192547994e-06,
+      "loss": 0.9395,
+      "step": 12473
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6592407676340825e-06,
+      "loss": 0.885,
+      "step": 12474
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6585464226154995e-06,
+      "loss": 1.0447,
+      "step": 12475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6578520842125e-06,
+      "loss": 0.9917,
+      "step": 12476
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.65715775243854e-06,
+      "loss": 0.8117,
+      "step": 12477
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.656463427307069e-06,
+      "loss": 0.9124,
+      "step": 12478
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.655769108831542e-06,
+      "loss": 1.0307,
+      "step": 12479
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6550747970254115e-06,
+      "loss": 1.0313,
+      "step": 12480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.654380491902128e-06,
+      "loss": 1.0868,
+      "step": 12481
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.653686193475144e-06,
+      "loss": 0.9422,
+      "step": 12482
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.652991901757912e-06,
+      "loss": 1.0057,
+      "step": 12483
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.652297616763882e-06,
+      "loss": 0.8979,
+      "step": 12484
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6516033385065084e-06,
+      "loss": 0.8287,
+      "step": 12485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.65090906699924e-06,
+      "loss": 1.0736,
+      "step": 12486
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.65021480225553e-06,
+      "loss": 0.8715,
+      "step": 12487
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.649520544288828e-06,
+      "loss": 0.9297,
+      "step": 12488
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6488262931125875e-06,
+      "loss": 1.1203,
+      "step": 12489
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.648132048740256e-06,
+      "loss": 0.9863,
+      "step": 12490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6474378111852866e-06,
+      "loss": 0.9195,
+      "step": 12491
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.64674358046113e-06,
+      "loss": 0.8826,
+      "step": 12492
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.646049356581236e-06,
+      "loss": 1.0025,
+      "step": 12493
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.645355139559057e-06,
+      "loss": 0.9912,
+      "step": 12494
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.644660929408039e-06,
+      "loss": 0.9943,
+      "step": 12495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.643966726141637e-06,
+      "loss": 1.0735,
+      "step": 12496
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.643272529773299e-06,
+      "loss": 0.8901,
+      "step": 12497
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.642578340316474e-06,
+      "loss": 0.9256,
+      "step": 12498
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.641884157784613e-06,
+      "loss": 0.9692,
+      "step": 12499
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.641189982191165e-06,
+      "loss": 0.9251,
+      "step": 12500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.64049581354958e-06,
+      "loss": 0.8454,
+      "step": 12501
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.639801651873305e-06,
+      "loss": 0.8253,
+      "step": 12502
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.639107497175793e-06,
+      "loss": 0.8199,
+      "step": 12503
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.638413349470491e-06,
+      "loss": 0.9616,
+      "step": 12504
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.637719208770849e-06,
+      "loss": 1.0088,
+      "step": 12505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.637025075090315e-06,
+      "loss": 1.0471,
+      "step": 12506
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6363309484423376e-06,
+      "loss": 0.9662,
+      "step": 12507
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6356368288403654e-06,
+      "loss": 1.0848,
+      "step": 12508
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.634942716297848e-06,
+      "loss": 1.06,
+      "step": 12509
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.634248610828232e-06,
+      "loss": 0.9264,
+      "step": 12510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.633554512444966e-06,
+      "loss": 0.9994,
+      "step": 12511
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.632860421161498e-06,
+      "loss": 1.1608,
+      "step": 12512
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.632166336991275e-06,
+      "loss": 0.8826,
+      "step": 12513
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.631472259947748e-06,
+      "loss": 1.0537,
+      "step": 12514
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.630778190044361e-06,
+      "loss": 1.0538,
+      "step": 12515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.630084127294564e-06,
+      "loss": 0.9703,
+      "step": 12516
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.629390071711802e-06,
+      "loss": 1.1318,
+      "step": 12517
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6286960233095235e-06,
+      "loss": 0.7791,
+      "step": 12518
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.628001982101177e-06,
+      "loss": 1.08,
+      "step": 12519
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.627307948100205e-06,
+      "loss": 0.8935,
+      "step": 12520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.626613921320059e-06,
+      "loss": 0.9037,
+      "step": 12521
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.625919901774183e-06,
+      "loss": 0.9084,
+      "step": 12522
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.625225889476024e-06,
+      "loss": 0.9494,
+      "step": 12523
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.624531884439028e-06,
+      "loss": 0.9645,
+      "step": 12524
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6238378866766405e-06,
+      "loss": 1.0071,
+      "step": 12525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.62314389620231e-06,
+      "loss": 0.7943,
+      "step": 12526
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.622449913029482e-06,
+      "loss": 0.9046,
+      "step": 12527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6217559371716005e-06,
+      "loss": 0.888,
+      "step": 12528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.621061968642112e-06,
+      "loss": 0.9095,
+      "step": 12529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.620368007454461e-06,
+      "loss": 0.9457,
+      "step": 12530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.619674053622095e-06,
+      "loss": 0.9604,
+      "step": 12531
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.618980107158457e-06,
+      "loss": 0.9974,
+      "step": 12532
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.618286168076993e-06,
+      "loss": 0.9634,
+      "step": 12533
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.617592236391149e-06,
+      "loss": 0.7513,
+      "step": 12534
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.616898312114368e-06,
+      "loss": 0.9977,
+      "step": 12535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.616204395260095e-06,
+      "loss": 1.0176,
+      "step": 12536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.615510485841773e-06,
+      "loss": 0.8876,
+      "step": 12537
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.61481658387285e-06,
+      "loss": 0.9815,
+      "step": 12538
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.614122689366769e-06,
+      "loss": 1.0579,
+      "step": 12539
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.613428802336972e-06,
+      "loss": 0.9977,
+      "step": 12540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.612734922796906e-06,
+      "loss": 0.9258,
+      "step": 12541
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.612041050760012e-06,
+      "loss": 0.7761,
+      "step": 12542
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.611347186239734e-06,
+      "loss": 0.8734,
+      "step": 12543
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6106533292495174e-06,
+      "loss": 0.9153,
+      "step": 12544
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.609959479802803e-06,
+      "loss": 0.9583,
+      "step": 12545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.609265637913036e-06,
+      "loss": 0.7635,
+      "step": 12546
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.608571803593659e-06,
+      "loss": 0.9133,
+      "step": 12547
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.607877976858113e-06,
+      "loss": 0.9645,
+      "step": 12548
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.607184157719843e-06,
+      "loss": 0.8965,
+      "step": 12549
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.606490346192291e-06,
+      "loss": 0.9155,
+      "step": 12550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6057965422889e-06,
+      "loss": 0.8454,
+      "step": 12551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.605102746023111e-06,
+      "loss": 0.8576,
+      "step": 12552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.604408957408366e-06,
+      "loss": 0.9017,
+      "step": 12553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.603715176458109e-06,
+      "loss": 1.109,
+      "step": 12554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.60302140318578e-06,
+      "loss": 0.9866,
+      "step": 12555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.602327637604821e-06,
+      "loss": 0.9103,
+      "step": 12556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.601633879728674e-06,
+      "loss": 0.9606,
+      "step": 12557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.60094012957078e-06,
+      "loss": 0.8982,
+      "step": 12558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.60024638714458e-06,
+      "loss": 0.8447,
+      "step": 12559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.599552652463516e-06,
+      "loss": 1.0058,
+      "step": 12560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.598858925541028e-06,
+      "loss": 0.873,
+      "step": 12561
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.598165206390558e-06,
+      "loss": 0.8704,
+      "step": 12562
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.597471495025546e-06,
+      "loss": 0.8893,
+      "step": 12563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.596777791459433e-06,
+      "loss": 1.0085,
+      "step": 12564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.596084095705658e-06,
+      "loss": 1.0422,
+      "step": 12565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.595390407777663e-06,
+      "loss": 0.7745,
+      "step": 12566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.594696727688887e-06,
+      "loss": 0.9983,
+      "step": 12567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.594003055452769e-06,
+      "loss": 0.9375,
+      "step": 12568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5933093910827505e-06,
+      "loss": 1.0472,
+      "step": 12569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5926157345922696e-06,
+      "loss": 1.0371,
+      "step": 12570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.591922085994764e-06,
+      "loss": 0.8731,
+      "step": 12571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.591228445303679e-06,
+      "loss": 0.9157,
+      "step": 12572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.590534812532449e-06,
+      "loss": 0.9844,
+      "step": 12573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.589841187694515e-06,
+      "loss": 0.9946,
+      "step": 12574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.589147570803314e-06,
+      "loss": 0.9425,
+      "step": 12575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.588453961872286e-06,
+      "loss": 0.7973,
+      "step": 12576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.58776036091487e-06,
+      "loss": 0.9247,
+      "step": 12577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.587066767944503e-06,
+      "loss": 0.9206,
+      "step": 12578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5863731829746225e-06,
+      "loss": 1.0529,
+      "step": 12579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5856796060186685e-06,
+      "loss": 0.8407,
+      "step": 12580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.584986037090079e-06,
+      "loss": 1.0059,
+      "step": 12581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.58429247620229e-06,
+      "loss": 0.8148,
+      "step": 12582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.583598923368739e-06,
+      "loss": 1.1062,
+      "step": 12583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.582905378602865e-06,
+      "loss": 0.8638,
+      "step": 12584
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.582211841918106e-06,
+      "loss": 0.9789,
+      "step": 12585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5815183133278965e-06,
+      "loss": 0.964,
+      "step": 12586
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5808247928456756e-06,
+      "loss": 1.0068,
+      "step": 12587
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.580131280484879e-06,
+      "loss": 1.0534,
+      "step": 12588
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5794377762589434e-06,
+      "loss": 1.0268,
+      "step": 12589
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.578744280181306e-06,
+      "loss": 0.9519,
+      "step": 12590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.578050792265402e-06,
+      "loss": 0.9687,
+      "step": 12591
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.577357312524668e-06,
+      "loss": 0.9218,
+      "step": 12592
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.576663840972541e-06,
+      "loss": 0.8746,
+      "step": 12593
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.575970377622456e-06,
+      "loss": 0.8831,
+      "step": 12594
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.575276922487846e-06,
+      "loss": 0.992,
+      "step": 12595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.574583475582152e-06,
+      "loss": 1.0911,
+      "step": 12596
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5738900369188065e-06,
+      "loss": 0.9786,
+      "step": 12597
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.573196606511245e-06,
+      "loss": 0.9283,
+      "step": 12598
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5725031843729016e-06,
+      "loss": 0.7714,
+      "step": 12599
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.571809770517212e-06,
+      "loss": 0.9695,
+      "step": 12600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.571116364957612e-06,
+      "loss": 0.9256,
+      "step": 12601
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.570422967707535e-06,
+      "loss": 0.9019,
+      "step": 12602
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5697295787804155e-06,
+      "loss": 0.8966,
+      "step": 12603
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.569036198189687e-06,
+      "loss": 0.9073,
+      "step": 12604
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.568342825948786e-06,
+      "loss": 0.9235,
+      "step": 12605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.567649462071142e-06,
+      "loss": 0.8051,
+      "step": 12606
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.566956106570193e-06,
+      "loss": 1.0109,
+      "step": 12607
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.566262759459371e-06,
+      "loss": 0.9152,
+      "step": 12608
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.565569420752111e-06,
+      "loss": 0.8747,
+      "step": 12609
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.564876090461844e-06,
+      "loss": 1.0867,
+      "step": 12610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.564182768602004e-06,
+      "loss": 0.7896,
+      "step": 12611
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5634894551860245e-06,
+      "loss": 0.9568,
+      "step": 12612
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.562796150227337e-06,
+      "loss": 0.8881,
+      "step": 12613
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.562102853739376e-06,
+      "loss": 0.7884,
+      "step": 12614
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.561409565735573e-06,
+      "loss": 0.7243,
+      "step": 12615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.560716286229359e-06,
+      "loss": 0.9979,
+      "step": 12616
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.560023015234167e-06,
+      "loss": 0.8997,
+      "step": 12617
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.559329752763429e-06,
+      "loss": 0.9511,
+      "step": 12618
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.558636498830578e-06,
+      "loss": 0.8856,
+      "step": 12619
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.557943253449045e-06,
+      "loss": 1.0565,
+      "step": 12620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.557250016632262e-06,
+      "loss": 0.9061,
+      "step": 12621
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.556556788393659e-06,
+      "loss": 0.9135,
+      "step": 12622
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.555863568746667e-06,
+      "loss": 0.9578,
+      "step": 12623
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.555170357704717e-06,
+      "loss": 1.2862,
+      "step": 12624
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.554477155281242e-06,
+      "loss": 0.8288,
+      "step": 12625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.55378396148967e-06,
+      "loss": 1.0043,
+      "step": 12626
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.553090776343433e-06,
+      "loss": 0.9375,
+      "step": 12627
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.552397599855961e-06,
+      "loss": 1.2196,
+      "step": 12628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.551704432040683e-06,
+      "loss": 0.886,
+      "step": 12629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.551011272911028e-06,
+      "loss": 0.8459,
+      "step": 12630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.55031812248043e-06,
+      "loss": 1.0423,
+      "step": 12631
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.549624980762317e-06,
+      "loss": 1.0454,
+      "step": 12632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.548931847770118e-06,
+      "loss": 0.9277,
+      "step": 12633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.548238723517262e-06,
+      "loss": 0.8667,
+      "step": 12634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.547545608017177e-06,
+      "loss": 0.926,
+      "step": 12635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.546852501283294e-06,
+      "loss": 1.1242,
+      "step": 12636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.546159403329041e-06,
+      "loss": 0.83,
+      "step": 12637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5454663141678475e-06,
+      "loss": 0.7576,
+      "step": 12638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.54477323381314e-06,
+      "loss": 0.9122,
+      "step": 12639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.544080162278348e-06,
+      "loss": 0.7583,
+      "step": 12640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.543387099576898e-06,
+      "loss": 0.9499,
+      "step": 12641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.542694045722221e-06,
+      "loss": 1.0411,
+      "step": 12642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.542001000727744e-06,
+      "loss": 0.9861,
+      "step": 12643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.541307964606893e-06,
+      "loss": 0.96,
+      "step": 12644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.540614937373095e-06,
+      "loss": 0.9824,
+      "step": 12645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.539921919039779e-06,
+      "loss": 0.9795,
+      "step": 12646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.539228909620372e-06,
+      "loss": 0.8493,
+      "step": 12647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5385359091283005e-06,
+      "loss": 1.0188,
+      "step": 12648
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.537842917576991e-06,
+      "loss": 0.8392,
+      "step": 12649
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.537149934979869e-06,
+      "loss": 0.8994,
+      "step": 12650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.536456961350363e-06,
+      "loss": 1.044,
+      "step": 12651
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.535763996701898e-06,
+      "loss": 0.8984,
+      "step": 12652
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.535071041047898e-06,
+      "loss": 0.915,
+      "step": 12653
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.534378094401793e-06,
+      "loss": 0.9156,
+      "step": 12654
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.533685156777008e-06,
+      "loss": 0.8245,
+      "step": 12655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.532992228186966e-06,
+      "loss": 1.0665,
+      "step": 12656
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.532299308645095e-06,
+      "loss": 1.0879,
+      "step": 12657
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.531606398164818e-06,
+      "loss": 1.0327,
+      "step": 12658
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.530913496759561e-06,
+      "loss": 0.8215,
+      "step": 12659
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5302206044427476e-06,
+      "loss": 1.1601,
+      "step": 12660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.529527721227804e-06,
+      "loss": 0.9853,
+      "step": 12661
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.528834847128155e-06,
+      "loss": 0.7783,
+      "step": 12662
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.528141982157223e-06,
+      "loss": 0.9412,
+      "step": 12663
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.527449126328431e-06,
+      "loss": 0.8983,
+      "step": 12664
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.526756279655207e-06,
+      "loss": 0.9594,
+      "step": 12665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.526063442150973e-06,
+      "loss": 0.8635,
+      "step": 12666
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.525370613829153e-06,
+      "loss": 1.0363,
+      "step": 12667
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.524677794703168e-06,
+      "loss": 0.8994,
+      "step": 12668
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.523984984786444e-06,
+      "loss": 0.8949,
+      "step": 12669
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.523292184092403e-06,
+      "loss": 0.9871,
+      "step": 12670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.522599392634468e-06,
+      "loss": 0.8893,
+      "step": 12671
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.521906610426061e-06,
+      "loss": 0.8296,
+      "step": 12672
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.521213837480605e-06,
+      "loss": 0.9074,
+      "step": 12673
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.520521073811523e-06,
+      "loss": 1.1418,
+      "step": 12674
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.519828319432235e-06,
+      "loss": 0.8842,
+      "step": 12675
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.519135574356164e-06,
+      "loss": 0.9464,
+      "step": 12676
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.518442838596733e-06,
+      "loss": 0.8656,
+      "step": 12677
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.517750112167364e-06,
+      "loss": 0.8502,
+      "step": 12678
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.517057395081476e-06,
+      "loss": 0.9192,
+      "step": 12679
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.516364687352492e-06,
+      "loss": 1.0651,
+      "step": 12680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5156719889938325e-06,
+      "loss": 0.907,
+      "step": 12681
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.514979300018918e-06,
+      "loss": 0.8479,
+      "step": 12682
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.51428662044117e-06,
+      "loss": 0.8858,
+      "step": 12683
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.513593950274008e-06,
+      "loss": 0.7741,
+      "step": 12684
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.512901289530853e-06,
+      "loss": 0.8623,
+      "step": 12685
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.512208638225124e-06,
+      "loss": 0.8758,
+      "step": 12686
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.511515996370244e-06,
+      "loss": 0.8551,
+      "step": 12687
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.510823363979627e-06,
+      "loss": 0.8484,
+      "step": 12688
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5101307410666994e-06,
+      "loss": 1.02,
+      "step": 12689
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.509438127644877e-06,
+      "loss": 0.743,
+      "step": 12690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50874552372758e-06,
+      "loss": 1.0323,
+      "step": 12691
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.508052929328227e-06,
+      "loss": 0.9017,
+      "step": 12692
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5073603444602374e-06,
+      "loss": 0.8183,
+      "step": 12693
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.506667769137029e-06,
+      "loss": 0.9143,
+      "step": 12694
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.505975203372021e-06,
+      "loss": 0.901,
+      "step": 12695
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50528264717863e-06,
+      "loss": 0.8949,
+      "step": 12696
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.504590100570276e-06,
+      "loss": 0.9097,
+      "step": 12697
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5038975635603774e-06,
+      "loss": 0.9838,
+      "step": 12698
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.503205036162348e-06,
+      "loss": 0.9558,
+      "step": 12699
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.502512518389611e-06,
+      "loss": 0.9384,
+      "step": 12700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5018200102555805e-06,
+      "loss": 1.0286,
+      "step": 12701
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.501127511773674e-06,
+      "loss": 1.016,
+      "step": 12702
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50043502295731e-06,
+      "loss": 0.8828,
+      "step": 12703
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4997425438199035e-06,
+      "loss": 0.8483,
+      "step": 12704
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.499050074374872e-06,
+      "loss": 0.9727,
+      "step": 12705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.498357614635631e-06,
+      "loss": 0.8463,
+      "step": 12706
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4976651646155986e-06,
+      "loss": 0.8813,
+      "step": 12707
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.496972724328189e-06,
+      "loss": 0.9439,
+      "step": 12708
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.496280293786818e-06,
+      "loss": 0.8424,
+      "step": 12709
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.495587873004903e-06,
+      "loss": 0.8599,
+      "step": 12710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.494895461995857e-06,
+      "loss": 0.9927,
+      "step": 12711
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.494203060773099e-06,
+      "loss": 0.8896,
+      "step": 12712
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4935106693500405e-06,
+      "loss": 1.0332,
+      "step": 12713
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.492818287740098e-06,
+      "loss": 1.0285,
+      "step": 12714
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.492125915956687e-06,
+      "loss": 0.9445,
+      "step": 12715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.491433554013221e-06,
+      "loss": 0.9144,
+      "step": 12716
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.490741201923114e-06,
+      "loss": 0.8807,
+      "step": 12717
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.490048859699782e-06,
+      "loss": 0.8399,
+      "step": 12718
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.489356527356636e-06,
+      "loss": 0.9264,
+      "step": 12719
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.488664204907092e-06,
+      "loss": 0.7327,
+      "step": 12720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.487971892364563e-06,
+      "loss": 1.0847,
+      "step": 12721
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4872795897424606e-06,
+      "loss": 0.931,
+      "step": 12722
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.486587297054202e-06,
+      "loss": 0.94,
+      "step": 12723
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.485895014313198e-06,
+      "loss": 0.9419,
+      "step": 12724
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.485202741532861e-06,
+      "loss": 0.9161,
+      "step": 12725
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4845104787266045e-06,
+      "loss": 0.9299,
+      "step": 12726
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.483818225907841e-06,
+      "loss": 0.9935,
+      "step": 12727
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.483125983089981e-06,
+      "loss": 0.8135,
+      "step": 12728
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.482433750286439e-06,
+      "loss": 0.9562,
+      "step": 12729
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4817415275106255e-06,
+      "loss": 1.0054,
+      "step": 12730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.481049314775952e-06,
+      "loss": 0.8748,
+      "step": 12731
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.480357112095831e-06,
+      "loss": 0.9379,
+      "step": 12732
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.479664919483672e-06,
+      "loss": 1.0147,
+      "step": 12733
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.478972736952887e-06,
+      "loss": 0.9425,
+      "step": 12734
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4782805645168885e-06,
+      "loss": 1.0451,
+      "step": 12735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.477588402189084e-06,
+      "loss": 0.9117,
+      "step": 12736
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.476896249982887e-06,
+      "loss": 1.0626,
+      "step": 12737
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.476204107911707e-06,
+      "loss": 0.8261,
+      "step": 12738
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.475511975988952e-06,
+      "loss": 0.8528,
+      "step": 12739
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.474819854228034e-06,
+      "loss": 0.987,
+      "step": 12740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.474127742642363e-06,
+      "loss": 0.9604,
+      "step": 12741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.473435641245347e-06,
+      "loss": 0.8611,
+      "step": 12742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.472743550050396e-06,
+      "loss": 1.0691,
+      "step": 12743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4720514690709195e-06,
+      "loss": 0.9508,
+      "step": 12744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.471359398320325e-06,
+      "loss": 0.8668,
+      "step": 12745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4706673378120215e-06,
+      "loss": 0.9707,
+      "step": 12746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.46997528755942e-06,
+      "loss": 0.8442,
+      "step": 12747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.469283247575926e-06,
+      "loss": 0.9135,
+      "step": 12748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.46859121787495e-06,
+      "loss": 0.9991,
+      "step": 12749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.467899198469898e-06,
+      "loss": 0.9515,
+      "step": 12750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.467207189374178e-06,
+      "loss": 0.956,
+      "step": 12751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.466515190601199e-06,
+      "loss": 0.7952,
+      "step": 12752
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4658232021643665e-06,
+      "loss": 1.0042,
+      "step": 12753
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.465131224077088e-06,
+      "loss": 0.8749,
+      "step": 12754
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4644392563527715e-06,
+      "loss": 0.7342,
+      "step": 12755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.463747299004823e-06,
+      "loss": 1.0352,
+      "step": 12756
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.463055352046646e-06,
+      "loss": 0.9538,
+      "step": 12757
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.462363415491652e-06,
+      "loss": 1.1159,
+      "step": 12758
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.461671489353245e-06,
+      "loss": 1.1065,
+      "step": 12759
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4609795736448306e-06,
+      "loss": 1.0881,
+      "step": 12760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.460287668379815e-06,
+      "loss": 0.9621,
+      "step": 12761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4595957735716035e-06,
+      "loss": 0.741,
+      "step": 12762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4589038892336005e-06,
+      "loss": 1.0224,
+      "step": 12763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.458212015379212e-06,
+      "loss": 0.8272,
+      "step": 12764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.457520152021844e-06,
+      "loss": 1.0381,
+      "step": 12765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.456828299174899e-06,
+      "loss": 0.9953,
+      "step": 12766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.456136456851783e-06,
+      "loss": 0.944,
+      "step": 12767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.455444625065899e-06,
+      "loss": 1.2312,
+      "step": 12768
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.454752803830652e-06,
+      "loss": 0.8641,
+      "step": 12769
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.454060993159446e-06,
+      "loss": 1.1286,
+      "step": 12770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.453369193065685e-06,
+      "loss": 1.1066,
+      "step": 12771
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.45267740356277e-06,
+      "loss": 1.0919,
+      "step": 12772
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.451985624664109e-06,
+      "loss": 1.0042,
+      "step": 12773
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4512938563831e-06,
+      "loss": 0.9718,
+      "step": 12774
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.450602098733149e-06,
+      "loss": 0.9091,
+      "step": 12775
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4499103517276575e-06,
+      "loss": 0.762,
+      "step": 12776
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.449218615380029e-06,
+      "loss": 0.915,
+      "step": 12777
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.448526889703664e-06,
+      "loss": 0.9311,
+      "step": 12778
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.447835174711965e-06,
+      "loss": 0.8982,
+      "step": 12779
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.447143470418336e-06,
+      "loss": 1.0298,
+      "step": 12780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.446451776836172e-06,
+      "loss": 0.8489,
+      "step": 12781
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.445760093978884e-06,
+      "loss": 0.8415,
+      "step": 12782
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.445068421859867e-06,
+      "loss": 1.0685,
+      "step": 12783
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.444376760492524e-06,
+      "loss": 0.955,
+      "step": 12784
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.443685109890254e-06,
+      "loss": 0.9162,
+      "step": 12785
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.442993470066461e-06,
+      "loss": 1.0991,
+      "step": 12786
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.442301841034541e-06,
+      "loss": 1.0746,
+      "step": 12787
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4416102228078975e-06,
+      "loss": 0.893,
+      "step": 12788
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4409186153999296e-06,
+      "loss": 1.0358,
+      "step": 12789
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.440227018824036e-06,
+      "loss": 0.8434,
+      "step": 12790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.439535433093617e-06,
+      "loss": 0.9477,
+      "step": 12791
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.43884385822207e-06,
+      "loss": 1.0539,
+      "step": 12792
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4381522942227985e-06,
+      "loss": 0.9675,
+      "step": 12793
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.437460741109198e-06,
+      "loss": 1.0098,
+      "step": 12794
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.436769198894668e-06,
+      "loss": 0.8888,
+      "step": 12795
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.436077667592607e-06,
+      "loss": 0.8262,
+      "step": 12796
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.435386147216414e-06,
+      "loss": 0.958,
+      "step": 12797
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.434694637779486e-06,
+      "loss": 0.808,
+      "step": 12798
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.434003139295221e-06,
+      "loss": 0.8127,
+      "step": 12799
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.433311651777017e-06,
+      "loss": 0.9485,
+      "step": 12800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.432620175238272e-06,
+      "loss": 0.9483,
+      "step": 12801
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4319287096923805e-06,
+      "loss": 0.9666,
+      "step": 12802
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.431237255152743e-06,
+      "loss": 1.0119,
+      "step": 12803
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.430545811632753e-06,
+      "loss": 1.0407,
+      "step": 12804
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4298543791458095e-06,
+      "loss": 0.8051,
+      "step": 12805
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.429162957705308e-06,
+      "loss": 0.9413,
+      "step": 12806
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.428471547324645e-06,
+      "loss": 0.9435,
+      "step": 12807
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.427780148017216e-06,
+      "loss": 0.9837,
+      "step": 12808
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4270887597964176e-06,
+      "loss": 0.8887,
+      "step": 12809
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.426397382675643e-06,
+      "loss": 0.9637,
+      "step": 12810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.425706016668291e-06,
+      "loss": 0.9164,
+      "step": 12811
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.425014661787753e-06,
+      "loss": 1.1714,
+      "step": 12812
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4243233180474265e-06,
+      "loss": 0.9036,
+      "step": 12813
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.423631985460704e-06,
+      "loss": 0.7257,
+      "step": 12814
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.42294066404098e-06,
+      "loss": 1.0303,
+      "step": 12815
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.422249353801651e-06,
+      "loss": 0.875,
+      "step": 12816
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.421558054756111e-06,
+      "loss": 1.0528,
+      "step": 12817
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.420866766917751e-06,
+      "loss": 0.9928,
+      "step": 12818
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.420175490299967e-06,
+      "loss": 0.914,
+      "step": 12819
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.419484224916151e-06,
+      "loss": 1.027,
+      "step": 12820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.418792970779697e-06,
+      "loss": 0.9448,
+      "step": 12821
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.418101727903996e-06,
+      "loss": 1.0058,
+      "step": 12822
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4174104963024435e-06,
+      "loss": 0.9085,
+      "step": 12823
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.41671927598843e-06,
+      "loss": 0.9297,
+      "step": 12824
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.416028066975347e-06,
+      "loss": 0.9119,
+      "step": 12825
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.41533686927659e-06,
+      "loss": 0.9545,
+      "step": 12826
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4146456829055455e-06,
+      "loss": 0.8617,
+      "step": 12827
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.41395450787561e-06,
+      "loss": 0.9091,
+      "step": 12828
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.413263344200173e-06,
+      "loss": 0.912,
+      "step": 12829
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.412572191892626e-06,
+      "loss": 0.7862,
+      "step": 12830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.411881050966359e-06,
+      "loss": 0.9928,
+      "step": 12831
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.411189921434763e-06,
+      "loss": 1.0446,
+      "step": 12832
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.410498803311229e-06,
+      "loss": 1.0023,
+      "step": 12833
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.409807696609148e-06,
+      "loss": 0.8582,
+      "step": 12834
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.409116601341908e-06,
+      "loss": 0.8602,
+      "step": 12835
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.408425517522899e-06,
+      "loss": 0.8048,
+      "step": 12836
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.407734445165512e-06,
+      "loss": 0.9454,
+      "step": 12837
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.407043384283136e-06,
+      "loss": 1.034,
+      "step": 12838
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4063523348891575e-06,
+      "loss": 1.0634,
+      "step": 12839
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.405661296996971e-06,
+      "loss": 0.9547,
+      "step": 12840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.404970270619961e-06,
+      "loss": 0.9469,
+      "step": 12841
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.404279255771517e-06,
+      "loss": 0.8435,
+      "step": 12842
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.403588252465028e-06,
+      "loss": 1.1077,
+      "step": 12843
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.402897260713881e-06,
+      "loss": 0.9652,
+      "step": 12844
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.402206280531464e-06,
+      "loss": 0.9326,
+      "step": 12845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.401515311931164e-06,
+      "loss": 0.8715,
+      "step": 12846
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.40082435492637e-06,
+      "loss": 1.1146,
+      "step": 12847
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.400133409530468e-06,
+      "loss": 0.8919,
+      "step": 12848
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.399442475756846e-06,
+      "loss": 0.9676,
+      "step": 12849
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.398751553618886e-06,
+      "loss": 1.0804,
+      "step": 12850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.398060643129981e-06,
+      "loss": 0.9831,
+      "step": 12851
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.397369744303515e-06,
+      "loss": 1.0239,
+      "step": 12852
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.396678857152874e-06,
+      "loss": 1.0782,
+      "step": 12853
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.395987981691442e-06,
+      "loss": 1.0186,
+      "step": 12854
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.395297117932607e-06,
+      "loss": 0.915,
+      "step": 12855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.394606265889752e-06,
+      "loss": 1.0525,
+      "step": 12856
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3939154255762635e-06,
+      "loss": 0.8107,
+      "step": 12857
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.393224597005527e-06,
+      "loss": 0.8989,
+      "step": 12858
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3925337801909255e-06,
+      "loss": 0.9598,
+      "step": 12859
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3918429751458444e-06,
+      "loss": 0.9114,
+      "step": 12860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.391152181883668e-06,
+      "loss": 1.1113,
+      "step": 12861
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.390461400417779e-06,
+      "loss": 0.9421,
+      "step": 12862
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3897706307615625e-06,
+      "loss": 0.882,
+      "step": 12863
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.389079872928401e-06,
+      "loss": 0.8484,
+      "step": 12864
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.38838912693168e-06,
+      "loss": 0.8991,
+      "step": 12865
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.387698392784781e-06,
+      "loss": 1.0795,
+      "step": 12866
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.387007670501086e-06,
+      "loss": 0.8481,
+      "step": 12867
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.386316960093978e-06,
+      "loss": 0.8709,
+      "step": 12868
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.38562626157684e-06,
+      "loss": 1.0841,
+      "step": 12869
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.384935574963053e-06,
+      "loss": 1.0558,
+      "step": 12870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3842449002660015e-06,
+      "loss": 1.0151,
+      "step": 12871
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.383554237499064e-06,
+      "loss": 1.048,
+      "step": 12872
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.382863586675624e-06,
+      "loss": 0.8255,
+      "step": 12873
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.38217294780906e-06,
+      "loss": 0.8747,
+      "step": 12874
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.381482320912756e-06,
+      "loss": 1.1438,
+      "step": 12875
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.380791706000093e-06,
+      "loss": 1.0729,
+      "step": 12876
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.380101103084449e-06,
+      "loss": 0.8644,
+      "step": 12877
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.379410512179207e-06,
+      "loss": 0.9557,
+      "step": 12878
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.378719933297745e-06,
+      "loss": 1.0583,
+      "step": 12879
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.378029366453443e-06,
+      "loss": 0.9173,
+      "step": 12880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.377338811659681e-06,
+      "loss": 0.9026,
+      "step": 12881
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.376648268929839e-06,
+      "loss": 0.9907,
+      "step": 12882
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.375957738277294e-06,
+      "loss": 0.996,
+      "step": 12883
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.375267219715427e-06,
+      "loss": 0.9187,
+      "step": 12884
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.374576713257614e-06,
+      "loss": 1.2253,
+      "step": 12885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.373886218917238e-06,
+      "loss": 1.0608,
+      "step": 12886
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.373195736707673e-06,
+      "loss": 0.9163,
+      "step": 12887
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.372505266642299e-06,
+      "loss": 1.0195,
+      "step": 12888
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.371814808734493e-06,
+      "loss": 0.9527,
+      "step": 12889
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.371124362997633e-06,
+      "loss": 0.8758,
+      "step": 12890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.370433929445096e-06,
+      "loss": 0.9041,
+      "step": 12891
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.369743508090259e-06,
+      "loss": 1.049,
+      "step": 12892
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.369053098946498e-06,
+      "loss": 1.1819,
+      "step": 12893
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.368362702027189e-06,
+      "loss": 0.8755,
+      "step": 12894
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3676723173457105e-06,
+      "loss": 0.9575,
+      "step": 12895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.366981944915438e-06,
+      "loss": 0.9429,
+      "step": 12896
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.366291584749744e-06,
+      "loss": 1.0158,
+      "step": 12897
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.365601236862009e-06,
+      "loss": 0.8882,
+      "step": 12898
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.364910901265607e-06,
+      "loss": 0.8672,
+      "step": 12899
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.364220577973911e-06,
+      "loss": 1.0281,
+      "step": 12900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.363530267000297e-06,
+      "loss": 1.0495,
+      "step": 12901
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.36283996835814e-06,
+      "loss": 0.9307,
+      "step": 12902
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.362149682060814e-06,
+      "loss": 0.9527,
+      "step": 12903
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.361459408121694e-06,
+      "loss": 0.9603,
+      "step": 12904
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.360769146554152e-06,
+      "loss": 0.9613,
+      "step": 12905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.360078897371564e-06,
+      "loss": 1.0467,
+      "step": 12906
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3593886605873005e-06,
+      "loss": 0.9226,
+      "step": 12907
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.358698436214736e-06,
+      "loss": 1.0308,
+      "step": 12908
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.358008224267245e-06,
+      "loss": 1.0384,
+      "step": 12909
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3573180247581994e-06,
+      "loss": 0.8508,
+      "step": 12910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.356627837700971e-06,
+      "loss": 0.9267,
+      "step": 12911
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.355937663108933e-06,
+      "loss": 0.903,
+      "step": 12912
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.355247500995455e-06,
+      "loss": 0.907,
+      "step": 12913
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.354557351373912e-06,
+      "loss": 0.9315,
+      "step": 12914
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.353867214257673e-06,
+      "loss": 0.9826,
+      "step": 12915
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.353177089660111e-06,
+      "loss": 0.897,
+      "step": 12916
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.352486977594596e-06,
+      "loss": 1.1126,
+      "step": 12917
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3517968780744986e-06,
+      "loss": 0.8671,
+      "step": 12918
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.351106791113189e-06,
+      "loss": 0.8977,
+      "step": 12919
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.350416716724038e-06,
+      "loss": 1.0691,
+      "step": 12920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.349726654920417e-06,
+      "loss": 0.9602,
+      "step": 12921
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3490366057156945e-06,
+      "loss": 0.7947,
+      "step": 12922
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.348346569123239e-06,
+      "loss": 0.8149,
+      "step": 12923
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.347656545156422e-06,
+      "loss": 0.8755,
+      "step": 12924
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.346966533828611e-06,
+      "loss": 0.8216,
+      "step": 12925
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.346276535153175e-06,
+      "loss": 0.7972,
+      "step": 12926
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.345586549143484e-06,
+      "loss": 0.8057,
+      "step": 12927
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.344896575812904e-06,
+      "loss": 0.9203,
+      "step": 12928
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3442066151748045e-06,
+      "loss": 0.9439,
+      "step": 12929
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.343516667242553e-06,
+      "loss": 0.9064,
+      "step": 12930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3428267320295176e-06,
+      "loss": 0.7334,
+      "step": 12931
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.342136809549063e-06,
+      "loss": 0.7764,
+      "step": 12932
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.341446899814561e-06,
+      "loss": 1.1404,
+      "step": 12933
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.340757002839375e-06,
+      "loss": 0.9241,
+      "step": 12934
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.340067118636873e-06,
+      "loss": 0.9252,
+      "step": 12935
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3393772472204205e-06,
+      "loss": 0.896,
+      "step": 12936
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.338687388603384e-06,
+      "loss": 1.0602,
+      "step": 12937
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.337997542799129e-06,
+      "loss": 0.9061,
+      "step": 12938
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.337307709821023e-06,
+      "loss": 0.9563,
+      "step": 12939
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.336617889682428e-06,
+      "loss": 0.9363,
+      "step": 12940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.335928082396712e-06,
+      "loss": 0.9176,
+      "step": 12941
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.335238287977239e-06,
+      "loss": 0.9374,
+      "step": 12942
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.334548506437371e-06,
+      "loss": 0.844,
+      "step": 12943
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.333858737790476e-06,
+      "loss": 0.8967,
+      "step": 12944
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.333168982049918e-06,
+      "loss": 0.8681,
+      "step": 12945
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.332479239229059e-06,
+      "loss": 0.7515,
+      "step": 12946
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.331789509341263e-06,
+      "loss": 0.9718,
+      "step": 12947
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.331099792399895e-06,
+      "loss": 1.1078,
+      "step": 12948
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.330410088418315e-06,
+      "loss": 0.8782,
+      "step": 12949
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.32972039740989e-06,
+      "loss": 0.8742,
+      "step": 12950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.329030719387978e-06,
+      "loss": 0.9069,
+      "step": 12951
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.328341054365945e-06,
+      "loss": 0.8566,
+      "step": 12952
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.327651402357151e-06,
+      "loss": 0.8604,
+      "step": 12953
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.326961763374959e-06,
+      "loss": 0.8181,
+      "step": 12954
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.326272137432728e-06,
+      "loss": 0.9941,
+      "step": 12955
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.325582524543824e-06,
+      "loss": 1.0257,
+      "step": 12956
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.324892924721605e-06,
+      "loss": 0.8395,
+      "step": 12957
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.324203337979432e-06,
+      "loss": 0.9433,
+      "step": 12958
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.323513764330667e-06,
+      "loss": 0.7294,
+      "step": 12959
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.322824203788669e-06,
+      "loss": 0.9322,
+      "step": 12960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.322134656366798e-06,
+      "loss": 0.8574,
+      "step": 12961
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.321445122078415e-06,
+      "loss": 1.0298,
+      "step": 12962
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.320755600936879e-06,
+      "loss": 0.8919,
+      "step": 12963
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.320066092955549e-06,
+      "loss": 0.948,
+      "step": 12964
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.319376598147783e-06,
+      "loss": 0.8598,
+      "step": 12965
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.318687116526942e-06,
+      "loss": 0.9189,
+      "step": 12966
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3179976481063805e-06,
+      "loss": 1.1038,
+      "step": 12967
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.317308192899463e-06,
+      "loss": 0.992,
+      "step": 12968
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.316618750919545e-06,
+      "loss": 1.0872,
+      "step": 12969
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.315929322179982e-06,
+      "loss": 0.9013,
+      "step": 12970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3152399066941345e-06,
+      "loss": 0.9593,
+      "step": 12971
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.314550504475358e-06,
+      "loss": 1.0761,
+      "step": 12972
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3138611155370095e-06,
+      "loss": 0.8712,
+      "step": 12973
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.313171739892447e-06,
+      "loss": 0.9683,
+      "step": 12974
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3124823775550265e-06,
+      "loss": 0.9658,
+      "step": 12975
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.311793028538104e-06,
+      "loss": 0.9467,
+      "step": 12976
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.311103692855034e-06,
+      "loss": 0.8851,
+      "step": 12977
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.310414370519173e-06,
+      "loss": 0.8847,
+      "step": 12978
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.309725061543879e-06,
+      "loss": 0.9044,
+      "step": 12979
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.309035765942505e-06,
+      "loss": 0.9891,
+      "step": 12980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.308346483728407e-06,
+      "loss": 1.1052,
+      "step": 12981
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.307657214914938e-06,
+      "loss": 0.8898,
+      "step": 12982
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.306967959515454e-06,
+      "loss": 0.8419,
+      "step": 12983
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3062787175433085e-06,
+      "loss": 1.0038,
+      "step": 12984
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3055894890118544e-06,
+      "loss": 0.7908,
+      "step": 12985
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.304900273934447e-06,
+      "loss": 0.955,
+      "step": 12986
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3042110723244405e-06,
+      "loss": 0.9243,
+      "step": 12987
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.303521884195186e-06,
+      "loss": 0.9312,
+      "step": 12988
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.302832709560036e-06,
+      "loss": 1.0247,
+      "step": 12989
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.302143548432344e-06,
+      "loss": 1.0335,
+      "step": 12990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.301454400825462e-06,
+      "loss": 0.912,
+      "step": 12991
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.300765266752744e-06,
+      "loss": 0.8347,
+      "step": 12992
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.300076146227541e-06,
+      "loss": 0.8517,
+      "step": 12993
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.299387039263203e-06,
+      "loss": 0.8673,
+      "step": 12994
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.298697945873082e-06,
+      "loss": 0.7269,
+      "step": 12995
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.298008866070529e-06,
+      "loss": 1.021,
+      "step": 12996
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2973197998688956e-06,
+      "loss": 1.0342,
+      "step": 12997
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.296630747281531e-06,
+      "loss": 0.8852,
+      "step": 12998
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.295941708321787e-06,
+      "loss": 0.8755,
+      "step": 12999
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.295252683003013e-06,
+      "loss": 0.8842,
+      "step": 13000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.294563671338555e-06,
+      "loss": 1.0018,
+      "step": 13001
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.293874673341769e-06,
+      "loss": 1.0715,
+      "step": 13002
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.293185689026e-06,
+      "loss": 0.8297,
+      "step": 13003
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.292496718404598e-06,
+      "loss": 0.9936,
+      "step": 13004
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.291807761490913e-06,
+      "loss": 0.8878,
+      "step": 13005
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.291118818298291e-06,
+      "loss": 0.8677,
+      "step": 13006
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.290429888840081e-06,
+      "loss": 0.8866,
+      "step": 13007
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.289740973129632e-06,
+      "loss": 1.0219,
+      "step": 13008
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.289052071180289e-06,
+      "loss": 0.7961,
+      "step": 13009
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.288363183005402e-06,
+      "loss": 0.9,
+      "step": 13010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.287674308618316e-06,
+      "loss": 0.943,
+      "step": 13011
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.286985448032379e-06,
+      "loss": 0.9254,
+      "step": 13012
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.286296601260937e-06,
+      "loss": 0.8956,
+      "step": 13013
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.285607768317336e-06,
+      "loss": 0.9889,
+      "step": 13014
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.284918949214924e-06,
+      "loss": 0.9018,
+      "step": 13015
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.284230143967043e-06,
+      "loss": 0.8211,
+      "step": 13016
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.283541352587043e-06,
+      "loss": 0.8495,
+      "step": 13017
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.282852575088266e-06,
+      "loss": 0.9221,
+      "step": 13018
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.282163811484057e-06,
+      "loss": 0.883,
+      "step": 13019
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2814750617877615e-06,
+      "loss": 0.9229,
+      "step": 13020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.280786326012724e-06,
+      "loss": 1.0536,
+      "step": 13021
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.280097604172288e-06,
+      "loss": 0.8901,
+      "step": 13022
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.279408896279797e-06,
+      "loss": 0.7485,
+      "step": 13023
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.278720202348597e-06,
+      "loss": 0.9794,
+      "step": 13024
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.278031522392026e-06,
+      "loss": 0.9646,
+      "step": 13025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.277342856423433e-06,
+      "loss": 0.8294,
+      "step": 13026
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.276654204456158e-06,
+      "loss": 0.8638,
+      "step": 13027
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.275965566503544e-06,
+      "loss": 0.8911,
+      "step": 13028
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2752769425789334e-06,
+      "loss": 0.9399,
+      "step": 13029
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.274588332695667e-06,
+      "loss": 0.9684,
+      "step": 13030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.273899736867086e-06,
+      "loss": 0.8593,
+      "step": 13031
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.273211155106535e-06,
+      "loss": 0.9483,
+      "step": 13032
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.272522587427352e-06,
+      "loss": 1.0122,
+      "step": 13033
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.271834033842879e-06,
+      "loss": 0.8916,
+      "step": 13034
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.271145494366456e-06,
+      "loss": 0.9256,
+      "step": 13035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2704569690114215e-06,
+      "loss": 0.9059,
+      "step": 13036
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.269768457791121e-06,
+      "loss": 1.0633,
+      "step": 13037
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.26907996071889e-06,
+      "loss": 0.9028,
+      "step": 13038
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.268391477808069e-06,
+      "loss": 0.9928,
+      "step": 13039
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.267703009071997e-06,
+      "loss": 0.9097,
+      "step": 13040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.267014554524014e-06,
+      "loss": 0.9493,
+      "step": 13041
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.266326114177457e-06,
+      "loss": 1.024,
+      "step": 13042
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.265637688045665e-06,
+      "loss": 0.7976,
+      "step": 13043
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.264949276141977e-06,
+      "loss": 1.1175,
+      "step": 13044
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2642608784797295e-06,
+      "loss": 0.9963,
+      "step": 13045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.26357249507226e-06,
+      "loss": 0.8463,
+      "step": 13046
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.262884125932908e-06,
+      "loss": 0.8796,
+      "step": 13047
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.262195771075006e-06,
+      "loss": 0.989,
+      "step": 13048
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.261507430511896e-06,
+      "loss": 0.8723,
+      "step": 13049
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.260819104256912e-06,
+      "loss": 1.0142,
+      "step": 13050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2601307923233895e-06,
+      "loss": 1.0548,
+      "step": 13051
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.259442494724665e-06,
+      "loss": 0.975,
+      "step": 13052
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.258754211474074e-06,
+      "loss": 0.9691,
+      "step": 13053
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.258065942584953e-06,
+      "loss": 0.8711,
+      "step": 13054
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2573776880706346e-06,
+      "loss": 0.9286,
+      "step": 13055
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.256689447944457e-06,
+      "loss": 1.0056,
+      "step": 13056
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.256001222219751e-06,
+      "loss": 0.9372,
+      "step": 13057
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.255313010909852e-06,
+      "loss": 1.0945,
+      "step": 13058
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.254624814028096e-06,
+      "loss": 1.0855,
+      "step": 13059
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.253936631587813e-06,
+      "loss": 0.7591,
+      "step": 13060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.253248463602341e-06,
+      "loss": 0.9469,
+      "step": 13061
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.252560310085009e-06,
+      "loss": 0.9712,
+      "step": 13062
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.251872171049153e-06,
+      "loss": 0.9342,
+      "step": 13063
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.251184046508104e-06,
+      "loss": 1.0469,
+      "step": 13064
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.250495936475193e-06,
+      "loss": 0.8871,
+      "step": 13065
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.249807840963754e-06,
+      "loss": 0.8662,
+      "step": 13066
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.249119759987118e-06,
+      "loss": 0.824,
+      "step": 13067
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.248431693558616e-06,
+      "loss": 1.0226,
+      "step": 13068
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.247743641691579e-06,
+      "loss": 0.9382,
+      "step": 13069
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2470556043993395e-06,
+      "loss": 0.9209,
+      "step": 13070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.246367581695224e-06,
+      "loss": 0.9287,
+      "step": 13071
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.245679573592568e-06,
+      "loss": 0.9677,
+      "step": 13072
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2449915801047e-06,
+      "loss": 1.132,
+      "step": 13073
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2443036012449475e-06,
+      "loss": 0.804,
+      "step": 13074
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.243615637026641e-06,
+      "loss": 0.9179,
+      "step": 13075
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.242927687463111e-06,
+      "loss": 0.984,
+      "step": 13076
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.242239752567685e-06,
+      "loss": 0.9006,
+      "step": 13077
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.241551832353692e-06,
+      "loss": 0.8056,
+      "step": 13078
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.24086392683446e-06,
+      "loss": 0.9667,
+      "step": 13079
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2401760360233175e-06,
+      "loss": 0.8424,
+      "step": 13080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.239488159933592e-06,
+      "loss": 0.9069,
+      "step": 13081
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.23880029857861e-06,
+      "loss": 1.0455,
+      "step": 13082
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.238112451971698e-06,
+      "loss": 0.9713,
+      "step": 13083
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.237424620126187e-06,
+      "loss": 1.1231,
+      "step": 13084
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2367368030554e-06,
+      "loss": 0.8327,
+      "step": 13085
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.236049000772665e-06,
+      "loss": 0.86,
+      "step": 13086
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.235361213291307e-06,
+      "loss": 0.9515,
+      "step": 13087
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.234673440624653e-06,
+      "loss": 1.0001,
+      "step": 13088
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.233985682786026e-06,
+      "loss": 0.951,
+      "step": 13089
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2332979397887546e-06,
+      "loss": 0.9968,
+      "step": 13090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.23261021164616e-06,
+      "loss": 1.0661,
+      "step": 13091
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.231922498371569e-06,
+      "loss": 0.9291,
+      "step": 13092
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.231234799978305e-06,
+      "loss": 0.9425,
+      "step": 13093
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.230547116479691e-06,
+      "loss": 1.0726,
+      "step": 13094
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.229859447889053e-06,
+      "loss": 0.9563,
+      "step": 13095
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2291717942197145e-06,
+      "loss": 0.8802,
+      "step": 13096
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.228484155484998e-06,
+      "loss": 0.9985,
+      "step": 13097
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.227796531698225e-06,
+      "loss": 0.9024,
+      "step": 13098
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.227108922872719e-06,
+      "loss": 0.9247,
+      "step": 13099
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.226421329021802e-06,
+      "loss": 0.9837,
+      "step": 13100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.225733750158798e-06,
+      "loss": 0.9736,
+      "step": 13101
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.225046186297026e-06,
+      "loss": 1.0954,
+      "step": 13102
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.224358637449808e-06,
+      "loss": 1.0457,
+      "step": 13103
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.223671103630466e-06,
+      "loss": 0.7827,
+      "step": 13104
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.22298358485232e-06,
+      "loss": 0.9336,
+      "step": 13105
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.22229608112869e-06,
+      "loss": 0.9409,
+      "step": 13106
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.221608592472897e-06,
+      "loss": 0.8545,
+      "step": 13107
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.220921118898263e-06,
+      "loss": 0.9958,
+      "step": 13108
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2202336604181035e-06,
+      "loss": 0.9772,
+      "step": 13109
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.219546217045741e-06,
+      "loss": 1.0051,
+      "step": 13110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2188587887944936e-06,
+      "loss": 0.974,
+      "step": 13111
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2181713756776794e-06,
+      "loss": 0.9161,
+      "step": 13112
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.217483977708617e-06,
+      "loss": 0.8246,
+      "step": 13113
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2167965949006266e-06,
+      "loss": 1.1328,
+      "step": 13114
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.216109227267024e-06,
+      "loss": 0.9241,
+      "step": 13115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.215421874821126e-06,
+      "loss": 0.8509,
+      "step": 13116
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2147345375762525e-06,
+      "loss": 0.8055,
+      "step": 13117
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.214047215545717e-06,
+      "loss": 0.9601,
+      "step": 13118
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.21335990874284e-06,
+      "loss": 0.9662,
+      "step": 13119
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.212672617180937e-06,
+      "loss": 0.8408,
+      "step": 13120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2119853408733226e-06,
+      "loss": 0.8549,
+      "step": 13121
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2112980798333146e-06,
+      "loss": 0.7597,
+      "step": 13122
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2106108340742264e-06,
+      "loss": 0.8054,
+      "step": 13123
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2099236036093744e-06,
+      "loss": 1.0987,
+      "step": 13124
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.209236388452074e-06,
+      "loss": 0.9253,
+      "step": 13125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2085491886156385e-06,
+      "loss": 0.9513,
+      "step": 13126
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.207862004113383e-06,
+      "loss": 0.9337,
+      "step": 13127
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.207174834958622e-06,
+      "loss": 1.1322,
+      "step": 13128
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.206487681164666e-06,
+      "loss": 1.0994,
+      "step": 13129
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2058005427448334e-06,
+      "loss": 0.8368,
+      "step": 13130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2051134197124354e-06,
+      "loss": 0.9717,
+      "step": 13131
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.204426312080785e-06,
+      "loss": 1.0051,
+      "step": 13132
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.203739219863193e-06,
+      "loss": 0.7675,
+      "step": 13133
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.203052143072974e-06,
+      "loss": 1.0122,
+      "step": 13134
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.202365081723439e-06,
+      "loss": 1.15,
+      "step": 13135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.201678035827899e-06,
+      "loss": 0.987,
+      "step": 13136
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.200991005399666e-06,
+      "loss": 0.9532,
+      "step": 13137
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.200303990452052e-06,
+      "loss": 0.8822,
+      "step": 13138
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.199616990998366e-06,
+      "loss": 0.8572,
+      "step": 13139
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.198930007051919e-06,
+      "loss": 0.9988,
+      "step": 13140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.198243038626021e-06,
+      "loss": 0.998,
+      "step": 13141
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.197556085733983e-06,
+      "loss": 0.8448,
+      "step": 13142
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.196869148389114e-06,
+      "loss": 0.9354,
+      "step": 13143
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1961822266047225e-06,
+      "loss": 0.9298,
+      "step": 13144
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.195495320394118e-06,
+      "loss": 0.9814,
+      "step": 13145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.19480842977061e-06,
+      "loss": 0.9913,
+      "step": 13146
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.194121554747506e-06,
+      "loss": 0.8505,
+      "step": 13147
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.193434695338113e-06,
+      "loss": 0.9667,
+      "step": 13148
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1927478515557414e-06,
+      "loss": 0.9973,
+      "step": 13149
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.192061023413696e-06,
+      "loss": 0.7032,
+      "step": 13150
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.191374210925285e-06,
+      "loss": 0.8123,
+      "step": 13151
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.190687414103816e-06,
+      "loss": 0.923,
+      "step": 13152
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.190000632962591e-06,
+      "loss": 0.9902,
+      "step": 13153
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1893138675149235e-06,
+      "loss": 0.7962,
+      "step": 13154
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.188627117774115e-06,
+      "loss": 0.7853,
+      "step": 13155
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.187940383753474e-06,
+      "loss": 1.0001,
+      "step": 13156
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.187253665466302e-06,
+      "loss": 0.931,
+      "step": 13157
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.186566962925906e-06,
+      "loss": 0.9887,
+      "step": 13158
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1858802761455915e-06,
+      "loss": 0.9083,
+      "step": 13159
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.185193605138662e-06,
+      "loss": 1.0136,
+      "step": 13160
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.18450694991842e-06,
+      "loss": 0.9435,
+      "step": 13161
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.183820310498172e-06,
+      "loss": 1.0399,
+      "step": 13162
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.18313368689122e-06,
+      "loss": 0.8418,
+      "step": 13163
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1824470791108675e-06,
+      "loss": 0.889,
+      "step": 13164
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.181760487170418e-06,
+      "loss": 1.019,
+      "step": 13165
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.181073911083172e-06,
+      "loss": 0.8354,
+      "step": 13166
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.180387350862435e-06,
+      "loss": 0.9922,
+      "step": 13167
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.179700806521506e-06,
+      "loss": 0.9644,
+      "step": 13168
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.179014278073689e-06,
+      "loss": 1.0242,
+      "step": 13169
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.178327765532283e-06,
+      "loss": 0.8672,
+      "step": 13170
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.177641268910591e-06,
+      "loss": 1.011,
+      "step": 13171
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1769547882219115e-06,
+      "loss": 1.0019,
+      "step": 13172
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.176268323479547e-06,
+      "loss": 0.9039,
+      "step": 13173
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.175581874696798e-06,
+      "loss": 0.879,
+      "step": 13174
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1748954418869616e-06,
+      "loss": 0.8262,
+      "step": 13175
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1742090250633375e-06,
+      "loss": 1.091,
+      "step": 13176
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.173522624239228e-06,
+      "loss": 0.8785,
+      "step": 13177
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.17283623942793e-06,
+      "loss": 0.9106,
+      "step": 13178
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.172149870642743e-06,
+      "loss": 0.9162,
+      "step": 13179
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.171463517896964e-06,
+      "loss": 0.8897,
+      "step": 13180
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.170777181203892e-06,
+      "loss": 0.7696,
+      "step": 13181
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.170090860576823e-06,
+      "loss": 1.023,
+      "step": 13182
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.169404556029056e-06,
+      "loss": 0.7435,
+      "step": 13183
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.168718267573887e-06,
+      "loss": 1.031,
+      "step": 13184
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1680319952246135e-06,
+      "loss": 1.0478,
+      "step": 13185
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.167345738994531e-06,
+      "loss": 0.8533,
+      "step": 13186
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.166659498896933e-06,
+      "loss": 0.9939,
+      "step": 13187
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.165973274945122e-06,
+      "loss": 0.9121,
+      "step": 13188
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.165287067152388e-06,
+      "loss": 0.9911,
+      "step": 13189
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.164600875532029e-06,
+      "loss": 0.9102,
+      "step": 13190
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.163914700097338e-06,
+      "loss": 1.0087,
+      "step": 13191
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1632285408616095e-06,
+      "loss": 1.1056,
+      "step": 13192
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.162542397838139e-06,
+      "loss": 0.9209,
+      "step": 13193
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.161856271040219e-06,
+      "loss": 1.0786,
+      "step": 13194
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.161170160481145e-06,
+      "loss": 0.9959,
+      "step": 13195
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1604840661742065e-06,
+      "loss": 0.9532,
+      "step": 13196
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.159797988132701e-06,
+      "loss": 0.9547,
+      "step": 13197
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.159111926369918e-06,
+      "loss": 0.9192,
+      "step": 13198
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.158425880899149e-06,
+      "loss": 1.0645,
+      "step": 13199
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.15773985173369e-06,
+      "loss": 0.968,
+      "step": 13200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.157053838886828e-06,
+      "loss": 0.8055,
+      "step": 13201
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.156367842371859e-06,
+      "loss": 0.954,
+      "step": 13202
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.155681862202071e-06,
+      "loss": 0.895,
+      "step": 13203
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.154995898390756e-06,
+      "loss": 0.8024,
+      "step": 13204
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.154309950951203e-06,
+      "loss": 1.0947,
+      "step": 13205
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1536240198967024e-06,
+      "loss": 0.8397,
+      "step": 13206
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.152938105240546e-06,
+      "loss": 1.1437,
+      "step": 13207
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.152252206996021e-06,
+      "loss": 0.84,
+      "step": 13208
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.151566325176416e-06,
+      "loss": 0.8174,
+      "step": 13209
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.150880459795022e-06,
+      "loss": 0.9953,
+      "step": 13210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.150194610865124e-06,
+      "loss": 0.9156,
+      "step": 13211
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.149508778400015e-06,
+      "loss": 0.9243,
+      "step": 13212
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.14882296241298e-06,
+      "loss": 0.9457,
+      "step": 13213
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.148137162917309e-06,
+      "loss": 1.121,
+      "step": 13214
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.147451379926285e-06,
+      "loss": 0.952,
+      "step": 13215
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.146765613453198e-06,
+      "loss": 0.8503,
+      "step": 13216
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.146079863511333e-06,
+      "loss": 0.9412,
+      "step": 13217
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1453941301139774e-06,
+      "loss": 0.9354,
+      "step": 13218
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.144708413274418e-06,
+      "loss": 1.0257,
+      "step": 13219
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.144022713005937e-06,
+      "loss": 0.7284,
+      "step": 13220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.143337029321823e-06,
+      "loss": 0.8928,
+      "step": 13221
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.142651362235357e-06,
+      "loss": 1.0383,
+      "step": 13222
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1419657117598285e-06,
+      "loss": 0.9082,
+      "step": 13223
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.14128007790852e-06,
+      "loss": 0.9227,
+      "step": 13224
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.140594460694715e-06,
+      "loss": 0.906,
+      "step": 13225
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.139908860131698e-06,
+      "loss": 0.7856,
+      "step": 13226
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.139223276232751e-06,
+      "loss": 0.8261,
+      "step": 13227
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.138537709011158e-06,
+      "loss": 1.1561,
+      "step": 13228
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.137852158480202e-06,
+      "loss": 0.8238,
+      "step": 13229
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.137166624653164e-06,
+      "loss": 1.0799,
+      "step": 13230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.136481107543328e-06,
+      "loss": 0.9717,
+      "step": 13231
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.135795607163973e-06,
+      "loss": 1.0271,
+      "step": 13232
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.135110123528383e-06,
+      "loss": 0.8913,
+      "step": 13233
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.134424656649837e-06,
+      "loss": 0.9236,
+      "step": 13234
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.133739206541618e-06,
+      "loss": 0.7336,
+      "step": 13235
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.133053773217006e-06,
+      "loss": 1.1032,
+      "step": 13236
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.13236835668928e-06,
+      "loss": 0.962,
+      "step": 13237
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.131682956971721e-06,
+      "loss": 0.9838,
+      "step": 13238
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.130997574077607e-06,
+      "loss": 0.9927,
+      "step": 13239
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.130312208020217e-06,
+      "loss": 0.7943,
+      "step": 13240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1296268588128315e-06,
+      "loss": 0.8533,
+      "step": 13241
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.128941526468728e-06,
+      "loss": 0.962,
+      "step": 13242
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1282562110011854e-06,
+      "loss": 1.0214,
+      "step": 13243
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1275709124234805e-06,
+      "loss": 0.6502,
+      "step": 13244
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.12688563074889e-06,
+      "loss": 0.781,
+      "step": 13245
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.126200365990694e-06,
+      "loss": 1.1231,
+      "step": 13246
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.125515118162167e-06,
+      "loss": 0.7761,
+      "step": 13247
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.124829887276586e-06,
+      "loss": 0.8858,
+      "step": 13248
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.124144673347228e-06,
+      "loss": 1.0685,
+      "step": 13249
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.123459476387369e-06,
+      "loss": 1.0505,
+      "step": 13250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.122774296410284e-06,
+      "loss": 1.0071,
+      "step": 13251
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.122089133429246e-06,
+      "loss": 0.9199,
+      "step": 13252
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1214039874575335e-06,
+      "loss": 1.0377,
+      "step": 13253
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.120718858508419e-06,
+      "loss": 0.8346,
+      "step": 13254
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.120033746595178e-06,
+      "loss": 1.0749,
+      "step": 13255
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.119348651731082e-06,
+      "loss": 0.7916,
+      "step": 13256
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.118663573929406e-06,
+      "loss": 0.8854,
+      "step": 13257
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.117978513203424e-06,
+      "loss": 0.8726,
+      "step": 13258
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.117293469566409e-06,
+      "loss": 0.6879,
+      "step": 13259
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1166084430316315e-06,
+      "loss": 0.9786,
+      "step": 13260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.115923433612366e-06,
+      "loss": 1.0164,
+      "step": 13261
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1152384413218835e-06,
+      "loss": 0.937,
+      "step": 13262
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1145534661734545e-06,
+      "loss": 0.9064,
+      "step": 13263
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.113868508180352e-06,
+      "loss": 0.8717,
+      "step": 13264
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.113183567355846e-06,
+      "loss": 1.0012,
+      "step": 13265
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.112498643713207e-06,
+      "loss": 0.9534,
+      "step": 13266
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.111813737265706e-06,
+      "loss": 0.9925,
+      "step": 13267
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.111128848026613e-06,
+      "loss": 0.9356,
+      "step": 13268
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.110443976009194e-06,
+      "loss": 1.0228,
+      "step": 13269
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.109759121226723e-06,
+      "loss": 0.9536,
+      "step": 13270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1090742836924665e-06,
+      "loss": 0.9369,
+      "step": 13271
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1083894634196954e-06,
+      "loss": 1.1068,
+      "step": 13272
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.107704660421675e-06,
+      "loss": 0.9118,
+      "step": 13273
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.107019874711674e-06,
+      "loss": 0.8745,
+      "step": 13274
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.106335106302962e-06,
+      "loss": 1.0432,
+      "step": 13275
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.105650355208804e-06,
+      "loss": 0.8092,
+      "step": 13276
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1049656214424665e-06,
+      "loss": 0.8564,
+      "step": 13277
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.104280905017218e-06,
+      "loss": 0.9304,
+      "step": 13278
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.103596205946323e-06,
+      "loss": 1.057,
+      "step": 13279
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.102911524243047e-06,
+      "loss": 1.1231,
+      "step": 13280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.102226859920658e-06,
+      "loss": 0.8395,
+      "step": 13281
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.10154221299242e-06,
+      "loss": 1.0183,
+      "step": 13282
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.100857583471598e-06,
+      "loss": 1.0191,
+      "step": 13283
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.100172971371457e-06,
+      "loss": 0.814,
+      "step": 13284
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0994883767052605e-06,
+      "loss": 0.9372,
+      "step": 13285
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.098803799486272e-06,
+      "loss": 1.0429,
+      "step": 13286
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.098119239727754e-06,
+      "loss": 0.8697,
+      "step": 13287
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.097434697442974e-06,
+      "loss": 0.9696,
+      "step": 13288
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0967501726451905e-06,
+      "loss": 1.0539,
+      "step": 13289
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.096065665347667e-06,
+      "loss": 1.0249,
+      "step": 13290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.095381175563667e-06,
+      "loss": 1.2149,
+      "step": 13291
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.09469670330645e-06,
+      "loss": 0.8319,
+      "step": 13292
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0940122485892805e-06,
+      "loss": 0.881,
+      "step": 13293
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0933278114254185e-06,
+      "loss": 1.0708,
+      "step": 13294
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.092643391828124e-06,
+      "loss": 1.0221,
+      "step": 13295
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.091958989810657e-06,
+      "loss": 1.0215,
+      "step": 13296
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.09127460538628e-06,
+      "loss": 0.9052,
+      "step": 13297
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.090590238568251e-06,
+      "loss": 1.074,
+      "step": 13298
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.08990588936983e-06,
+      "loss": 0.8938,
+      "step": 13299
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.089221557804276e-06,
+      "loss": 0.9617,
+      "step": 13300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.088537243884847e-06,
+      "loss": 0.9662,
+      "step": 13301
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.087852947624803e-06,
+      "loss": 1.0567,
+      "step": 13302
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.087168669037401e-06,
+      "loss": 1.122,
+      "step": 13303
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.086484408135896e-06,
+      "loss": 0.9337,
+      "step": 13304
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.085800164933551e-06,
+      "loss": 1.0613,
+      "step": 13305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.085115939443621e-06,
+      "loss": 0.8312,
+      "step": 13306
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.084431731679362e-06,
+      "loss": 0.8646,
+      "step": 13307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0837475416540305e-06,
+      "loss": 0.9433,
+      "step": 13308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.083063369380882e-06,
+      "loss": 0.9785,
+      "step": 13309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.082379214873173e-06,
+      "loss": 0.865,
+      "step": 13310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.081695078144159e-06,
+      "loss": 0.9059,
+      "step": 13311
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.081010959207094e-06,
+      "loss": 0.863,
+      "step": 13312
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.080326858075233e-06,
+      "loss": 0.8859,
+      "step": 13313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.079642774761831e-06,
+      "loss": 0.9217,
+      "step": 13314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.078958709280139e-06,
+      "loss": 1.0667,
+      "step": 13315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.078274661643415e-06,
+      "loss": 0.8958,
+      "step": 13316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.077590631864911e-06,
+      "loss": 0.8525,
+      "step": 13317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0769066199578785e-06,
+      "loss": 0.9789,
+      "step": 13318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.076222625935572e-06,
+      "loss": 0.9646,
+      "step": 13319
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.075538649811242e-06,
+      "loss": 0.9078,
+      "step": 13320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.074854691598141e-06,
+      "loss": 0.8899,
+      "step": 13321
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0741707513095204e-06,
+      "loss": 0.9256,
+      "step": 13322
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.073486828958631e-06,
+      "loss": 1.0017,
+      "step": 13323
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.072802924558724e-06,
+      "loss": 0.9527,
+      "step": 13324
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.072119038123051e-06,
+      "loss": 0.9452,
+      "step": 13325
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.071435169664861e-06,
+      "loss": 0.8162,
+      "step": 13326
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.070751319197402e-06,
+      "loss": 0.8965,
+      "step": 13327
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.070067486733928e-06,
+      "loss": 1.0359,
+      "step": 13328
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.069383672287684e-06,
+      "loss": 0.9838,
+      "step": 13329
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.068699875871921e-06,
+      "loss": 0.7653,
+      "step": 13330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.068016097499887e-06,
+      "loss": 0.8403,
+      "step": 13331
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.067332337184829e-06,
+      "loss": 0.9228,
+      "step": 13332
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.066648594939996e-06,
+      "loss": 0.8736,
+      "step": 13333
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.065964870778634e-06,
+      "loss": 0.9875,
+      "step": 13334
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.065281164713991e-06,
+      "loss": 1.088,
+      "step": 13335
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.064597476759313e-06,
+      "loss": 0.8982,
+      "step": 13336
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.063913806927848e-06,
+      "loss": 0.8645,
+      "step": 13337
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.063230155232837e-06,
+      "loss": 0.9934,
+      "step": 13338
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0625465216875326e-06,
+      "loss": 0.9343,
+      "step": 13339
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.061862906305177e-06,
+      "loss": 0.9961,
+      "step": 13340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0611793090990135e-06,
+      "loss": 0.9537,
+      "step": 13341
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.060495730082288e-06,
+      "loss": 0.9656,
+      "step": 13342
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.059812169268245e-06,
+      "loss": 1.0487,
+      "step": 13343
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.059128626670128e-06,
+      "loss": 0.9105,
+      "step": 13344
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.058445102301181e-06,
+      "loss": 0.8573,
+      "step": 13345
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.057761596174646e-06,
+      "loss": 1.0267,
+      "step": 13346
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.057078108303766e-06,
+      "loss": 0.8919,
+      "step": 13347
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.056394638701784e-06,
+      "loss": 0.941,
+      "step": 13348
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.055711187381942e-06,
+      "loss": 0.97,
+      "step": 13349
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.05502775435748e-06,
+      "loss": 0.9144,
+      "step": 13350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.054344339641643e-06,
+      "loss": 1.0626,
+      "step": 13351
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.053660943247669e-06,
+      "loss": 0.9073,
+      "step": 13352
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0529775651888e-06,
+      "loss": 0.8544,
+      "step": 13353
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.052294205478276e-06,
+      "loss": 0.9883,
+      "step": 13354
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.051610864129336e-06,
+      "loss": 0.8951,
+      "step": 13355
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.050927541155221e-06,
+      "loss": 0.914,
+      "step": 13356
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0502442365691695e-06,
+      "loss": 0.9081,
+      "step": 13357
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.04956095038442e-06,
+      "loss": 0.9962,
+      "step": 13358
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.048877682614212e-06,
+      "loss": 0.6747,
+      "step": 13359
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.048194433271783e-06,
+      "loss": 0.9792,
+      "step": 13360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0475112023703715e-06,
+      "loss": 1.0499,
+      "step": 13361
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0468279899232114e-06,
+      "loss": 0.9179,
+      "step": 13362
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.046144795943546e-06,
+      "loss": 0.8472,
+      "step": 13363
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.045461620444607e-06,
+      "loss": 0.8446,
+      "step": 13364
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.044778463439634e-06,
+      "loss": 0.9462,
+      "step": 13365
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.044095324941861e-06,
+      "loss": 1.0344,
+      "step": 13366
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.043412204964526e-06,
+      "loss": 1.0933,
+      "step": 13367
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0427291035208605e-06,
+      "loss": 1.0253,
+      "step": 13368
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.042046020624102e-06,
+      "loss": 1.0571,
+      "step": 13369
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0413629562874845e-06,
+      "loss": 0.7625,
+      "step": 13370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0406799105242425e-06,
+      "loss": 0.9715,
+      "step": 13371
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.039996883347609e-06,
+      "loss": 0.965,
+      "step": 13372
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.039313874770817e-06,
+      "loss": 0.8937,
+      "step": 13373
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.038630884807102e-06,
+      "loss": 0.9387,
+      "step": 13374
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.037947913469696e-06,
+      "loss": 0.862,
+      "step": 13375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.03726496077183e-06,
+      "loss": 1.055,
+      "step": 13376
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0365820267267384e-06,
+      "loss": 0.8166,
+      "step": 13377
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.035899111347651e-06,
+      "loss": 0.8785,
+      "step": 13378
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0352162146477994e-06,
+      "loss": 0.9074,
+      "step": 13379
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.034533336640415e-06,
+      "loss": 1.2276,
+      "step": 13380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.033850477338728e-06,
+      "loss": 1.0631,
+      "step": 13381
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.033167636755969e-06,
+      "loss": 1.0594,
+      "step": 13382
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.032484814905368e-06,
+      "loss": 0.8684,
+      "step": 13383
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.031802011800154e-06,
+      "loss": 1.1218,
+      "step": 13384
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.031119227453555e-06,
+      "loss": 0.7623,
+      "step": 13385
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.030436461878803e-06,
+      "loss": 1.0681,
+      "step": 13386
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.029753715089124e-06,
+      "loss": 0.9471,
+      "step": 13387
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.029070987097747e-06,
+      "loss": 0.8172,
+      "step": 13388
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.028388277917899e-06,
+      "loss": 0.9736,
+      "step": 13389
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.027705587562808e-06,
+      "loss": 1.0291,
+      "step": 13390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0270229160457e-06,
+      "loss": 0.8971,
+      "step": 13391
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.026340263379804e-06,
+      "loss": 1.072,
+      "step": 13392
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.025657629578343e-06,
+      "loss": 0.9193,
+      "step": 13393
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0249750146545445e-06,
+      "loss": 1.0635,
+      "step": 13394
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.024292418621634e-06,
+      "loss": 0.8065,
+      "step": 13395
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.023609841492837e-06,
+      "loss": 1.0539,
+      "step": 13396
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.022927283281376e-06,
+      "loss": 0.9906,
+      "step": 13397
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0222447440004785e-06,
+      "loss": 1.0287,
+      "step": 13398
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.021562223663368e-06,
+      "loss": 0.9628,
+      "step": 13399
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.020879722283267e-06,
+      "loss": 0.7837,
+      "step": 13400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.020197239873399e-06,
+      "loss": 0.8874,
+      "step": 13401
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.019514776446988e-06,
+      "loss": 0.9251,
+      "step": 13402
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.018832332017255e-06,
+      "loss": 1.0119,
+      "step": 13403
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.018149906597424e-06,
+      "loss": 0.9037,
+      "step": 13404
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.017467500200716e-06,
+      "loss": 0.9728,
+      "step": 13405
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.016785112840351e-06,
+      "loss": 1.0319,
+      "step": 13406
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.016102744529552e-06,
+      "loss": 0.8561,
+      "step": 13407
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.015420395281539e-06,
+      "loss": 0.92,
+      "step": 13408
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.014738065109532e-06,
+      "loss": 0.9365,
+      "step": 13409
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.014055754026752e-06,
+      "loss": 1.0196,
+      "step": 13410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.013373462046418e-06,
+      "loss": 1.0062,
+      "step": 13411
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0126911891817485e-06,
+      "loss": 0.9449,
+      "step": 13412
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.012008935445964e-06,
+      "loss": 0.9257,
+      "step": 13413
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.011326700852284e-06,
+      "loss": 0.9202,
+      "step": 13414
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0106444854139235e-06,
+      "loss": 0.923,
+      "step": 13415
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.009962289144101e-06,
+      "loss": 0.9857,
+      "step": 13416
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.009280112056035e-06,
+      "loss": 0.8003,
+      "step": 13417
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.008597954162943e-06,
+      "loss": 0.8359,
+      "step": 13418
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0079158154780405e-06,
+      "loss": 0.8781,
+      "step": 13419
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0072336960145425e-06,
+      "loss": 1.0111,
+      "step": 13420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.006551595785667e-06,
+      "loss": 0.8582,
+      "step": 13421
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.00586951480463e-06,
+      "loss": 0.9302,
+      "step": 13422
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.005187453084645e-06,
+      "loss": 0.9489,
+      "step": 13423
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.004505410638928e-06,
+      "loss": 0.8645,
+      "step": 13424
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.003823387480693e-06,
+      "loss": 0.8413,
+      "step": 13425
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.003141383623153e-06,
+      "loss": 1.0176,
+      "step": 13426
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.002459399079523e-06,
+      "loss": 0.9704,
+      "step": 13427
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.001777433863016e-06,
+      "loss": 1.164,
+      "step": 13428
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.001095487986845e-06,
+      "loss": 0.8274,
+      "step": 13429
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.000413561464222e-06,
+      "loss": 1.0803,
+      "step": 13430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.999731654308357e-06,
+      "loss": 0.9122,
+      "step": 13431
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9990497665324666e-06,
+      "loss": 0.9147,
+      "step": 13432
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.99836789814976e-06,
+      "loss": 1.1078,
+      "step": 13433
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.997686049173448e-06,
+      "loss": 0.9444,
+      "step": 13434
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.997004219616742e-06,
+      "loss": 1.0087,
+      "step": 13435
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.996322409492852e-06,
+      "loss": 1.065,
+      "step": 13436
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.995640618814986e-06,
+      "loss": 0.9898,
+      "step": 13437
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.994958847596357e-06,
+      "loss": 1.0552,
+      "step": 13438
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.994277095850172e-06,
+      "loss": 0.9434,
+      "step": 13439
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.993595363589639e-06,
+      "loss": 0.888,
+      "step": 13440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9929136508279684e-06,
+      "loss": 0.8421,
+      "step": 13441
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.992231957578367e-06,
+      "loss": 0.9318,
+      "step": 13442
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.991550283854042e-06,
+      "loss": 0.8957,
+      "step": 13443
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.990868629668203e-06,
+      "loss": 0.9378,
+      "step": 13444
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.990186995034055e-06,
+      "loss": 1.0455,
+      "step": 13445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.989505379964805e-06,
+      "loss": 0.843,
+      "step": 13446
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.988823784473659e-06,
+      "loss": 0.9088,
+      "step": 13447
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.988142208573822e-06,
+      "loss": 1.0014,
+      "step": 13448
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.987460652278501e-06,
+      "loss": 0.9083,
+      "step": 13449
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.986779115600901e-06,
+      "loss": 1.0253,
+      "step": 13450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9860975985542245e-06,
+      "loss": 0.8175,
+      "step": 13451
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.985416101151677e-06,
+      "loss": 0.9489,
+      "step": 13452
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.984734623406463e-06,
+      "loss": 0.866,
+      "step": 13453
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.984053165331785e-06,
+      "loss": 1.1061,
+      "step": 13454
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.983371726940844e-06,
+      "loss": 0.8761,
+      "step": 13455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.982690308246848e-06,
+      "loss": 0.8992,
+      "step": 13456
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.982008909262996e-06,
+      "loss": 0.9323,
+      "step": 13457
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.981327530002492e-06,
+      "loss": 1.046,
+      "step": 13458
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.980646170478535e-06,
+      "loss": 0.9338,
+      "step": 13459
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.979964830704326e-06,
+      "loss": 0.8098,
+      "step": 13460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9792835106930685e-06,
+      "loss": 0.9583,
+      "step": 13461
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.978602210457962e-06,
+      "loss": 0.9285,
+      "step": 13462
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.977920930012204e-06,
+      "loss": 1.0209,
+      "step": 13463
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.977239669368998e-06,
+      "loss": 0.9174,
+      "step": 13464
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.97655842854154e-06,
+      "loss": 0.9191,
+      "step": 13465
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.975877207543029e-06,
+      "loss": 0.8432,
+      "step": 13466
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.975196006386666e-06,
+      "loss": 0.8239,
+      "step": 13467
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.974514825085648e-06,
+      "loss": 0.913,
+      "step": 13468
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.973833663653174e-06,
+      "loss": 0.8627,
+      "step": 13469
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.973152522102439e-06,
+      "loss": 0.9256,
+      "step": 13470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9724714004466406e-06,
+      "loss": 0.9072,
+      "step": 13471
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.971790298698977e-06,
+      "loss": 1.0048,
+      "step": 13472
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.971109216872642e-06,
+      "loss": 1.0759,
+      "step": 13473
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.970428154980833e-06,
+      "loss": 0.9785,
+      "step": 13474
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.969747113036745e-06,
+      "loss": 0.9445,
+      "step": 13475
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.969066091053573e-06,
+      "loss": 0.7834,
+      "step": 13476
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.968385089044511e-06,
+      "loss": 0.8882,
+      "step": 13477
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.967704107022754e-06,
+      "loss": 0.9175,
+      "step": 13478
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.967023145001496e-06,
+      "loss": 0.9099,
+      "step": 13479
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9663422029939305e-06,
+      "loss": 0.9548,
+      "step": 13480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.96566128101325e-06,
+      "loss": 0.8677,
+      "step": 13481
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.964980379072648e-06,
+      "loss": 0.9107,
+      "step": 13482
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.964299497185315e-06,
+      "loss": 0.9922,
+      "step": 13483
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9636186353644465e-06,
+      "loss": 0.9509,
+      "step": 13484
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.962937793623231e-06,
+      "loss": 1.1903,
+      "step": 13485
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9622569719748596e-06,
+      "loss": 0.9148,
+      "step": 13486
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.961576170432525e-06,
+      "loss": 0.877,
+      "step": 13487
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.960895389009416e-06,
+      "loss": 0.87,
+      "step": 13488
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.960214627718724e-06,
+      "loss": 1.1673,
+      "step": 13489
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.959533886573635e-06,
+      "loss": 0.9448,
+      "step": 13490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9588531655873425e-06,
+      "loss": 0.8615,
+      "step": 13491
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.958172464773034e-06,
+      "loss": 0.9857,
+      "step": 13492
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.957491784143898e-06,
+      "loss": 0.7975,
+      "step": 13493
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.956811123713123e-06,
+      "loss": 0.8636,
+      "step": 13494
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.956130483493894e-06,
+      "loss": 0.9075,
+      "step": 13495
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.955449863499401e-06,
+      "loss": 0.8956,
+      "step": 13496
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.95476926374283e-06,
+      "loss": 0.8204,
+      "step": 13497
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9540886842373674e-06,
+      "loss": 0.8636,
+      "step": 13498
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9534081249962e-06,
+      "loss": 0.9899,
+      "step": 13499
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.952727586032512e-06,
+      "loss": 0.9038,
+      "step": 13500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.952047067359488e-06,
+      "loss": 0.9404,
+      "step": 13501
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.951366568990316e-06,
+      "loss": 0.9432,
+      "step": 13502
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.950686090938179e-06,
+      "loss": 1.0292,
+      "step": 13503
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.950005633216261e-06,
+      "loss": 1.0128,
+      "step": 13504
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.949325195837745e-06,
+      "loss": 0.965,
+      "step": 13505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.948644778815816e-06,
+      "loss": 1.1383,
+      "step": 13506
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.947964382163655e-06,
+      "loss": 0.8478,
+      "step": 13507
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.947284005894445e-06,
+      "loss": 0.9565,
+      "step": 13508
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.94660365002137e-06,
+      "loss": 0.9492,
+      "step": 13509
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.945923314557609e-06,
+      "loss": 1.1121,
+      "step": 13510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.945242999516345e-06,
+      "loss": 1.0044,
+      "step": 13511
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.944562704910759e-06,
+      "loss": 0.8126,
+      "step": 13512
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.94388243075403e-06,
+      "loss": 0.9098,
+      "step": 13513
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.943202177059341e-06,
+      "loss": 0.8653,
+      "step": 13514
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.942521943839869e-06,
+      "loss": 1.2589,
+      "step": 13515
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9418417311087954e-06,
+      "loss": 1.1171,
+      "step": 13516
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.941161538879298e-06,
+      "loss": 1.0195,
+      "step": 13517
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.940481367164555e-06,
+      "loss": 0.8063,
+      "step": 13518
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.939801215977746e-06,
+      "loss": 0.98,
+      "step": 13519
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.939121085332047e-06,
+      "loss": 0.996,
+      "step": 13520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.938440975240637e-06,
+      "loss": 0.9876,
+      "step": 13521
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9377608857166925e-06,
+      "loss": 0.98,
+      "step": 13522
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.93708081677339e-06,
+      "loss": 1.1567,
+      "step": 13523
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.936400768423903e-06,
+      "loss": 0.9352,
+      "step": 13524
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.935720740681412e-06,
+      "loss": 0.929,
+      "step": 13525
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.93504073355909e-06,
+      "loss": 0.8776,
+      "step": 13526
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.934360747070113e-06,
+      "loss": 0.9659,
+      "step": 13527
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.933680781227654e-06,
+      "loss": 0.7849,
+      "step": 13528
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.933000836044889e-06,
+      "loss": 0.9121,
+      "step": 13529
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.932320911534989e-06,
+      "loss": 0.9197,
+      "step": 13530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.931641007711131e-06,
+      "loss": 0.9005,
+      "step": 13531
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.930961124586485e-06,
+      "loss": 0.9904,
+      "step": 13532
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.930281262174225e-06,
+      "loss": 0.798,
+      "step": 13533
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9296014204875235e-06,
+      "loss": 1.077,
+      "step": 13534
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9289215995395505e-06,
+      "loss": 0.9311,
+      "step": 13535
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.928241799343478e-06,
+      "loss": 0.8808,
+      "step": 13536
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.927562019912479e-06,
+      "loss": 0.8923,
+      "step": 13537
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.926882261259723e-06,
+      "loss": 0.9547,
+      "step": 13538
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.926202523398378e-06,
+      "loss": 0.9407,
+      "step": 13539
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.925522806341617e-06,
+      "loss": 1.0239,
+      "step": 13540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.924843110102607e-06,
+      "loss": 0.9577,
+      "step": 13541
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9241634346945185e-06,
+      "loss": 0.9984,
+      "step": 13542
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.923483780130519e-06,
+      "loss": 0.9077,
+      "step": 13543
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.922804146423778e-06,
+      "loss": 1.0125,
+      "step": 13544
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9221245335874615e-06,
+      "loss": 0.9398,
+      "step": 13545
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.921444941634737e-06,
+      "loss": 1.0861,
+      "step": 13546
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.920765370578773e-06,
+      "loss": 0.9278,
+      "step": 13547
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.920085820432733e-06,
+      "loss": 1.0318,
+      "step": 13548
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.919406291209787e-06,
+      "loss": 0.9123,
+      "step": 13549
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.918726782923099e-06,
+      "loss": 0.9582,
+      "step": 13550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.918047295585835e-06,
+      "loss": 0.8829,
+      "step": 13551
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.917367829211159e-06,
+      "loss": 0.9859,
+      "step": 13552
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.916688383812235e-06,
+      "loss": 0.9016,
+      "step": 13553
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.916008959402229e-06,
+      "loss": 0.9446,
+      "step": 13554
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9153295559943025e-06,
+      "loss": 0.8454,
+      "step": 13555
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.91465017360162e-06,
+      "loss": 0.9126,
+      "step": 13556
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.913970812237344e-06,
+      "loss": 1.1145,
+      "step": 13557
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.913291471914638e-06,
+      "loss": 0.9845,
+      "step": 13558
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.91261215264666e-06,
+      "loss": 0.8706,
+      "step": 13559
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.911932854446578e-06,
+      "loss": 1.0265,
+      "step": 13560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.91125357732755e-06,
+      "loss": 1.1023,
+      "step": 13561
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.910574321302737e-06,
+      "loss": 0.9846,
+      "step": 13562
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.909895086385299e-06,
+      "loss": 0.9184,
+      "step": 13563
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.909215872588396e-06,
+      "loss": 0.8757,
+      "step": 13564
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.908536679925189e-06,
+      "loss": 0.9109,
+      "step": 13565
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.907857508408836e-06,
+      "loss": 0.8202,
+      "step": 13566
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.907178358052496e-06,
+      "loss": 0.9504,
+      "step": 13567
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.906499228869327e-06,
+      "loss": 0.7339,
+      "step": 13568
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.905820120872488e-06,
+      "loss": 0.9482,
+      "step": 13569
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.905141034075135e-06,
+      "loss": 0.8334,
+      "step": 13570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.904461968490426e-06,
+      "loss": 0.9792,
+      "step": 13571
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.903782924131518e-06,
+      "loss": 0.8734,
+      "step": 13572
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9031039010115675e-06,
+      "loss": 1.0762,
+      "step": 13573
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.90242489914373e-06,
+      "loss": 0.8384,
+      "step": 13574
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.901745918541162e-06,
+      "loss": 0.9524,
+      "step": 13575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.901066959217017e-06,
+      "loss": 0.8498,
+      "step": 13576
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.900388021184451e-06,
+      "loss": 0.9664,
+      "step": 13577
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.899709104456617e-06,
+      "loss": 0.8497,
+      "step": 13578
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.89903020904667e-06,
+      "loss": 0.9384,
+      "step": 13579
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.898351334967763e-06,
+      "loss": 1.0311,
+      "step": 13580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.897672482233048e-06,
+      "loss": 0.9002,
+      "step": 13581
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.896993650855679e-06,
+      "loss": 1.0358,
+      "step": 13582
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8963148408488065e-06,
+      "loss": 0.7765,
+      "step": 13583
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8956360522255856e-06,
+      "loss": 0.8325,
+      "step": 13584
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.894957284999165e-06,
+      "loss": 0.8503,
+      "step": 13585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.894278539182697e-06,
+      "loss": 0.8139,
+      "step": 13586
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.89359981478933e-06,
+      "loss": 0.9176,
+      "step": 13587
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.892921111832217e-06,
+      "loss": 0.8363,
+      "step": 13588
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8922424303245065e-06,
+      "loss": 0.9784,
+      "step": 13589
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.891563770279347e-06,
+      "loss": 0.9939,
+      "step": 13590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.890885131709888e-06,
+      "loss": 0.9123,
+      "step": 13591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.890206514629279e-06,
+      "loss": 0.9685,
+      "step": 13592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.889527919050666e-06,
+      "loss": 0.8542,
+      "step": 13593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.888849344987197e-06,
+      "loss": 1.1293,
+      "step": 13594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.88817079245202e-06,
+      "loss": 1.079,
+      "step": 13595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.887492261458282e-06,
+      "loss": 0.9745,
+      "step": 13596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.88681375201913e-06,
+      "loss": 1.0327,
+      "step": 13597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.886135264147709e-06,
+      "loss": 1.1114,
+      "step": 13598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.885456797857164e-06,
+      "loss": 0.8447,
+      "step": 13599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.884778353160641e-06,
+      "loss": 1.0497,
+      "step": 13600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.884099930071284e-06,
+      "loss": 1.0413,
+      "step": 13601
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.883421528602238e-06,
+      "loss": 0.747,
+      "step": 13602
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8827431487666465e-06,
+      "loss": 0.9817,
+      "step": 13603
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.882064790577653e-06,
+      "loss": 1.0513,
+      "step": 13604
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.881386454048401e-06,
+      "loss": 0.9643,
+      "step": 13605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.880708139192031e-06,
+      "loss": 1.1607,
+      "step": 13606
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.880029846021688e-06,
+      "loss": 0.9128,
+      "step": 13607
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8793515745505125e-06,
+      "loss": 0.9818,
+      "step": 13608
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.878673324791647e-06,
+      "loss": 0.9108,
+      "step": 13609
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.877995096758232e-06,
+      "loss": 0.9337,
+      "step": 13610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.877316890463406e-06,
+      "loss": 0.888,
+      "step": 13611
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.876638705920312e-06,
+      "loss": 0.914,
+      "step": 13612
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.875960543142087e-06,
+      "loss": 0.808,
+      "step": 13613
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8752824021418715e-06,
+      "loss": 0.98,
+      "step": 13614
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8746042829328046e-06,
+      "loss": 0.9088,
+      "step": 13615
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.873926185528025e-06,
+      "loss": 0.8971,
+      "step": 13616
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.873248109940667e-06,
+      "loss": 0.8285,
+      "step": 13617
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8725700561838745e-06,
+      "loss": 1.0127,
+      "step": 13618
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.871892024270781e-06,
+      "loss": 1.0061,
+      "step": 13619
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8712140142145235e-06,
+      "loss": 1.1168,
+      "step": 13620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.870536026028238e-06,
+      "loss": 0.9281,
+      "step": 13621
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8698580597250615e-06,
+      "loss": 0.9767,
+      "step": 13622
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.869180115318129e-06,
+      "loss": 1.0615,
+      "step": 13623
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.868502192820574e-06,
+      "loss": 0.8376,
+      "step": 13624
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.867824292245534e-06,
+      "loss": 0.8354,
+      "step": 13625
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.86714641360614e-06,
+      "loss": 1.0722,
+      "step": 13626
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8664685569155274e-06,
+      "loss": 1.1512,
+      "step": 13627
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.86579072218683e-06,
+      "loss": 0.9561,
+      "step": 13628
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.865112909433179e-06,
+      "loss": 1.0377,
+      "step": 13629
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.864435118667709e-06,
+      "loss": 0.9833,
+      "step": 13630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.863757349903551e-06,
+      "loss": 0.9118,
+      "step": 13631
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.863079603153836e-06,
+      "loss": 0.7203,
+      "step": 13632
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.862401878431696e-06,
+      "loss": 1.0433,
+      "step": 13633
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.861724175750262e-06,
+      "loss": 0.8745,
+      "step": 13634
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.861046495122664e-06,
+      "loss": 1.0153,
+      "step": 13635
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.860368836562031e-06,
+      "loss": 0.8152,
+      "step": 13636
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.859691200081494e-06,
+      "loss": 0.8827,
+      "step": 13637
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.859013585694181e-06,
+      "loss": 0.9977,
+      "step": 13638
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.858335993413221e-06,
+      "loss": 0.9381,
+      "step": 13639
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.857658423251742e-06,
+      "loss": 0.9024,
+      "step": 13640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.85698087522287e-06,
+      "loss": 1.0863,
+      "step": 13641
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8563033493397365e-06,
+      "loss": 0.9733,
+      "step": 13642
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8556258456154665e-06,
+      "loss": 1.016,
+      "step": 13643
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.854948364063186e-06,
+      "loss": 1.0444,
+      "step": 13644
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8542709046960205e-06,
+      "loss": 0.8564,
+      "step": 13645
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.853593467527097e-06,
+      "loss": 0.883,
+      "step": 13646
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.85291605256954e-06,
+      "loss": 0.9434,
+      "step": 13647
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.852238659836474e-06,
+      "loss": 1.0214,
+      "step": 13648
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.851561289341023e-06,
+      "loss": 0.9586,
+      "step": 13649
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.850883941096313e-06,
+      "loss": 0.9192,
+      "step": 13650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.850206615115465e-06,
+      "loss": 0.9036,
+      "step": 13651
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.849529311411601e-06,
+      "loss": 1.0327,
+      "step": 13652
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.848852029997847e-06,
+      "loss": 0.8535,
+      "step": 13653
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.848174770887325e-06,
+      "loss": 1.0784,
+      "step": 13654
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.847497534093156e-06,
+      "loss": 0.839,
+      "step": 13655
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.846820319628459e-06,
+      "loss": 0.9326,
+      "step": 13656
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.846143127506358e-06,
+      "loss": 1.0005,
+      "step": 13657
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.845465957739971e-06,
+      "loss": 0.9778,
+      "step": 13658
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8447888103424206e-06,
+      "loss": 0.9431,
+      "step": 13659
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.844111685326824e-06,
+      "loss": 1.0029,
+      "step": 13660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8434345827063024e-06,
+      "loss": 0.8507,
+      "step": 13661
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.842757502493972e-06,
+      "loss": 0.9958,
+      "step": 13662
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.842080444702953e-06,
+      "loss": 0.9259,
+      "step": 13663
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.841403409346363e-06,
+      "loss": 1.0602,
+      "step": 13664
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.840726396437319e-06,
+      "loss": 0.9928,
+      "step": 13665
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.840049405988937e-06,
+      "loss": 0.8904,
+      "step": 13666
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.839372438014336e-06,
+      "loss": 1.0602,
+      "step": 13667
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.83869549252663e-06,
+      "loss": 1.0018,
+      "step": 13668
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.838018569538937e-06,
+      "loss": 0.9034,
+      "step": 13669
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.837341669064369e-06,
+      "loss": 0.9565,
+      "step": 13670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.836664791116043e-06,
+      "loss": 0.9697,
+      "step": 13671
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.835987935707073e-06,
+      "loss": 1.0522,
+      "step": 13672
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.835311102850572e-06,
+      "loss": 0.9721,
+      "step": 13673
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.834634292559655e-06,
+      "loss": 0.866,
+      "step": 13674
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8339575048474345e-06,
+      "loss": 1.0381,
+      "step": 13675
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.83328073972702e-06,
+      "loss": 0.8681,
+      "step": 13676
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.832603997211528e-06,
+      "loss": 0.9165,
+      "step": 13677
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.83192727731407e-06,
+      "loss": 1.1019,
+      "step": 13678
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.831250580047755e-06,
+      "loss": 1.0154,
+      "step": 13679
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.830573905425696e-06,
+      "loss": 0.9644,
+      "step": 13680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8298972534610015e-06,
+      "loss": 0.9505,
+      "step": 13681
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8292206241667824e-06,
+      "loss": 0.8993,
+      "step": 13682
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.828544017556148e-06,
+      "loss": 1.0724,
+      "step": 13683
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.827867433642209e-06,
+      "loss": 0.8844,
+      "step": 13684
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8271908724380705e-06,
+      "loss": 0.9077,
+      "step": 13685
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.826514333956843e-06,
+      "loss": 0.9586,
+      "step": 13686
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.825837818211634e-06,
+      "loss": 0.9605,
+      "step": 13687
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.825161325215552e-06,
+      "loss": 0.8879,
+      "step": 13688
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.824484854981703e-06,
+      "loss": 1.0525,
+      "step": 13689
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.823808407523192e-06,
+      "loss": 0.9727,
+      "step": 13690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.823131982853127e-06,
+      "loss": 0.9284,
+      "step": 13691
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.822455580984613e-06,
+      "loss": 0.8758,
+      "step": 13692
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.821779201930755e-06,
+      "loss": 0.9599,
+      "step": 13693
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8211028457046575e-06,
+      "loss": 0.9637,
+      "step": 13694
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.820426512319426e-06,
+      "loss": 0.9999,
+      "step": 13695
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.819750201788162e-06,
+      "loss": 0.877,
+      "step": 13696
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8190739141239705e-06,
+      "loss": 0.8096,
+      "step": 13697
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.818397649339954e-06,
+      "loss": 0.7811,
+      "step": 13698
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.817721407449213e-06,
+      "loss": 1.0223,
+      "step": 13699
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8170451884648534e-06,
+      "loss": 0.749,
+      "step": 13700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8163689923999755e-06,
+      "loss": 0.8451,
+      "step": 13701
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.815692819267679e-06,
+      "loss": 0.9101,
+      "step": 13702
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.815016669081066e-06,
+      "loss": 0.8424,
+      "step": 13703
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8143405418532362e-06,
+      "loss": 0.8271,
+      "step": 13704
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8136644375972896e-06,
+      "loss": 0.8451,
+      "step": 13705
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8129883563263252e-06,
+      "loss": 0.8169,
+      "step": 13706
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.812312298053442e-06,
+      "loss": 1.0015,
+      "step": 13707
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.811636262791739e-06,
+      "loss": 0.9687,
+      "step": 13708
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8109602505543136e-06,
+      "loss": 1.1299,
+      "step": 13709
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.810284261354262e-06,
+      "loss": 1.0276,
+      "step": 13710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8096082952046843e-06,
+      "loss": 1.0491,
+      "step": 13711
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.808932352118676e-06,
+      "loss": 0.9907,
+      "step": 13712
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.808256432109334e-06,
+      "loss": 0.9792,
+      "step": 13713
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8075805351897523e-06,
+      "loss": 1.0622,
+      "step": 13714
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8069046613730276e-06,
+      "loss": 0.8057,
+      "step": 13715
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8062288106722547e-06,
+      "loss": 0.8835,
+      "step": 13716
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.805552983100528e-06,
+      "loss": 1.1194,
+      "step": 13717
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8048771786709405e-06,
+      "loss": 0.8853,
+      "step": 13718
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.804201397396587e-06,
+      "loss": 0.7623,
+      "step": 13719
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.80352563929056e-06,
+      "loss": 1.0268,
+      "step": 13720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.802849904365953e-06,
+      "loss": 0.838,
+      "step": 13721
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8021741926358555e-06,
+      "loss": 1.013,
+      "step": 13722
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8014985041133627e-06,
+      "loss": 1.0013,
+      "step": 13723
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.800822838811565e-06,
+      "loss": 1.1026,
+      "step": 13724
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.800147196743552e-06,
+      "loss": 0.9733,
+      "step": 13725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7994715779224154e-06,
+      "loss": 0.9717,
+      "step": 13726
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7987959823612447e-06,
+      "loss": 0.9834,
+      "step": 13727
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7981204100731297e-06,
+      "loss": 0.7053,
+      "step": 13728
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.797444861071158e-06,
+      "loss": 0.9435,
+      "step": 13729
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.79676933536842e-06,
+      "loss": 0.9469,
+      "step": 13730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7960938329780033e-06,
+      "loss": 0.8428,
+      "step": 13731
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7954183539129953e-06,
+      "loss": 0.9392,
+      "step": 13732
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.794742898186484e-06,
+      "loss": 0.8722,
+      "step": 13733
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7940674658115527e-06,
+      "loss": 0.9697,
+      "step": 13734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7933920568012927e-06,
+      "loss": 0.8216,
+      "step": 13735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7927166711687884e-06,
+      "loss": 1.0422,
+      "step": 13736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7920413089271248e-06,
+      "loss": 0.9021,
+      "step": 13737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.791365970089387e-06,
+      "loss": 0.9681,
+      "step": 13738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.790690654668659e-06,
+      "loss": 1.0482,
+      "step": 13739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7900153626780255e-06,
+      "loss": 0.9764,
+      "step": 13740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7893400941305695e-06,
+      "loss": 1.0616,
+      "step": 13741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7886648490393744e-06,
+      "loss": 0.9047,
+      "step": 13742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.787989627417524e-06,
+      "loss": 0.8769,
+      "step": 13743
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.787314429278099e-06,
+      "loss": 0.8449,
+      "step": 13744
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7866392546341797e-06,
+      "loss": 1.0325,
+      "step": 13745
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7859641034988513e-06,
+      "loss": 1.0139,
+      "step": 13746
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7852889758851942e-06,
+      "loss": 1.0435,
+      "step": 13747
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.784613871806287e-06,
+      "loss": 0.9589,
+      "step": 13748
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.78393879127521e-06,
+      "loss": 0.8597,
+      "step": 13749
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.783263734305044e-06,
+      "loss": 0.8375,
+      "step": 13750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7825887009088656e-06,
+      "loss": 0.9633,
+      "step": 13751
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7819136910997567e-06,
+      "loss": 0.9457,
+      "step": 13752
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.781238704890793e-06,
+      "loss": 1.0118,
+      "step": 13753
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7805637422950523e-06,
+      "loss": 1.1019,
+      "step": 13754
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.779888803325613e-06,
+      "loss": 1.0452,
+      "step": 13755
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.779213887995552e-06,
+      "loss": 1.0471,
+      "step": 13756
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.778538996317943e-06,
+      "loss": 0.9172,
+      "step": 13757
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7778641283058655e-06,
+      "loss": 0.9841,
+      "step": 13758
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7771892839723928e-06,
+      "loss": 0.9072,
+      "step": 13759
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7765144633306006e-06,
+      "loss": 0.9754,
+      "step": 13760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7758396663935633e-06,
+      "loss": 0.9675,
+      "step": 13761
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7751648931743544e-06,
+      "loss": 0.8833,
+      "step": 13762
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7744901436860484e-06,
+      "loss": 0.8551,
+      "step": 13763
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7738154179417174e-06,
+      "loss": 1.0113,
+      "step": 13764
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.773140715954435e-06,
+      "loss": 0.9204,
+      "step": 13765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7724660377372725e-06,
+      "loss": 1.11,
+      "step": 13766
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.771791383303302e-06,
+      "loss": 0.9542,
+      "step": 13767
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.771116752665593e-06,
+      "loss": 0.8746,
+      "step": 13768
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.77044214583722e-06,
+      "loss": 0.9945,
+      "step": 13769
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.769767562831252e-06,
+      "loss": 0.8118,
+      "step": 13770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7690930036607575e-06,
+      "loss": 1.1017,
+      "step": 13771
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.768418468338808e-06,
+      "loss": 1.0169,
+      "step": 13772
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7677439568784707e-06,
+      "loss": 0.9882,
+      "step": 13773
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7670694692928146e-06,
+      "loss": 0.7489,
+      "step": 13774
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.766395005594908e-06,
+      "loss": 0.9857,
+      "step": 13775
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7657205657978184e-06,
+      "loss": 0.9645,
+      "step": 13776
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.765046149914613e-06,
+      "loss": 0.921,
+      "step": 13777
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7643717579583583e-06,
+      "loss": 0.9487,
+      "step": 13778
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7636973899421202e-06,
+      "loss": 1.0202,
+      "step": 13779
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7630230458789643e-06,
+      "loss": 1.0403,
+      "step": 13780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7623487257819563e-06,
+      "loss": 0.9718,
+      "step": 13781
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7616744296641617e-06,
+      "loss": 1.0733,
+      "step": 13782
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.761000157538644e-06,
+      "loss": 1.1228,
+      "step": 13783
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.760325909418467e-06,
+      "loss": 0.8431,
+      "step": 13784
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7596516853166943e-06,
+      "loss": 0.9361,
+      "step": 13785
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.758977485246389e-06,
+      "loss": 0.8659,
+      "step": 13786
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7583033092206132e-06,
+      "loss": 0.9313,
+      "step": 13787
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.75762915725243e-06,
+      "loss": 0.917,
+      "step": 13788
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7569550293548988e-06,
+      "loss": 0.9345,
+      "step": 13789
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7562809255410826e-06,
+      "loss": 0.9997,
+      "step": 13790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7556068458240413e-06,
+      "loss": 1.0027,
+      "step": 13791
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.754932790216833e-06,
+      "loss": 0.7379,
+      "step": 13792
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.754258758732522e-06,
+      "loss": 0.9647,
+      "step": 13793
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7535847513841644e-06,
+      "loss": 0.9516,
+      "step": 13794
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.75291076818482e-06,
+      "loss": 0.8951,
+      "step": 13795
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7522368091475458e-06,
+      "loss": 0.8395,
+      "step": 13796
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7515628742854006e-06,
+      "loss": 0.8169,
+      "step": 13797
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.750888963611442e-06,
+      "loss": 0.7935,
+      "step": 13798
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.750215077138726e-06,
+      "loss": 0.9687,
+      "step": 13799
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.749541214880309e-06,
+      "loss": 1.0041,
+      "step": 13800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.748867376849248e-06,
+      "loss": 0.8931,
+      "step": 13801
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7481935630585973e-06,
+      "loss": 0.9821,
+      "step": 13802
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7475197735214108e-06,
+      "loss": 0.8532,
+      "step": 13803
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.746846008250746e-06,
+      "loss": 0.9356,
+      "step": 13804
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7461722672596564e-06,
+      "loss": 0.9267,
+      "step": 13805
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.745498550561194e-06,
+      "loss": 0.878,
+      "step": 13806
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7448248581684125e-06,
+      "loss": 0.9846,
+      "step": 13807
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.744151190094365e-06,
+      "loss": 0.8306,
+      "step": 13808
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7434775463521034e-06,
+      "loss": 0.8767,
+      "step": 13809
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.742803926954679e-06,
+      "loss": 0.9501,
+      "step": 13810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7421303319151434e-06,
+      "loss": 0.9271,
+      "step": 13811
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.741456761246548e-06,
+      "loss": 0.8617,
+      "step": 13812
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7407832149619416e-06,
+      "loss": 0.8198,
+      "step": 13813
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7401096930743753e-06,
+      "loss": 0.9511,
+      "step": 13814
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.739436195596896e-06,
+      "loss": 0.9025,
+      "step": 13815
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.738762722542556e-06,
+      "loss": 1.0614,
+      "step": 13816
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7380892739244024e-06,
+      "loss": 0.8008,
+      "step": 13817
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7374158497554824e-06,
+      "loss": 0.8709,
+      "step": 13818
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7367424500488448e-06,
+      "loss": 0.8827,
+      "step": 13819
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.736069074817535e-06,
+      "loss": 0.8425,
+      "step": 13820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7353957240746e-06,
+      "loss": 0.9555,
+      "step": 13821
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.734722397833086e-06,
+      "loss": 1.0113,
+      "step": 13822
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7340490961060387e-06,
+      "loss": 0.8797,
+      "step": 13823
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7333758189065034e-06,
+      "loss": 1.0695,
+      "step": 13824
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7327025662475234e-06,
+      "loss": 0.9557,
+      "step": 13825
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7320293381421448e-06,
+      "loss": 0.8479,
+      "step": 13826
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.731356134603407e-06,
+      "loss": 1.0441,
+      "step": 13827
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.730682955644359e-06,
+      "loss": 1.0942,
+      "step": 13828
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7300098012780404e-06,
+      "loss": 0.9331,
+      "step": 13829
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7293366715174945e-06,
+      "loss": 0.9931,
+      "step": 13830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.728663566375762e-06,
+      "loss": 0.8538,
+      "step": 13831
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7279904858658843e-06,
+      "loss": 0.9117,
+      "step": 13832
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.727317430000903e-06,
+      "loss": 0.8694,
+      "step": 13833
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.726644398793857e-06,
+      "loss": 0.6556,
+      "step": 13834
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7259713922577877e-06,
+      "loss": 0.9053,
+      "step": 13835
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.725298410405733e-06,
+      "loss": 0.7901,
+      "step": 13836
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.724625453250733e-06,
+      "loss": 0.8818,
+      "step": 13837
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7239525208058227e-06,
+      "loss": 0.95,
+      "step": 13838
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.723279613084045e-06,
+      "loss": 0.9904,
+      "step": 13839
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.722606730098436e-06,
+      "loss": 0.9342,
+      "step": 13840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7219338718620314e-06,
+      "loss": 1.0021,
+      "step": 13841
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7212610383878684e-06,
+      "loss": 0.9228,
+      "step": 13842
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.720588229688983e-06,
+      "loss": 0.9153,
+      "step": 13843
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.719915445778409e-06,
+      "loss": 0.9452,
+      "step": 13844
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7192426866691844e-06,
+      "loss": 0.9477,
+      "step": 13845
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7185699523743413e-06,
+      "loss": 1.0579,
+      "step": 13846
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7178972429069147e-06,
+      "loss": 0.8922,
+      "step": 13847
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.717224558279939e-06,
+      "loss": 0.9218,
+      "step": 13848
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.716551898506446e-06,
+      "loss": 0.9392,
+      "step": 13849
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.715879263599468e-06,
+      "loss": 1.0799,
+      "step": 13850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.715206653572039e-06,
+      "loss": 0.7867,
+      "step": 13851
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.714534068437189e-06,
+      "loss": 0.9613,
+      "step": 13852
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.71386150820795e-06,
+      "loss": 0.929,
+      "step": 13853
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.713188972897353e-06,
+      "loss": 1.0876,
+      "step": 13854
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.712516462518427e-06,
+      "loss": 0.8354,
+      "step": 13855
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7118439770842035e-06,
+      "loss": 0.9332,
+      "step": 13856
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7111715166077107e-06,
+      "loss": 0.8922,
+      "step": 13857
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.710499081101977e-06,
+      "loss": 0.8814,
+      "step": 13858
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7098266705800312e-06,
+      "loss": 0.9467,
+      "step": 13859
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7091542850549005e-06,
+      "loss": 1.0058,
+      "step": 13860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.708481924539611e-06,
+      "loss": 0.8193,
+      "step": 13861
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.707809589047193e-06,
+      "loss": 0.9446,
+      "step": 13862
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7071372785906716e-06,
+      "loss": 1.0464,
+      "step": 13863
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.706464993183072e-06,
+      "loss": 0.8573,
+      "step": 13864
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.70579273283742e-06,
+      "loss": 1.0408,
+      "step": 13865
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.70512049756674e-06,
+      "loss": 1.0065,
+      "step": 13866
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7044482873840566e-06,
+      "loss": 0.9481,
+      "step": 13867
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7037761023023937e-06,
+      "loss": 1.0518,
+      "step": 13868
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.703103942334775e-06,
+      "loss": 0.8759,
+      "step": 13869
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.702431807494223e-06,
+      "loss": 1.1373,
+      "step": 13870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.701759697793761e-06,
+      "loss": 0.8317,
+      "step": 13871
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.701087613246409e-06,
+      "loss": 0.9849,
+      "step": 13872
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7004155538651897e-06,
+      "loss": 0.9786,
+      "step": 13873
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6997435196631255e-06,
+      "loss": 0.7458,
+      "step": 13874
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6990715106532353e-06,
+      "loss": 0.9024,
+      "step": 13875
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6983995268485396e-06,
+      "loss": 0.9221,
+      "step": 13876
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.697727568262058e-06,
+      "loss": 0.8392,
+      "step": 13877
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.697055634906809e-06,
+      "loss": 1.0828,
+      "step": 13878
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6963837267958115e-06,
+      "loss": 0.9056,
+      "step": 13879
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6957118439420835e-06,
+      "loss": 1.0586,
+      "step": 13880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6950399863586433e-06,
+      "loss": 0.9559,
+      "step": 13881
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6943681540585067e-06,
+      "loss": 0.8803,
+      "step": 13882
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6936963470546915e-06,
+      "loss": 0.8215,
+      "step": 13883
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6930245653602127e-06,
+      "loss": 0.9205,
+      "step": 13884
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6923528089880846e-06,
+      "loss": 0.8133,
+      "step": 13885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.691681077951327e-06,
+      "loss": 1.0491,
+      "step": 13886
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6910093722629504e-06,
+      "loss": 0.845,
+      "step": 13887
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.690337691935971e-06,
+      "loss": 0.9349,
+      "step": 13888
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.689666036983402e-06,
+      "loss": 0.9827,
+      "step": 13889
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.688994407418256e-06,
+      "loss": 0.928,
+      "step": 13890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6883228032535455e-06,
+      "loss": 0.9101,
+      "step": 13891
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6876512245022846e-06,
+      "loss": 1.1855,
+      "step": 13892
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.686979671177482e-06,
+      "loss": 0.8999,
+      "step": 13893
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.686308143292151e-06,
+      "loss": 0.8391,
+      "step": 13894
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6856366408593012e-06,
+      "loss": 0.8947,
+      "step": 13895
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.684965163891942e-06,
+      "loss": 0.894,
+      "step": 13896
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6842937124030866e-06,
+      "loss": 1.0126,
+      "step": 13897
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6836222864057414e-06,
+      "loss": 0.9317,
+      "step": 13898
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6829508859129166e-06,
+      "loss": 0.7629,
+      "step": 13899
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6822795109376188e-06,
+      "loss": 1.0178,
+      "step": 13900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6816081614928574e-06,
+      "loss": 0.8496,
+      "step": 13901
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6809368375916384e-06,
+      "loss": 0.9272,
+      "step": 13902
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6802655392469687e-06,
+      "loss": 0.9975,
+      "step": 13903
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6795942664718554e-06,
+      "loss": 0.8411,
+      "step": 13904
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.678923019279303e-06,
+      "loss": 0.9107,
+      "step": 13905
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.678251797682317e-06,
+      "loss": 0.9631,
+      "step": 13906
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.677580601693904e-06,
+      "loss": 0.8805,
+      "step": 13907
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6769094313270647e-06,
+      "loss": 0.9275,
+      "step": 13908
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.676238286594807e-06,
+      "loss": 0.896,
+      "step": 13909
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.675567167510132e-06,
+      "loss": 0.8997,
+      "step": 13910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6748960740860417e-06,
+      "loss": 1.0715,
+      "step": 13911
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6742250063355404e-06,
+      "loss": 0.9272,
+      "step": 13912
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.673553964271629e-06,
+      "loss": 0.8611,
+      "step": 13913
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.672882947907308e-06,
+      "loss": 1.1265,
+      "step": 13914
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6722119572555793e-06,
+      "loss": 0.8957,
+      "step": 13915
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.671540992329442e-06,
+      "loss": 0.8045,
+      "step": 13916
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.670870053141897e-06,
+      "loss": 0.865,
+      "step": 13917
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.670199139705944e-06,
+      "loss": 0.9987,
+      "step": 13918
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6695282520345803e-06,
+      "loss": 0.9974,
+      "step": 13919
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.668857390140803e-06,
+      "loss": 0.8698,
+      "step": 13920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6681865540376138e-06,
+      "loss": 0.9774,
+      "step": 13921
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.667515743738008e-06,
+      "loss": 0.8627,
+      "step": 13922
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6668449592549827e-06,
+      "loss": 0.8761,
+      "step": 13923
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6661742006015333e-06,
+      "loss": 0.9141,
+      "step": 13924
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6655034677906564e-06,
+      "loss": 1.0195,
+      "step": 13925
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.664832760835347e-06,
+      "loss": 0.8805,
+      "step": 13926
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.664162079748601e-06,
+      "loss": 0.9488,
+      "step": 13927
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6634914245434106e-06,
+      "loss": 0.9302,
+      "step": 13928
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.662820795232771e-06,
+      "loss": 0.7789,
+      "step": 13929
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6621501918296752e-06,
+      "loss": 0.9141,
+      "step": 13930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6614796143471153e-06,
+      "loss": 0.7756,
+      "step": 13931
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.660809062798085e-06,
+      "loss": 0.8238,
+      "step": 13932
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.660138537195575e-06,
+      "loss": 0.8437,
+      "step": 13933
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.659468037552577e-06,
+      "loss": 0.9612,
+      "step": 13934
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6587975638820817e-06,
+      "loss": 1.055,
+      "step": 13935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.658127116197079e-06,
+      "loss": 0.9198,
+      "step": 13936
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6574566945105592e-06,
+      "loss": 1.0873,
+      "step": 13937
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.656786298835512e-06,
+      "loss": 0.8918,
+      "step": 13938
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6561159291849248e-06,
+      "loss": 0.8835,
+      "step": 13939
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6554455855717873e-06,
+      "loss": 0.8879,
+      "step": 13940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6547752680090863e-06,
+      "loss": 0.7554,
+      "step": 13941
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.654104976509809e-06,
+      "loss": 0.9351,
+      "step": 13942
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6534347110869423e-06,
+      "loss": 0.851,
+      "step": 13943
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6527644717534735e-06,
+      "loss": 0.9757,
+      "step": 13944
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.652094258522387e-06,
+      "loss": 0.6967,
+      "step": 13945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6514240714066686e-06,
+      "loss": 1.0053,
+      "step": 13946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.650753910419303e-06,
+      "loss": 0.9377,
+      "step": 13947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.650083775573274e-06,
+      "loss": 1.0961,
+      "step": 13948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.649413666881566e-06,
+      "loss": 0.8314,
+      "step": 13949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6487435843571616e-06,
+      "loss": 0.8352,
+      "step": 13950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6480735280130443e-06,
+      "loss": 1.0459,
+      "step": 13951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.647403497862195e-06,
+      "loss": 1.0475,
+      "step": 13952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6467334939175967e-06,
+      "loss": 0.9412,
+      "step": 13953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6460635161922277e-06,
+      "loss": 1.0042,
+      "step": 13954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6453935646990735e-06,
+      "loss": 0.9474,
+      "step": 13955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.644723639451111e-06,
+      "loss": 1.0264,
+      "step": 13956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6440537404613207e-06,
+      "loss": 0.9329,
+      "step": 13957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.643383867742681e-06,
+      "loss": 0.8006,
+      "step": 13958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6427140213081724e-06,
+      "loss": 0.9176,
+      "step": 13959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6420442011707713e-06,
+      "loss": 0.973,
+      "step": 13960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6413744073434553e-06,
+      "loss": 0.9995,
+      "step": 13961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6407046398392023e-06,
+      "loss": 0.8746,
+      "step": 13962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6400348986709886e-06,
+      "loss": 0.8978,
+      "step": 13963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.63936518385179e-06,
+      "loss": 0.9985,
+      "step": 13964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.638695495394583e-06,
+      "loss": 0.9432,
+      "step": 13965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6380258333123397e-06,
+      "loss": 0.8603,
+      "step": 13966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.637356197618039e-06,
+      "loss": 1.0838,
+      "step": 13967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.636686588324653e-06,
+      "loss": 0.7047,
+      "step": 13968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6360170054451545e-06,
+      "loss": 1.1428,
+      "step": 13969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.635347448992517e-06,
+      "loss": 1.0619,
+      "step": 13970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6346779189797133e-06,
+      "loss": 1.0916,
+      "step": 13971
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6340084154197154e-06,
+      "loss": 0.9037,
+      "step": 13972
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.633338938325495e-06,
+      "loss": 0.9031,
+      "step": 13973
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6326694877100223e-06,
+      "loss": 0.9475,
+      "step": 13974
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6320000635862675e-06,
+      "loss": 0.8116,
+      "step": 13975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.631330665967202e-06,
+      "loss": 0.8459,
+      "step": 13976
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6306612948657937e-06,
+      "loss": 0.8576,
+      "step": 13977
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6299919502950108e-06,
+      "loss": 1.102,
+      "step": 13978
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6293226322678243e-06,
+      "loss": 0.9241,
+      "step": 13979
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6286533407972014e-06,
+      "loss": 0.9974,
+      "step": 13980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6279840758961094e-06,
+      "loss": 0.925,
+      "step": 13981
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.627314837577514e-06,
+      "loss": 0.9756,
+      "step": 13982
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6266456258543826e-06,
+      "loss": 0.9972,
+      "step": 13983
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.62597644073968e-06,
+      "loss": 1.0687,
+      "step": 13984
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.625307282246373e-06,
+      "loss": 0.876,
+      "step": 13985
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6246381503874258e-06,
+      "loss": 0.8995,
+      "step": 13986
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.623969045175802e-06,
+      "loss": 0.9887,
+      "step": 13987
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6232999666244655e-06,
+      "loss": 0.9195,
+      "step": 13988
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.622630914746378e-06,
+      "loss": 0.8799,
+      "step": 13989
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6219618895545062e-06,
+      "loss": 0.8991,
+      "step": 13990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6212928910618106e-06,
+      "loss": 0.8193,
+      "step": 13991
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.620623919281252e-06,
+      "loss": 0.8676,
+      "step": 13992
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6199549742257923e-06,
+      "loss": 1.0745,
+      "step": 13993
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.619286055908392e-06,
+      "loss": 0.8946,
+      "step": 13994
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.618617164342011e-06,
+      "loss": 0.9753,
+      "step": 13995
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.617948299539609e-06,
+      "loss": 0.9927,
+      "step": 13996
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6172794615141448e-06,
+      "loss": 0.7773,
+      "step": 13997
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6166106502785782e-06,
+      "loss": 0.9117,
+      "step": 13998
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6159418658458663e-06,
+      "loss": 0.9247,
+      "step": 13999
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.615273108228966e-06,
+      "loss": 0.8544,
+      "step": 14000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6146043774408347e-06,
+      "loss": 1.0011,
+      "step": 14001
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.613935673494431e-06,
+      "loss": 0.8355,
+      "step": 14002
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6132669964027077e-06,
+      "loss": 0.876,
+      "step": 14003
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.612598346178623e-06,
+      "loss": 0.7853,
+      "step": 14004
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.61192972283513e-06,
+      "loss": 0.7143,
+      "step": 14005
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6112611263851843e-06,
+      "loss": 1.0162,
+      "step": 14006
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.610592556841739e-06,
+      "loss": 0.8531,
+      "step": 14007
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6099240142177476e-06,
+      "loss": 1.0242,
+      "step": 14008
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6092554985261634e-06,
+      "loss": 1.0113,
+      "step": 14009
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.608587009779938e-06,
+      "loss": 0.9555,
+      "step": 14010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.607918547992024e-06,
+      "loss": 0.7911,
+      "step": 14011
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.607250113175372e-06,
+      "loss": 0.8839,
+      "step": 14012
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.606581705342932e-06,
+      "loss": 0.9987,
+      "step": 14013
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6059133245076566e-06,
+      "loss": 0.8541,
+      "step": 14014
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6052449706824943e-06,
+      "loss": 0.807,
+      "step": 14015
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.604576643880395e-06,
+      "loss": 1.1365,
+      "step": 14016
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.603908344114306e-06,
+      "loss": 0.959,
+      "step": 14017
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.603240071397176e-06,
+      "loss": 1.0201,
+      "step": 14018
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6025718257419532e-06,
+      "loss": 0.9946,
+      "step": 14019
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.601903607161584e-06,
+      "loss": 0.8811,
+      "step": 14020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6012354156690156e-06,
+      "loss": 1.0926,
+      "step": 14021
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.600567251277194e-06,
+      "loss": 0.9687,
+      "step": 14022
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5998991139990636e-06,
+      "loss": 0.857,
+      "step": 14023
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5992310038475697e-06,
+      "loss": 0.9574,
+      "step": 14024
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.598562920835659e-06,
+      "loss": 0.9587,
+      "step": 14025
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.597894864976273e-06,
+      "loss": 1.0042,
+      "step": 14026
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.597226836282356e-06,
+      "loss": 1.1003,
+      "step": 14027
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.596558834766851e-06,
+      "loss": 0.9481,
+      "step": 14028
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5958908604427e-06,
+      "loss": 0.991,
+      "step": 14029
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5952229133228457e-06,
+      "loss": 0.9053,
+      "step": 14030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5945549934202283e-06,
+      "loss": 0.9293,
+      "step": 14031
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5938871007477893e-06,
+      "loss": 0.7957,
+      "step": 14032
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.593219235318468e-06,
+      "loss": 0.8562,
+      "step": 14033
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.592551397145206e-06,
+      "loss": 0.9701,
+      "step": 14034
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5918835862409396e-06,
+      "loss": 1.0269,
+      "step": 14035
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5912158026186093e-06,
+      "loss": 0.9557,
+      "step": 14036
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.590548046291155e-06,
+      "loss": 0.943,
+      "step": 14037
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5898803172715124e-06,
+      "loss": 1.0038,
+      "step": 14038
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.589212615572618e-06,
+      "loss": 0.864,
+      "step": 14039
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5885449412074103e-06,
+      "loss": 0.7435,
+      "step": 14040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.587877294188824e-06,
+      "loss": 1.0482,
+      "step": 14041
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.587209674529794e-06,
+      "loss": 0.7984,
+      "step": 14042
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5865420822432563e-06,
+      "loss": 1.0041,
+      "step": 14043
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5858745173421454e-06,
+      "loss": 0.9737,
+      "step": 14044
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5852069798393953e-06,
+      "loss": 0.9816,
+      "step": 14045
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5845394697479386e-06,
+      "loss": 1.0123,
+      "step": 14046
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5838719870807064e-06,
+      "loss": 0.9285,
+      "step": 14047
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5832045318506352e-06,
+      "loss": 0.9724,
+      "step": 14048
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.582537104070655e-06,
+      "loss": 1.1452,
+      "step": 14049
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5818697037536974e-06,
+      "loss": 0.9487,
+      "step": 14050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.581202330912692e-06,
+      "loss": 0.8553,
+      "step": 14051
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.58053498556057e-06,
+      "loss": 0.8727,
+      "step": 14052
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.57986766771026e-06,
+      "loss": 0.8152,
+      "step": 14053
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5792003773746918e-06,
+      "loss": 0.9462,
+      "step": 14054
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5785331145667944e-06,
+      "loss": 0.9668,
+      "step": 14055
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5778658792994957e-06,
+      "loss": 0.907,
+      "step": 14056
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.577198671585723e-06,
+      "loss": 0.9821,
+      "step": 14057
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.576531491438403e-06,
+      "loss": 0.9187,
+      "step": 14058
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.575864338870462e-06,
+      "loss": 0.9417,
+      "step": 14059
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5751972138948264e-06,
+      "loss": 0.7247,
+      "step": 14060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5745301165244216e-06,
+      "loss": 1.0757,
+      "step": 14061
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5738630467721736e-06,
+      "loss": 0.9175,
+      "step": 14062
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5731960046510052e-06,
+      "loss": 0.8476,
+      "step": 14063
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5725289901738396e-06,
+      "loss": 1.0647,
+      "step": 14064
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5718620033536023e-06,
+      "loss": 0.9076,
+      "step": 14065
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.571195044203214e-06,
+      "loss": 1.0763,
+      "step": 14066
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.570528112735597e-06,
+      "loss": 0.8362,
+      "step": 14067
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5698612089636743e-06,
+      "loss": 0.9126,
+      "step": 14068
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5691943329003666e-06,
+      "loss": 1.0283,
+      "step": 14069
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5685274845585936e-06,
+      "loss": 0.8243,
+      "step": 14070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.567860663951274e-06,
+      "loss": 0.8843,
+      "step": 14071
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.567193871091331e-06,
+      "loss": 1.0107,
+      "step": 14072
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5665271059916816e-06,
+      "loss": 1.0485,
+      "step": 14073
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5658603686652447e-06,
+      "loss": 1.073,
+      "step": 14074
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.565193659124938e-06,
+      "loss": 0.8744,
+      "step": 14075
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5645269773836787e-06,
+      "loss": 0.7436,
+      "step": 14076
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5638603234543833e-06,
+      "loss": 1.0299,
+      "step": 14077
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5631936973499685e-06,
+      "loss": 0.9466,
+      "step": 14078
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5625270990833505e-06,
+      "loss": 0.8053,
+      "step": 14079
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.561860528667443e-06,
+      "loss": 1.0038,
+      "step": 14080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.561193986115161e-06,
+      "loss": 0.8704,
+      "step": 14081
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5605274714394185e-06,
+      "loss": 0.8754,
+      "step": 14082
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5598609846531305e-06,
+      "loss": 1.0723,
+      "step": 14083
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5591945257692106e-06,
+      "loss": 0.8403,
+      "step": 14084
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.558528094800569e-06,
+      "loss": 0.8673,
+      "step": 14085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.557861691760119e-06,
+      "loss": 0.9404,
+      "step": 14086
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5571953166607708e-06,
+      "loss": 0.8148,
+      "step": 14087
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.556528969515437e-06,
+      "loss": 0.9535,
+      "step": 14088
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5558626503370263e-06,
+      "loss": 1.0327,
+      "step": 14089
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5551963591384495e-06,
+      "loss": 1.0284,
+      "step": 14090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5545300959326155e-06,
+      "loss": 1.0223,
+      "step": 14091
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5538638607324334e-06,
+      "loss": 1.0177,
+      "step": 14092
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5531976535508107e-06,
+      "loss": 0.9453,
+      "step": 14093
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.552531474400654e-06,
+      "loss": 0.8577,
+      "step": 14094
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.551865323294873e-06,
+      "loss": 0.981,
+      "step": 14095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5511992002463723e-06,
+      "loss": 0.8551,
+      "step": 14096
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.550533105268059e-06,
+      "loss": 1.0176,
+      "step": 14097
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5498670383728374e-06,
+      "loss": 0.9132,
+      "step": 14098
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.549200999573613e-06,
+      "loss": 1.0239,
+      "step": 14099
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.548534988883291e-06,
+      "loss": 0.9508,
+      "step": 14100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5478690063147735e-06,
+      "loss": 1.0095,
+      "step": 14101
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5472030518809654e-06,
+      "loss": 1.0767,
+      "step": 14102
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5465371255947676e-06,
+      "loss": 1.0979,
+      "step": 14103
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5458712274690844e-06,
+      "loss": 0.9869,
+      "step": 14104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5452053575168157e-06,
+      "loss": 0.8453,
+      "step": 14105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5445395157508612e-06,
+      "loss": 0.8787,
+      "step": 14106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5438737021841257e-06,
+      "loss": 0.9906,
+      "step": 14107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5432079168295063e-06,
+      "loss": 0.9601,
+      "step": 14108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.542542159699904e-06,
+      "loss": 1.0669,
+      "step": 14109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.541876430808216e-06,
+      "loss": 1.0325,
+      "step": 14110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5412107301673413e-06,
+      "loss": 0.8886,
+      "step": 14111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5405450577901785e-06,
+      "loss": 1.1075,
+      "step": 14112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.539879413689624e-06,
+      "loss": 0.811,
+      "step": 14113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5392137978785747e-06,
+      "loss": 0.8561,
+      "step": 14114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.538548210369926e-06,
+      "loss": 0.984,
+      "step": 14115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5378826511765745e-06,
+      "loss": 0.689,
+      "step": 14116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5372171203114143e-06,
+      "loss": 0.9431,
+      "step": 14117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5365516177873416e-06,
+      "loss": 1.0199,
+      "step": 14118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5358861436172487e-06,
+      "loss": 0.8704,
+      "step": 14119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.53522069781403e-06,
+      "loss": 1.0845,
+      "step": 14120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.534555280390578e-06,
+      "loss": 0.9832,
+      "step": 14121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5338898913597853e-06,
+      "loss": 0.8234,
+      "step": 14122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.533224530734543e-06,
+      "loss": 0.9915,
+      "step": 14123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.532559198527743e-06,
+      "loss": 0.8288,
+      "step": 14124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.531893894752276e-06,
+      "loss": 0.8995,
+      "step": 14125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.531228619421031e-06,
+      "loss": 0.9047,
+      "step": 14126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5305633725468985e-06,
+      "loss": 0.9135,
+      "step": 14127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.529898154142768e-06,
+      "loss": 0.927,
+      "step": 14128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.529232964221524e-06,
+      "loss": 0.8554,
+      "step": 14129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.528567802796061e-06,
+      "loss": 0.9337,
+      "step": 14130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5279026698792617e-06,
+      "loss": 0.9309,
+      "step": 14131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.527237565484015e-06,
+      "loss": 0.9295,
+      "step": 14132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.526572489623207e-06,
+      "loss": 1.0342,
+      "step": 14133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5259074423097223e-06,
+      "loss": 0.9858,
+      "step": 14134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5252424235564464e-06,
+      "loss": 0.9391,
+      "step": 14135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5245774333762638e-06,
+      "loss": 0.9285,
+      "step": 14136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5239124717820595e-06,
+      "loss": 0.9705,
+      "step": 14137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5232475387867156e-06,
+      "loss": 1.0642,
+      "step": 14138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5225826344031155e-06,
+      "loss": 0.9484,
+      "step": 14139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.52191775864414e-06,
+      "loss": 1.0389,
+      "step": 14140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5212529115226744e-06,
+      "loss": 0.835,
+      "step": 14141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.520588093051599e-06,
+      "loss": 0.8486,
+      "step": 14142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.519923303243793e-06,
+      "loss": 0.9398,
+      "step": 14143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.519258542112137e-06,
+      "loss": 1.129,
+      "step": 14144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5185938096695117e-06,
+      "loss": 0.9987,
+      "step": 14145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5179291059287947e-06,
+      "loss": 1.0676,
+      "step": 14146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5172644309028652e-06,
+      "loss": 0.9339,
+      "step": 14147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5165997846046007e-06,
+      "loss": 1.0585,
+      "step": 14148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.51593516704688e-06,
+      "loss": 1.1192,
+      "step": 14149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5152705782425778e-06,
+      "loss": 0.8128,
+      "step": 14150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5146060182045714e-06,
+      "loss": 1.1213,
+      "step": 14151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5139414869457354e-06,
+      "loss": 0.9755,
+      "step": 14152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.513276984478947e-06,
+      "loss": 0.8869,
+      "step": 14153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5126125108170805e-06,
+      "loss": 0.8293,
+      "step": 14154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5119480659730087e-06,
+      "loss": 1.0882,
+      "step": 14155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5112836499596058e-06,
+      "loss": 1.0332,
+      "step": 14156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.510619262789744e-06,
+      "loss": 0.9727,
+      "step": 14157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.509954904476297e-06,
+      "loss": 1.0219,
+      "step": 14158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.509290575032135e-06,
+      "loss": 1.096,
+      "step": 14159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.50862627447013e-06,
+      "loss": 1.006,
+      "step": 14160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5079620028031525e-06,
+      "loss": 0.8613,
+      "step": 14161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5072977600440734e-06,
+      "loss": 0.997,
+      "step": 14162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.506633546205761e-06,
+      "loss": 1.0664,
+      "step": 14163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.505969361301083e-06,
+      "loss": 0.7966,
+      "step": 14164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5053052053429114e-06,
+      "loss": 1.0462,
+      "step": 14165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.504641078344113e-06,
+      "loss": 0.9699,
+      "step": 14166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5039769803175545e-06,
+      "loss": 0.9504,
+      "step": 14167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.503312911276102e-06,
+      "loss": 0.9595,
+      "step": 14168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5026488712326233e-06,
+      "loss": 0.9382,
+      "step": 14169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.501984860199983e-06,
+      "loss": 0.9864,
+      "step": 14170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5013208781910457e-06,
+      "loss": 0.7792,
+      "step": 14171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.500656925218676e-06,
+      "loss": 0.9953,
+      "step": 14172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.499993001295739e-06,
+      "loss": 0.9803,
+      "step": 14173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.499329106435097e-06,
+      "loss": 0.8642,
+      "step": 14174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4986652406496113e-06,
+      "loss": 0.879,
+      "step": 14175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.498001403952148e-06,
+      "loss": 1.0627,
+      "step": 14176
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4973375963555666e-06,
+      "loss": 1.0327,
+      "step": 14177
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4966738178727287e-06,
+      "loss": 0.9977,
+      "step": 14178
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.496010068516494e-06,
+      "loss": 1.0287,
+      "step": 14179
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.495346348299724e-06,
+      "loss": 1.075,
+      "step": 14180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4946826572352764e-06,
+      "loss": 0.9292,
+      "step": 14181
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4940189953360115e-06,
+      "loss": 1.0043,
+      "step": 14182
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4933553626147866e-06,
+      "loss": 1.0243,
+      "step": 14183
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4926917590844597e-06,
+      "loss": 0.9458,
+      "step": 14184
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4920281847578878e-06,
+      "loss": 0.9571,
+      "step": 14185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4913646396479284e-06,
+      "loss": 0.8842,
+      "step": 14186
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4907011237674354e-06,
+      "loss": 0.9134,
+      "step": 14187
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4900376371292666e-06,
+      "loss": 1.0559,
+      "step": 14188
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4893741797462767e-06,
+      "loss": 0.9631,
+      "step": 14189
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.488710751631319e-06,
+      "loss": 0.987,
+      "step": 14190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.488047352797248e-06,
+      "loss": 0.7063,
+      "step": 14191
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.487383983256916e-06,
+      "loss": 1.0278,
+      "step": 14192
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.486720643023177e-06,
+      "loss": 0.9714,
+      "step": 14193
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.486057332108882e-06,
+      "loss": 0.9297,
+      "step": 14194
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4853940505268824e-06,
+      "loss": 0.9285,
+      "step": 14195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4847307982900296e-06,
+      "loss": 0.9183,
+      "step": 14196
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.484067575411174e-06,
+      "loss": 0.9308,
+      "step": 14197
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4834043819031625e-06,
+      "loss": 0.9047,
+      "step": 14198
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.48274121777885e-06,
+      "loss": 1.0983,
+      "step": 14199
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4820780830510823e-06,
+      "loss": 0.9888,
+      "step": 14200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.481414977732707e-06,
+      "loss": 0.8829,
+      "step": 14201
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.480751901836572e-06,
+      "loss": 1.0319,
+      "step": 14202
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.480088855375525e-06,
+      "loss": 1.1279,
+      "step": 14203
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4794258383624115e-06,
+      "loss": 1.0793,
+      "step": 14204
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4787628508100767e-06,
+      "loss": 0.8849,
+      "step": 14205
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.478099892731367e-06,
+      "loss": 0.9,
+      "step": 14206
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4774369641391266e-06,
+      "loss": 0.8306,
+      "step": 14207
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4767740650461995e-06,
+      "loss": 0.9677,
+      "step": 14208
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.476111195465429e-06,
+      "loss": 0.8173,
+      "step": 14209
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4754483554096586e-06,
+      "loss": 0.8214,
+      "step": 14210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4747855448917307e-06,
+      "loss": 1.0084,
+      "step": 14211
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.474122763924487e-06,
+      "loss": 0.9133,
+      "step": 14212
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4734600125207686e-06,
+      "loss": 1.0463,
+      "step": 14213
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.472797290693416e-06,
+      "loss": 0.9242,
+      "step": 14214
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.47213459845527e-06,
+      "loss": 1.0394,
+      "step": 14215
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4714719358191695e-06,
+      "loss": 0.8004,
+      "step": 14216
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.470809302797953e-06,
+      "loss": 0.8896,
+      "step": 14217
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4701466994044596e-06,
+      "loss": 0.9197,
+      "step": 14218
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4694841256515266e-06,
+      "loss": 0.8149,
+      "step": 14219
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4688215815519923e-06,
+      "loss": 0.9799,
+      "step": 14220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4681590671186917e-06,
+      "loss": 0.8399,
+      "step": 14221
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.46749658236446e-06,
+      "loss": 0.8365,
+      "step": 14222
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4668341273021365e-06,
+      "loss": 1.0738,
+      "step": 14223
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.466171701944554e-06,
+      "loss": 0.7344,
+      "step": 14224
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4655093063045462e-06,
+      "loss": 0.8561,
+      "step": 14225
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4648469403949485e-06,
+      "loss": 0.6663,
+      "step": 14226
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4641846042285933e-06,
+      "loss": 0.9261,
+      "step": 14227
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.463522297818312e-06,
+      "loss": 1.1805,
+      "step": 14228
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.462860021176938e-06,
+      "loss": 0.9192,
+      "step": 14229
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4621977743173026e-06,
+      "loss": 0.9442,
+      "step": 14230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4615355572522358e-06,
+      "loss": 0.9823,
+      "step": 14231
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4608733699945696e-06,
+      "loss": 0.7551,
+      "step": 14232
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4602112125571295e-06,
+      "loss": 0.977,
+      "step": 14233
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4595490849527514e-06,
+      "loss": 1.0261,
+      "step": 14234
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4588869871942593e-06,
+      "loss": 1.0697,
+      "step": 14235
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.458224919294483e-06,
+      "loss": 0.9882,
+      "step": 14236
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4575628812662484e-06,
+      "loss": 0.9557,
+      "step": 14237
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.456900873122383e-06,
+      "loss": 0.8492,
+      "step": 14238
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.456238894875714e-06,
+      "loss": 0.9556,
+      "step": 14239
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4555769465390653e-06,
+      "loss": 1.0106,
+      "step": 14240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4549150281252635e-06,
+      "loss": 0.9491,
+      "step": 14241
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4542531396471324e-06,
+      "loss": 1.0753,
+      "step": 14242
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4535912811174965e-06,
+      "loss": 1.1014,
+      "step": 14243
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.452929452549178e-06,
+      "loss": 0.9725,
+      "step": 14244
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4522676539549994e-06,
+      "loss": 0.9817,
+      "step": 14245
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4516058853477856e-06,
+      "loss": 0.8448,
+      "step": 14246
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.450944146740356e-06,
+      "loss": 0.9724,
+      "step": 14247
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4502824381455324e-06,
+      "loss": 0.8897,
+      "step": 14248
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4496207595761357e-06,
+      "loss": 1.1333,
+      "step": 14249
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4489591110449837e-06,
+      "loss": 0.7996,
+      "step": 14250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.448297492564898e-06,
+      "loss": 1.0354,
+      "step": 14251
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4476359041486963e-06,
+      "loss": 0.9938,
+      "step": 14252
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4469743458091966e-06,
+      "loss": 1.1448,
+      "step": 14253
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.446312817559217e-06,
+      "loss": 0.9416,
+      "step": 14254
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4456513194115737e-06,
+      "loss": 0.9518,
+      "step": 14255
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4449898513790834e-06,
+      "loss": 1.0019,
+      "step": 14256
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4443284134745606e-06,
+      "loss": 1.0455,
+      "step": 14257
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4436670057108235e-06,
+      "loss": 1.042,
+      "step": 14258
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.443005628100685e-06,
+      "loss": 0.9272,
+      "step": 14259
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4423442806569598e-06,
+      "loss": 0.9208,
+      "step": 14260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.441682963392461e-06,
+      "loss": 0.8873,
+      "step": 14261
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4410216763200006e-06,
+      "loss": 0.9559,
+      "step": 14262
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.440360419452392e-06,
+      "loss": 0.9198,
+      "step": 14263
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.439699192802446e-06,
+      "loss": 0.9241,
+      "step": 14264
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4390379963829747e-06,
+      "loss": 0.86,
+      "step": 14265
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.438376830206788e-06,
+      "loss": 0.8915,
+      "step": 14266
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4377156942866958e-06,
+      "loss": 0.8611,
+      "step": 14267
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4370545886355056e-06,
+      "loss": 0.9303,
+      "step": 14268
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.436393513266031e-06,
+      "loss": 0.774,
+      "step": 14269
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4357324681910763e-06,
+      "loss": 0.9862,
+      "step": 14270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.435071453423451e-06,
+      "loss": 1.0438,
+      "step": 14271
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4344104689759606e-06,
+      "loss": 0.9787,
+      "step": 14272
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.433749514861413e-06,
+      "loss": 0.8383,
+      "step": 14273
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4330885910926125e-06,
+      "loss": 1.0106,
+      "step": 14274
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4324276976823654e-06,
+      "loss": 0.8985,
+      "step": 14275
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.431766834643476e-06,
+      "loss": 0.7892,
+      "step": 14276
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.431106001988749e-06,
+      "loss": 0.945,
+      "step": 14277
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.430445199730987e-06,
+      "loss": 0.9993,
+      "step": 14278
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.429784427882993e-06,
+      "loss": 0.9702,
+      "step": 14279
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.429123686457568e-06,
+      "loss": 0.8463,
+      "step": 14280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4284629754675176e-06,
+      "loss": 0.83,
+      "step": 14281
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4278022949256394e-06,
+      "loss": 1.0526,
+      "step": 14282
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.427141644844736e-06,
+      "loss": 0.8792,
+      "step": 14283
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4264810252376057e-06,
+      "loss": 1.0789,
+      "step": 14284
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.425820436117049e-06,
+      "loss": 1.0347,
+      "step": 14285
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4251598774958644e-06,
+      "loss": 0.943,
+      "step": 14286
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4244993493868493e-06,
+      "loss": 0.9617,
+      "step": 14287
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.423838851802803e-06,
+      "loss": 0.8266,
+      "step": 14288
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4231783847565203e-06,
+      "loss": 0.8882,
+      "step": 14289
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4225179482607995e-06,
+      "loss": 0.9015,
+      "step": 14290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.421857542328433e-06,
+      "loss": 0.9048,
+      "step": 14291
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.421197166972221e-06,
+      "loss": 0.9751,
+      "step": 14292
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.420536822204956e-06,
+      "loss": 0.9119,
+      "step": 14293
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4198765080394315e-06,
+      "loss": 1.1615,
+      "step": 14294
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.419216224488441e-06,
+      "loss": 0.9746,
+      "step": 14295
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4185559715647777e-06,
+      "loss": 0.8144,
+      "step": 14296
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4178957492812336e-06,
+      "loss": 1.0181,
+      "step": 14297
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.417235557650601e-06,
+      "loss": 1.0447,
+      "step": 14298
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4165753966856704e-06,
+      "loss": 0.87,
+      "step": 14299
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.415915266399232e-06,
+      "loss": 0.8789,
+      "step": 14300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4152551668040756e-06,
+      "loss": 1.0342,
+      "step": 14301
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4145950979129916e-06,
+      "loss": 0.8762,
+      "step": 14302
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4139350597387667e-06,
+      "loss": 1.0932,
+      "step": 14303
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4132750522941903e-06,
+      "loss": 0.8024,
+      "step": 14304
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4126150755920514e-06,
+      "loss": 0.9859,
+      "step": 14305
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.411955129645135e-06,
+      "loss": 1.0665,
+      "step": 14306
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.411295214466227e-06,
+      "loss": 0.9358,
+      "step": 14307
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4106353300681138e-06,
+      "loss": 0.9511,
+      "step": 14308
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.409975476463581e-06,
+      "loss": 0.9259,
+      "step": 14309
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4093156536654126e-06,
+      "loss": 1.011,
+      "step": 14310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.408655861686393e-06,
+      "loss": 1.0767,
+      "step": 14311
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4079961005393047e-06,
+      "loss": 1.016,
+      "step": 14312
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4073363702369303e-06,
+      "loss": 1.0116,
+      "step": 14313
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.406676670792053e-06,
+      "loss": 0.8381,
+      "step": 14314
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.406017002217452e-06,
+      "loss": 0.974,
+      "step": 14315
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.405357364525912e-06,
+      "loss": 0.8893,
+      "step": 14316
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.404697757730211e-06,
+      "loss": 0.9039,
+      "step": 14317
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.404038181843129e-06,
+      "loss": 1.1444,
+      "step": 14318
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.403378636877445e-06,
+      "loss": 0.6809,
+      "step": 14319
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4027191228459383e-06,
+      "loss": 1.1755,
+      "step": 14320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.402059639761386e-06,
+      "loss": 0.8203,
+      "step": 14321
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.401400187636566e-06,
+      "loss": 0.8164,
+      "step": 14322
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4007407664842544e-06,
+      "loss": 0.829,
+      "step": 14323
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.400081376317228e-06,
+      "loss": 0.9333,
+      "step": 14324
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.399422017148262e-06,
+      "loss": 0.9905,
+      "step": 14325
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3987626889901287e-06,
+      "loss": 0.863,
+      "step": 14326
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3981033918556073e-06,
+      "loss": 1.0555,
+      "step": 14327
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3974441257574695e-06,
+      "loss": 1.0253,
+      "step": 14328
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3967848907084887e-06,
+      "loss": 0.9735,
+      "step": 14329
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3961256867214364e-06,
+      "loss": 0.9474,
+      "step": 14330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3954665138090853e-06,
+      "loss": 0.888,
+      "step": 14331
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.394807371984207e-06,
+      "loss": 0.9709,
+      "step": 14332
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.394148261259571e-06,
+      "loss": 0.7615,
+      "step": 14333
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3934891816479483e-06,
+      "loss": 0.9921,
+      "step": 14334
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3928301331621084e-06,
+      "loss": 0.9257,
+      "step": 14335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3921711158148197e-06,
+      "loss": 1.0026,
+      "step": 14336
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3915121296188515e-06,
+      "loss": 0.7431,
+      "step": 14337
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.39085317458697e-06,
+      "loss": 0.9204,
+      "step": 14338
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.390194250731944e-06,
+      "loss": 0.9129,
+      "step": 14339
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.389535358066539e-06,
+      "loss": 1.0499,
+      "step": 14340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3888764966035213e-06,
+      "loss": 0.94,
+      "step": 14341
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.388217666355656e-06,
+      "loss": 1.0352,
+      "step": 14342
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3875588673357075e-06,
+      "loss": 0.8557,
+      "step": 14343
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3869000995564407e-06,
+      "loss": 0.8437,
+      "step": 14344
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3862413630306186e-06,
+      "loss": 0.9873,
+      "step": 14345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3855826577710036e-06,
+      "loss": 1.0096,
+      "step": 14346
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3849239837903587e-06,
+      "loss": 1.142,
+      "step": 14347
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.384265341101445e-06,
+      "loss": 0.9477,
+      "step": 14348
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.383606729717024e-06,
+      "loss": 0.9185,
+      "step": 14349
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.382948149649854e-06,
+      "loss": 0.8489,
+      "step": 14350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.382289600912698e-06,
+      "loss": 0.9508,
+      "step": 14351
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3816310835183153e-06,
+      "loss": 1.0056,
+      "step": 14352
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.380972597479463e-06,
+      "loss": 0.9273,
+      "step": 14353
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3803141428088986e-06,
+      "loss": 0.8652,
+      "step": 14354
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3796557195193807e-06,
+      "loss": 1.0875,
+      "step": 14355
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3789973276236656e-06,
+      "loss": 0.8977,
+      "step": 14356
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.37833896713451e-06,
+      "loss": 0.9729,
+      "step": 14357
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3776806380646686e-06,
+      "loss": 0.7502,
+      "step": 14358
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.377022340426897e-06,
+      "loss": 0.9539,
+      "step": 14359
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.376364074233949e-06,
+      "loss": 0.8483,
+      "step": 14360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3757058394985787e-06,
+      "loss": 0.9595,
+      "step": 14361
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3750476362335396e-06,
+      "loss": 0.8588,
+      "step": 14362
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.374389464451583e-06,
+      "loss": 1.0596,
+      "step": 14363
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3737313241654625e-06,
+      "loss": 0.8306,
+      "step": 14364
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.373073215387929e-06,
+      "loss": 0.8174,
+      "step": 14365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.372415138131733e-06,
+      "loss": 0.9975,
+      "step": 14366
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3717570924096245e-06,
+      "loss": 1.0112,
+      "step": 14367
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3710990782343533e-06,
+      "loss": 0.8422,
+      "step": 14368
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3704410956186674e-06,
+      "loss": 0.9555,
+      "step": 14369
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.369783144575316e-06,
+      "loss": 0.9112,
+      "step": 14370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.369125225117046e-06,
+      "loss": 0.875,
+      "step": 14371
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3684673372566044e-06,
+      "loss": 0.7875,
+      "step": 14372
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3678094810067387e-06,
+      "loss": 0.9146,
+      "step": 14373
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.367151656380193e-06,
+      "loss": 1.037,
+      "step": 14374
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.366493863389715e-06,
+      "loss": 0.9061,
+      "step": 14375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.365836102048048e-06,
+      "loss": 0.892,
+      "step": 14376
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.365178372367935e-06,
+      "loss": 0.9196,
+      "step": 14377
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3645206743621205e-06,
+      "loss": 0.9855,
+      "step": 14378
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3638630080433467e-06,
+      "loss": 0.8418,
+      "step": 14379
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3632053734243565e-06,
+      "loss": 0.874,
+      "step": 14380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3625477705178906e-06,
+      "loss": 0.9523,
+      "step": 14381
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.36189019933669e-06,
+      "loss": 1.0507,
+      "step": 14382
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3612326598934953e-06,
+      "loss": 0.9441,
+      "step": 14383
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3605751522010443e-06,
+      "loss": 1.0202,
+      "step": 14384
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3599176762720797e-06,
+      "loss": 0.8995,
+      "step": 14385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.359260232119338e-06,
+      "loss": 0.9118,
+      "step": 14386
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3586028197555577e-06,
+      "loss": 0.9593,
+      "step": 14387
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3579454391934744e-06,
+      "loss": 0.8934,
+      "step": 14388
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3572880904458267e-06,
+      "loss": 0.9895,
+      "step": 14389
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.356630773525349e-06,
+      "loss": 1.0797,
+      "step": 14390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3559734884447775e-06,
+      "loss": 1.037,
+      "step": 14391
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.355316235216847e-06,
+      "loss": 1.0222,
+      "step": 14392
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3546590138542913e-06,
+      "loss": 0.9281,
+      "step": 14393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3540018243698436e-06,
+      "loss": 1.1454,
+      "step": 14394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.353344666776237e-06,
+      "loss": 0.8002,
+      "step": 14395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3526875410862035e-06,
+      "loss": 0.8667,
+      "step": 14396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3520304473124767e-06,
+      "loss": 0.953,
+      "step": 14397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3513733854677855e-06,
+      "loss": 0.929,
+      "step": 14398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.35071635556486e-06,
+      "loss": 0.8434,
+      "step": 14399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3500593576164326e-06,
+      "loss": 0.8607,
+      "step": 14400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.34940239163523e-06,
+      "loss": 1.0424,
+      "step": 14401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3487454576339824e-06,
+      "loss": 1.0419,
+      "step": 14402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.348088555625416e-06,
+      "loss": 0.7521,
+      "step": 14403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.34743168562226e-06,
+      "loss": 1.0718,
+      "step": 14404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3467748476372396e-06,
+      "loss": 0.8417,
+      "step": 14405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.346118041683082e-06,
+      "loss": 0.9974,
+      "step": 14406
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3454612677725116e-06,
+      "loss": 0.9291,
+      "step": 14407
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3448045259182527e-06,
+      "loss": 0.8986,
+      "step": 14408
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.344147816133032e-06,
+      "loss": 0.8521,
+      "step": 14409
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.343491138429572e-06,
+      "loss": 0.686,
+      "step": 14410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3428344928205947e-06,
+      "loss": 0.913,
+      "step": 14411
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3421778793188243e-06,
+      "loss": 0.9312,
+      "step": 14412
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3415212979369804e-06,
+      "loss": 0.8788,
+      "step": 14413
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.340864748687786e-06,
+      "loss": 0.8418,
+      "step": 14414
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3402082315839603e-06,
+      "loss": 0.8114,
+      "step": 14415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.339551746638224e-06,
+      "loss": 0.924,
+      "step": 14416
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3388952938632957e-06,
+      "loss": 1.01,
+      "step": 14417
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.338238873271894e-06,
+      "loss": 0.7836,
+      "step": 14418
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.337582484876736e-06,
+      "loss": 0.9946,
+      "step": 14419
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.336926128690541e-06,
+      "loss": 0.7513,
+      "step": 14420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3362698047260257e-06,
+      "loss": 0.9049,
+      "step": 14421
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3356135129959055e-06,
+      "loss": 0.9479,
+      "step": 14422
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.334957253512896e-06,
+      "loss": 0.9699,
+      "step": 14423
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3343010262897125e-06,
+      "loss": 0.9343,
+      "step": 14424
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.333644831339068e-06,
+      "loss": 0.7959,
+      "step": 14425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.332988668673677e-06,
+      "loss": 0.9459,
+      "step": 14426
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.332332538306253e-06,
+      "loss": 0.7789,
+      "step": 14427
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3316764402495074e-06,
+      "loss": 0.8717,
+      "step": 14428
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3310203745161518e-06,
+      "loss": 0.9368,
+      "step": 14429
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.330364341118899e-06,
+      "loss": 0.8741,
+      "step": 14430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3297083400704565e-06,
+      "loss": 1.0439,
+      "step": 14431
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.329052371383537e-06,
+      "loss": 0.9125,
+      "step": 14432
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3283964350708484e-06,
+      "loss": 0.8985,
+      "step": 14433
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3277405311451006e-06,
+      "loss": 1.0382,
+      "step": 14434
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.327084659619001e-06,
+      "loss": 0.9539,
+      "step": 14435
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.326428820505256e-06,
+      "loss": 1.1408,
+      "step": 14436
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3257730138165726e-06,
+      "loss": 1.0032,
+      "step": 14437
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3251172395656576e-06,
+      "loss": 0.8791,
+      "step": 14438
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.324461497765217e-06,
+      "loss": 0.9767,
+      "step": 14439
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3238057884279533e-06,
+      "loss": 1.0551,
+      "step": 14440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3231501115665733e-06,
+      "loss": 1.0244,
+      "step": 14441
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3224944671937786e-06,
+      "loss": 0.798,
+      "step": 14442
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.321838855322272e-06,
+      "loss": 1.0863,
+      "step": 14443
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3211832759647583e-06,
+      "loss": 0.9118,
+      "step": 14444
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.320527729133938e-06,
+      "loss": 0.8547,
+      "step": 14445
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3198722148425123e-06,
+      "loss": 0.8412,
+      "step": 14446
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.319216733103181e-06,
+      "loss": 0.9918,
+      "step": 14447
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.318561283928644e-06,
+      "loss": 0.7685,
+      "step": 14448
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.317905867331601e-06,
+      "loss": 0.8991,
+      "step": 14449
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3172504833247504e-06,
+      "loss": 0.7853,
+      "step": 14450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.31659513192079e-06,
+      "loss": 0.7523,
+      "step": 14451
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3159398131324164e-06,
+      "loss": 0.924,
+      "step": 14452
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3152845269723276e-06,
+      "loss": 0.9909,
+      "step": 14453
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3146292734532183e-06,
+      "loss": 1.0116,
+      "step": 14454
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3139740525877844e-06,
+      "loss": 1.0066,
+      "step": 14455
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3133188643887216e-06,
+      "loss": 0.9235,
+      "step": 14456
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.312663708868723e-06,
+      "loss": 0.956,
+      "step": 14457
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3120085860404816e-06,
+      "loss": 0.8782,
+      "step": 14458
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3113534959166914e-06,
+      "loss": 1.0699,
+      "step": 14459
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.310698438510045e-06,
+      "loss": 0.9859,
+      "step": 14460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3100434138332317e-06,
+      "loss": 1.0196,
+      "step": 14461
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.309388421898944e-06,
+      "loss": 0.8607,
+      "step": 14462
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.308733462719873e-06,
+      "loss": 0.8631,
+      "step": 14463
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3080785363087077e-06,
+      "loss": 1.0092,
+      "step": 14464
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3074236426781366e-06,
+      "loss": 1.0573,
+      "step": 14465
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3067687818408475e-06,
+      "loss": 1.149,
+      "step": 14466
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.30611395380953e-06,
+      "loss": 1.064,
+      "step": 14467
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3054591585968705e-06,
+      "loss": 0.9329,
+      "step": 14468
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.304804396215555e-06,
+      "loss": 1.0659,
+      "step": 14469
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3041496666782707e-06,
+      "loss": 1.0227,
+      "step": 14470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3034949699977015e-06,
+      "loss": 0.8165,
+      "step": 14471
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.302840306186532e-06,
+      "loss": 1.0875,
+      "step": 14472
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3021856752574478e-06,
+      "loss": 0.9021,
+      "step": 14473
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.30153107722313e-06,
+      "loss": 0.8937,
+      "step": 14474
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3008765120962623e-06,
+      "loss": 0.8812,
+      "step": 14475
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.300221979889527e-06,
+      "loss": 0.943,
+      "step": 14476
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.299567480615603e-06,
+      "loss": 0.8998,
+      "step": 14477
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.298913014287176e-06,
+      "loss": 0.9105,
+      "step": 14478
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.298258580916923e-06,
+      "loss": 0.8459,
+      "step": 14479
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.297604180517524e-06,
+      "loss": 0.9834,
+      "step": 14480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2969498131016574e-06,
+      "loss": 0.9605,
+      "step": 14481
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.296295478682002e-06,
+      "loss": 0.8473,
+      "step": 14482
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.295641177271236e-06,
+      "loss": 0.8952,
+      "step": 14483
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.294986908882035e-06,
+      "loss": 1.0312,
+      "step": 14484
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2943326735270766e-06,
+      "loss": 0.9865,
+      "step": 14485
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.293678471219035e-06,
+      "loss": 0.9073,
+      "step": 14486
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.293024301970586e-06,
+      "loss": 0.7911,
+      "step": 14487
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.292370165794404e-06,
+      "loss": 1.0596,
+      "step": 14488
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.291716062703162e-06,
+      "loss": 0.956,
+      "step": 14489
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2910619927095344e-06,
+      "loss": 0.8462,
+      "step": 14490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2904079558261926e-06,
+      "loss": 0.8008,
+      "step": 14491
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2897539520658095e-06,
+      "loss": 0.9841,
+      "step": 14492
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.289099981441056e-06,
+      "loss": 1.004,
+      "step": 14493
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2884460439646015e-06,
+      "loss": 0.9063,
+      "step": 14494
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.287792139649117e-06,
+      "loss": 1.0286,
+      "step": 14495
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2871382685072717e-06,
+      "loss": 0.9924,
+      "step": 14496
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.286484430551733e-06,
+      "loss": 0.8396,
+      "step": 14497
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2858306257951704e-06,
+      "loss": 0.8193,
+      "step": 14498
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2851768542502504e-06,
+      "loss": 0.9636,
+      "step": 14499
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2845231159296404e-06,
+      "loss": 0.9536,
+      "step": 14500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2838694108460036e-06,
+      "loss": 0.8194,
+      "step": 14501
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.283215739012009e-06,
+      "loss": 1.0268,
+      "step": 14502
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2825621004403206e-06,
+      "loss": 0.9267,
+      "step": 14503
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.281908495143602e-06,
+      "loss": 1.004,
+      "step": 14504
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2812549231345163e-06,
+      "loss": 0.9918,
+      "step": 14505
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2806013844257267e-06,
+      "loss": 1.0183,
+      "step": 14506
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2799478790298955e-06,
+      "loss": 0.8906,
+      "step": 14507
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.279294406959684e-06,
+      "loss": 0.9202,
+      "step": 14508
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.278640968227753e-06,
+      "loss": 0.9695,
+      "step": 14509
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.277987562846763e-06,
+      "loss": 0.9159,
+      "step": 14510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.277334190829372e-06,
+      "loss": 1.055,
+      "step": 14511
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.27668085218824e-06,
+      "loss": 0.9789,
+      "step": 14512
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2760275469360263e-06,
+      "loss": 0.807,
+      "step": 14513
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2753742750853885e-06,
+      "loss": 0.8757,
+      "step": 14514
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2747210366489825e-06,
+      "loss": 0.8538,
+      "step": 14515
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2740678316394654e-06,
+      "loss": 0.8704,
+      "step": 14516
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.273414660069493e-06,
+      "loss": 1.0398,
+      "step": 14517
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.272761521951719e-06,
+      "loss": 0.9706,
+      "step": 14518
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.272108417298799e-06,
+      "loss": 0.9625,
+      "step": 14519
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.271455346123387e-06,
+      "loss": 0.8596,
+      "step": 14520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2708023084381346e-06,
+      "loss": 0.8695,
+      "step": 14521
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2701493042556955e-06,
+      "loss": 0.8809,
+      "step": 14522
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.269496333588722e-06,
+      "loss": 1.075,
+      "step": 14523
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2688433964498624e-06,
+      "loss": 0.9464,
+      "step": 14524
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.268190492851771e-06,
+      "loss": 0.9419,
+      "step": 14525
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.267537622807096e-06,
+      "loss": 0.8844,
+      "step": 14526
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2668847863284862e-06,
+      "loss": 0.7716,
+      "step": 14527
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2662319834285905e-06,
+      "loss": 1.1252,
+      "step": 14528
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.265579214120057e-06,
+      "loss": 0.9616,
+      "step": 14529
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.264926478415533e-06,
+      "loss": 0.9603,
+      "step": 14530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.264273776327665e-06,
+      "loss": 0.8634,
+      "step": 14531
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.263621107869098e-06,
+      "loss": 1.036,
+      "step": 14532
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.262968473052479e-06,
+      "loss": 0.989,
+      "step": 14533
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.262315871890452e-06,
+      "loss": 0.9779,
+      "step": 14534
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2616633043956604e-06,
+      "loss": 0.9026,
+      "step": 14535
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.261010770580746e-06,
+      "loss": 1.0223,
+      "step": 14536
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2603582704583547e-06,
+      "loss": 0.8306,
+      "step": 14537
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2597058040411283e-06,
+      "loss": 1.1043,
+      "step": 14538
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2590533713417073e-06,
+      "loss": 0.8094,
+      "step": 14539
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2584009723727315e-06,
+      "loss": 1.1164,
+      "step": 14540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2577486071468425e-06,
+      "loss": 1.0937,
+      "step": 14541
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.257096275676679e-06,
+      "loss": 0.6588,
+      "step": 14542
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.256443977974879e-06,
+      "loss": 0.9359,
+      "step": 14543
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.255791714054082e-06,
+      "loss": 0.9952,
+      "step": 14544
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2551394839269244e-06,
+      "loss": 0.7991,
+      "step": 14545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.254487287606044e-06,
+      "loss": 0.9948,
+      "step": 14546
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2538351251040757e-06,
+      "loss": 1.0022,
+      "step": 14547
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2531829964336565e-06,
+      "loss": 0.96,
+      "step": 14548
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2525309016074204e-06,
+      "loss": 0.8258,
+      "step": 14549
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2518788406380018e-06,
+      "loss": 0.8857,
+      "step": 14550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2512268135380337e-06,
+      "loss": 0.9518,
+      "step": 14551
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2505748203201496e-06,
+      "loss": 1.0018,
+      "step": 14552
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.249922860996982e-06,
+      "loss": 0.9915,
+      "step": 14553
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2492709355811613e-06,
+      "loss": 0.901,
+      "step": 14554
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2486190440853193e-06,
+      "loss": 0.9808,
+      "step": 14555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2479671865220864e-06,
+      "loss": 0.8474,
+      "step": 14556
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2473153629040923e-06,
+      "loss": 0.9975,
+      "step": 14557
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2466635732439646e-06,
+      "loss": 1.0786,
+      "step": 14558
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2460118175543305e-06,
+      "loss": 0.9483,
+      "step": 14559
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2453600958478216e-06,
+      "loss": 0.9681,
+      "step": 14560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2447084081370627e-06,
+      "loss": 1.0095,
+      "step": 14561
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.24405675443468e-06,
+      "loss": 0.9805,
+      "step": 14562
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2434051347533e-06,
+      "loss": 1.0055,
+      "step": 14563
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.242753549105547e-06,
+      "loss": 0.9279,
+      "step": 14564
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.242101997504046e-06,
+      "loss": 1.0178,
+      "step": 14565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2414504799614187e-06,
+      "loss": 1.0246,
+      "step": 14566
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.24079899649029e-06,
+      "loss": 1.0048,
+      "step": 14567
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2401475471032816e-06,
+      "loss": 0.8793,
+      "step": 14568
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.239496131813016e-06,
+      "loss": 0.9606,
+      "step": 14569
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2388447506321103e-06,
+      "loss": 0.8393,
+      "step": 14570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.238193403573191e-06,
+      "loss": 0.8311,
+      "step": 14571
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.237542090648874e-06,
+      "loss": 0.9756,
+      "step": 14572
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.23689081187178e-06,
+      "loss": 1.0034,
+      "step": 14573
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.236239567254526e-06,
+      "loss": 1.0464,
+      "step": 14574
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2355883568097306e-06,
+      "loss": 0.9224,
+      "step": 14575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2349371805500106e-06,
+      "loss": 0.9647,
+      "step": 14576
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2342860384879825e-06,
+      "loss": 0.9323,
+      "step": 14577
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.233634930636261e-06,
+      "loss": 0.9234,
+      "step": 14578
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2329838570074636e-06,
+      "loss": 0.8294,
+      "step": 14579
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.232332817614202e-06,
+      "loss": 0.8805,
+      "step": 14580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2316818124690915e-06,
+      "loss": 0.9231,
+      "step": 14581
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2310308415847437e-06,
+      "loss": 1.003,
+      "step": 14582
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2303799049737728e-06,
+      "loss": 1.0162,
+      "step": 14583
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2297290026487893e-06,
+      "loss": 0.8004,
+      "step": 14584
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.229078134622406e-06,
+      "loss": 0.9293,
+      "step": 14585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2284273009072314e-06,
+      "loss": 0.8959,
+      "step": 14586
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.227776501515876e-06,
+      "loss": 0.9529,
+      "step": 14587
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.227125736460949e-06,
+      "loss": 0.8959,
+      "step": 14588
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2264750057550594e-06,
+      "loss": 0.9506,
+      "step": 14589
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.225824309410813e-06,
+      "loss": 0.8157,
+      "step": 14590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.225173647440819e-06,
+      "loss": 0.9478,
+      "step": 14591
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2245230198576827e-06,
+      "loss": 1.0793,
+      "step": 14592
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2238724266740107e-06,
+      "loss": 1.0186,
+      "step": 14593
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2232218679024045e-06,
+      "loss": 0.9475,
+      "step": 14594
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2225713435554747e-06,
+      "loss": 1.1374,
+      "step": 14595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2219208536458213e-06,
+      "loss": 1.0314,
+      "step": 14596
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.221270398186048e-06,
+      "loss": 0.9592,
+      "step": 14597
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.220619977188757e-06,
+      "loss": 0.7999,
+      "step": 14598
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2199695906665507e-06,
+      "loss": 1.0861,
+      "step": 14599
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.21931923863203e-06,
+      "loss": 0.9292,
+      "step": 14600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.218668921097795e-06,
+      "loss": 0.8926,
+      "step": 14601
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2180186380764456e-06,
+      "loss": 1.0328,
+      "step": 14602
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.217368389580581e-06,
+      "loss": 0.8312,
+      "step": 14603
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2167181756227984e-06,
+      "loss": 1.0075,
+      "step": 14604
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2160679962156956e-06,
+      "loss": 1.0388,
+      "step": 14605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2154178513718727e-06,
+      "loss": 0.8256,
+      "step": 14606
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2147677411039236e-06,
+      "loss": 0.9407,
+      "step": 14607
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2141176654244455e-06,
+      "loss": 0.8297,
+      "step": 14608
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2134676243460317e-06,
+      "loss": 0.9512,
+      "step": 14609
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2128176178812775e-06,
+      "loss": 0.922,
+      "step": 14610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2121676460427765e-06,
+      "loss": 0.9612,
+      "step": 14611
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2115177088431226e-06,
+      "loss": 0.9418,
+      "step": 14612
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2108678062949063e-06,
+      "loss": 1.0718,
+      "step": 14613
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2102179384107215e-06,
+      "loss": 0.8523,
+      "step": 14614
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.209568105203157e-06,
+      "loss": 0.9171,
+      "step": 14615
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.208918306684805e-06,
+      "loss": 0.9207,
+      "step": 14616
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2082685428682535e-06,
+      "loss": 0.8223,
+      "step": 14617
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2076188137660934e-06,
+      "loss": 1.0271,
+      "step": 14618
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2069691193909123e-06,
+      "loss": 0.9717,
+      "step": 14619
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.206319459755298e-06,
+      "loss": 0.9381,
+      "step": 14620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.205669834871837e-06,
+      "loss": 1.0129,
+      "step": 14621
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2050202447531164e-06,
+      "loss": 0.9296,
+      "step": 14622
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.204370689411721e-06,
+      "loss": 1.0744,
+      "step": 14623
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2037211688602353e-06,
+      "loss": 1.0018,
+      "step": 14624
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.203071683111246e-06,
+      "loss": 0.7853,
+      "step": 14625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2024222321773346e-06,
+      "loss": 0.9272,
+      "step": 14626
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2017728160710848e-06,
+      "loss": 0.7756,
+      "step": 14627
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2011234348050784e-06,
+      "loss": 0.9599,
+      "step": 14628
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2004740883918956e-06,
+      "loss": 0.9301,
+      "step": 14629
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1998247768441216e-06,
+      "loss": 0.8844,
+      "step": 14630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.199175500174333e-06,
+      "loss": 0.8861,
+      "step": 14631
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1985262583951113e-06,
+      "loss": 0.8384,
+      "step": 14632
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.197877051519035e-06,
+      "loss": 0.7283,
+      "step": 14633
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.197227879558682e-06,
+      "loss": 1.0862,
+      "step": 14634
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.19657874252663e-06,
+      "loss": 1.0449,
+      "step": 14635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1959296404354555e-06,
+      "loss": 1.0808,
+      "step": 14636
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.195280573297735e-06,
+      "loss": 0.9122,
+      "step": 14637
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.194631541126044e-06,
+      "loss": 0.9639,
+      "step": 14638
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1939825439329574e-06,
+      "loss": 0.9461,
+      "step": 14639
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.193333581731048e-06,
+      "loss": 0.8682,
+      "step": 14640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1926846545328917e-06,
+      "loss": 1.0413,
+      "step": 14641
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1920357623510604e-06,
+      "loss": 0.9964,
+      "step": 14642
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.191386905198126e-06,
+      "loss": 1.0826,
+      "step": 14643
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1907380830866608e-06,
+      "loss": 0.8549,
+      "step": 14644
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.190089296029234e-06,
+      "loss": 0.796,
+      "step": 14645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1894405440384168e-06,
+      "loss": 0.9221,
+      "step": 14646
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1887918271267784e-06,
+      "loss": 0.831,
+      "step": 14647
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.188143145306888e-06,
+      "loss": 1.0621,
+      "step": 14648
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.187494498591312e-06,
+      "loss": 0.9945,
+      "step": 14649
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.18684588699262e-06,
+      "loss": 1.0705,
+      "step": 14650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.186197310523377e-06,
+      "loss": 0.8004,
+      "step": 14651
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1855487691961474e-06,
+      "loss": 1.0298,
+      "step": 14652
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1849002630235013e-06,
+      "loss": 0.9743,
+      "step": 14653
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.184251792018001e-06,
+      "loss": 0.9739,
+      "step": 14654
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.18360335619221e-06,
+      "loss": 0.8997,
+      "step": 14655
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1829549555586924e-06,
+      "loss": 0.9611,
+      "step": 14656
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.18230659013001e-06,
+      "loss": 0.9618,
+      "step": 14657
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1816582599187252e-06,
+      "loss": 0.9583,
+      "step": 14658
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1810099649373993e-06,
+      "loss": 0.9859,
+      "step": 14659
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1803617051985923e-06,
+      "loss": 0.793,
+      "step": 14660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.179713480714865e-06,
+      "loss": 0.9808,
+      "step": 14661
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.179065291498776e-06,
+      "loss": 0.9226,
+      "step": 14662
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1784171375628815e-06,
+      "loss": 0.9378,
+      "step": 14663
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.177769018919744e-06,
+      "loss": 0.935,
+      "step": 14664
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1771209355819187e-06,
+      "loss": 0.9327,
+      "step": 14665
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.176472887561961e-06,
+      "loss": 0.9404,
+      "step": 14666
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1758248748724287e-06,
+      "loss": 1.1436,
+      "step": 14667
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1751768975258745e-06,
+      "loss": 1.0695,
+      "step": 14668
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1745289555348547e-06,
+      "loss": 0.949,
+      "step": 14669
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.173881048911922e-06,
+      "loss": 1.1913,
+      "step": 14670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1732331776696294e-06,
+      "loss": 1.0501,
+      "step": 14671
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1725853418205297e-06,
+      "loss": 1.0734,
+      "step": 14672
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1719375413771747e-06,
+      "loss": 1.1341,
+      "step": 14673
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1712897763521156e-06,
+      "loss": 0.9075,
+      "step": 14674
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.170642046757901e-06,
+      "loss": 0.9454,
+      "step": 14675
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.169994352607082e-06,
+      "loss": 1.0613,
+      "step": 14676
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1693466939122085e-06,
+      "loss": 0.8461,
+      "step": 14677
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1686990706858265e-06,
+      "loss": 0.8954,
+      "step": 14678
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.168051482940485e-06,
+      "loss": 0.9878,
+      "step": 14679
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1674039306887306e-06,
+      "loss": 0.701,
+      "step": 14680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1667564139431094e-06,
+      "loss": 0.7145,
+      "step": 14681
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1661089327161666e-06,
+      "loss": 0.8754,
+      "step": 14682
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1654614870204476e-06,
+      "loss": 0.8057,
+      "step": 14683
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.164814076868496e-06,
+      "loss": 0.9242,
+      "step": 14684
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.164166702272855e-06,
+      "loss": 0.7366,
+      "step": 14685
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.163519363246068e-06,
+      "loss": 0.9147,
+      "step": 14686
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1628720598006746e-06,
+      "loss": 0.8675,
+      "step": 14687
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1622247919492205e-06,
+      "loss": 0.9547,
+      "step": 14688
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1615775597042443e-06,
+      "loss": 0.858,
+      "step": 14689
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.160930363078285e-06,
+      "loss": 0.9425,
+      "step": 14690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1602832020838837e-06,
+      "loss": 0.9426,
+      "step": 14691
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1596360767335777e-06,
+      "loss": 0.8905,
+      "step": 14692
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1589889870399055e-06,
+      "loss": 0.7241,
+      "step": 14693
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.158341933015404e-06,
+      "loss": 0.7611,
+      "step": 14694
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.15769491467261e-06,
+      "loss": 0.8866,
+      "step": 14695
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.157047932024058e-06,
+      "loss": 0.9073,
+      "step": 14696
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.156400985082285e-06,
+      "loss": 0.9146,
+      "step": 14697
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1557540738598232e-06,
+      "loss": 1.0939,
+      "step": 14698
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1551071983692095e-06,
+      "loss": 0.8902,
+      "step": 14699
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.154460358622975e-06,
+      "loss": 1.0114,
+      "step": 14700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1538135546336522e-06,
+      "loss": 0.9417,
+      "step": 14701
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1531667864137736e-06,
+      "loss": 0.8464,
+      "step": 14702
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1525200539758697e-06,
+      "loss": 0.9642,
+      "step": 14703
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1518733573324707e-06,
+      "loss": 1.0036,
+      "step": 14704
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1512266964961058e-06,
+      "loss": 0.9865,
+      "step": 14705
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1505800714793044e-06,
+      "loss": 1.0798,
+      "step": 14706
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.149933482294595e-06,
+      "loss": 0.8575,
+      "step": 14707
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1492869289545037e-06,
+      "loss": 0.7652,
+      "step": 14708
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.148640411471559e-06,
+      "loss": 0.9028,
+      "step": 14709
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.147993929858285e-06,
+      "loss": 0.9011,
+      "step": 14710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.14734748412721e-06,
+      "loss": 0.9934,
+      "step": 14711
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1467010742908573e-06,
+      "loss": 0.9213,
+      "step": 14712
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1460547003617502e-06,
+      "loss": 0.9781,
+      "step": 14713
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1454083623524136e-06,
+      "loss": 0.9402,
+      "step": 14714
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.144762060275369e-06,
+      "loss": 0.9092,
+      "step": 14715
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1441157941431384e-06,
+      "loss": 0.9285,
+      "step": 14716
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1434695639682432e-06,
+      "loss": 0.9047,
+      "step": 14717
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1428233697632044e-06,
+      "loss": 0.8426,
+      "step": 14718
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1421772115405415e-06,
+      "loss": 0.8013,
+      "step": 14719
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1415310893127736e-06,
+      "loss": 0.9996,
+      "step": 14720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1408850030924174e-06,
+      "loss": 1.051,
+      "step": 14721
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.140238952891994e-06,
+      "loss": 0.9211,
+      "step": 14722
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.13959293872402e-06,
+      "loss": 1.1738,
+      "step": 14723
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1389469606010103e-06,
+      "loss": 0.9186,
+      "step": 14724
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1383010185354804e-06,
+      "loss": 0.922,
+      "step": 14725
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.137655112539947e-06,
+      "loss": 0.9675,
+      "step": 14726
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1370092426269226e-06,
+      "loss": 1.0187,
+      "step": 14727
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1363634088089224e-06,
+      "loss": 0.9497,
+      "step": 14728
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1357176110984578e-06,
+      "loss": 0.5818,
+      "step": 14729
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.135071849508041e-06,
+      "loss": 0.9061,
+      "step": 14730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.134426124050185e-06,
+      "loss": 0.9363,
+      "step": 14731
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1337804347374e-06,
+      "loss": 0.9358,
+      "step": 14732
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1331347815821937e-06,
+      "loss": 0.8054,
+      "step": 14733
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1324891645970788e-06,
+      "loss": 0.7567,
+      "step": 14734
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.131843583794563e-06,
+      "loss": 0.8575,
+      "step": 14735
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1311980391871544e-06,
+      "loss": 0.9092,
+      "step": 14736
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.130552530787359e-06,
+      "loss": 1.0464,
+      "step": 14737
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.129907058607685e-06,
+      "loss": 1.0398,
+      "step": 14738
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.129261622660637e-06,
+      "loss": 0.8379,
+      "step": 14739
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1286162229587208e-06,
+      "loss": 0.9528,
+      "step": 14740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.127970859514441e-06,
+      "loss": 1.0539,
+      "step": 14741
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1273255323403005e-06,
+      "loss": 0.7757,
+      "step": 14742
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.126680241448804e-06,
+      "loss": 0.8522,
+      "step": 14743
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1260349868524526e-06,
+      "loss": 1.0848,
+      "step": 14744
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1253897685637457e-06,
+      "loss": 0.9689,
+      "step": 14745
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1247445865951894e-06,
+      "loss": 0.9901,
+      "step": 14746
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1240994409592804e-06,
+      "loss": 0.9042,
+      "step": 14747
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1234543316685206e-06,
+      "loss": 0.9692,
+      "step": 14748
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1228092587354066e-06,
+      "loss": 0.6945,
+      "step": 14749
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.122164222172438e-06,
+      "loss": 0.9563,
+      "step": 14750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.121519221992111e-06,
+      "loss": 0.9966,
+      "step": 14751
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.120874258206923e-06,
+      "loss": 0.8164,
+      "step": 14752
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1202293308293703e-06,
+      "loss": 0.8071,
+      "step": 14753
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1195844398719477e-06,
+      "loss": 1.0206,
+      "step": 14754
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.11893958534715e-06,
+      "loss": 0.9839,
+      "step": 14755
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.118294767267469e-06,
+      "loss": 0.9828,
+      "step": 14756
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1176499856454022e-06,
+      "loss": 0.6548,
+      "step": 14757
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1170052404934402e-06,
+      "loss": 1.0549,
+      "step": 14758
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.116360531824074e-06,
+      "loss": 0.9629,
+      "step": 14759
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.115715859649796e-06,
+      "loss": 0.9604,
+      "step": 14760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1150712239830948e-06,
+      "loss": 0.9628,
+      "step": 14761
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1144266248364618e-06,
+      "loss": 1.0116,
+      "step": 14762
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.113782062222385e-06,
+      "loss": 0.8816,
+      "step": 14763
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1131375361533533e-06,
+      "loss": 1.0525,
+      "step": 14764
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1124930466418534e-06,
+      "loss": 0.8916,
+      "step": 14765
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.111848593700373e-06,
+      "loss": 1.151,
+      "step": 14766
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.111204177341397e-06,
+      "loss": 0.9636,
+      "step": 14767
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.110559797577411e-06,
+      "loss": 0.9582,
+      "step": 14768
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.109915454420901e-06,
+      "loss": 1.0243,
+      "step": 14769
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1092711478843508e-06,
+      "loss": 0.9361,
+      "step": 14770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.108626877980243e-06,
+      "loss": 1.0294,
+      "step": 14771
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1079826447210604e-06,
+      "loss": 0.8546,
+      "step": 14772
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.107338448119285e-06,
+      "loss": 0.8306,
+      "step": 14773
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.106694288187398e-06,
+      "loss": 0.9843,
+      "step": 14774
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.106050164937879e-06,
+      "loss": 0.6581,
+      "step": 14775
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1054060783832097e-06,
+      "loss": 1.0607,
+      "step": 14776
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.104762028535867e-06,
+      "loss": 0.9534,
+      "step": 14777
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.10411801540833e-06,
+      "loss": 0.9188,
+      "step": 14778
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.103474039013077e-06,
+      "loss": 0.9731,
+      "step": 14779
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1028300993625815e-06,
+      "loss": 0.8143,
+      "step": 14780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1021861964693245e-06,
+      "loss": 0.8756,
+      "step": 14781
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1015423303457803e-06,
+      "loss": 0.9195,
+      "step": 14782
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1008985010044226e-06,
+      "loss": 0.9251,
+      "step": 14783
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.100254708457725e-06,
+      "loss": 1.0346,
+      "step": 14784
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.099610952718162e-06,
+      "loss": 0.927,
+      "step": 14785
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0989672337982052e-06,
+      "loss": 1.052,
+      "step": 14786
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0983235517103274e-06,
+      "loss": 0.9702,
+      "step": 14787
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0976799064669995e-06,
+      "loss": 1.0137,
+      "step": 14788
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0970362980806917e-06,
+      "loss": 0.939,
+      "step": 14789
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.096392726563874e-06,
+      "loss": 0.8918,
+      "step": 14790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0957491919290136e-06,
+      "loss": 0.9344,
+      "step": 14791
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.095105694188582e-06,
+      "loss": 0.8883,
+      "step": 14792
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.094462233355046e-06,
+      "loss": 0.985,
+      "step": 14793
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.093818809440872e-06,
+      "loss": 0.9179,
+      "step": 14794
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0931754224585263e-06,
+      "loss": 0.9536,
+      "step": 14795
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0925320724204743e-06,
+      "loss": 1.0335,
+      "step": 14796
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.09188875933918e-06,
+      "loss": 0.8712,
+      "step": 14797
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0912454832271087e-06,
+      "loss": 0.789,
+      "step": 14798
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0906022440967233e-06,
+      "loss": 1.0567,
+      "step": 14799
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0899590419604863e-06,
+      "loss": 0.9946,
+      "step": 14800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0893158768308595e-06,
+      "loss": 1.1159,
+      "step": 14801
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0886727487203037e-06,
+      "loss": 0.9817,
+      "step": 14802
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0880296576412796e-06,
+      "loss": 1.0064,
+      "step": 14803
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0873866036062474e-06,
+      "loss": 0.9451,
+      "step": 14804
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.086743586627666e-06,
+      "loss": 0.7977,
+      "step": 14805
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0861006067179945e-06,
+      "loss": 0.7938,
+      "step": 14806
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.085457663889689e-06,
+      "loss": 0.9547,
+      "step": 14807
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.084814758155207e-06,
+      "loss": 1.0469,
+      "step": 14808
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0841718895270045e-06,
+      "loss": 0.7713,
+      "step": 14809
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.083529058017537e-06,
+      "loss": 0.8568,
+      "step": 14810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.08288626363926e-06,
+      "loss": 1.0839,
+      "step": 14811
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.082243506404625e-06,
+      "loss": 1.0026,
+      "step": 14812
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.081600786326089e-06,
+      "loss": 0.9479,
+      "step": 14813
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0809581034160994e-06,
+      "loss": 0.9065,
+      "step": 14814
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.080315457687114e-06,
+      "loss": 0.9542,
+      "step": 14815
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0796728491515804e-06,
+      "loss": 0.8614,
+      "step": 14816
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0790302778219507e-06,
+      "loss": 0.6646,
+      "step": 14817
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.078387743710673e-06,
+      "loss": 0.8808,
+      "step": 14818
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.077745246830197e-06,
+      "loss": 0.9666,
+      "step": 14819
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0771027871929704e-06,
+      "loss": 0.8677,
+      "step": 14820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.076460364811442e-06,
+      "loss": 0.9841,
+      "step": 14821
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.075817979698057e-06,
+      "loss": 0.9547,
+      "step": 14822
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0751756318652624e-06,
+      "loss": 0.8668,
+      "step": 14823
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.074533321325503e-06,
+      "loss": 0.9357,
+      "step": 14824
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0738910480912242e-06,
+      "loss": 0.9595,
+      "step": 14825
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0732488121748684e-06,
+      "loss": 0.9346,
+      "step": 14826
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.07260661358888e-06,
+      "loss": 0.9917,
+      "step": 14827
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.071964452345702e-06,
+      "loss": 0.8034,
+      "step": 14828
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0713223284577754e-06,
+      "loss": 1.0092,
+      "step": 14829
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.070680241937541e-06,
+      "loss": 0.8193,
+      "step": 14830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0700381927974395e-06,
+      "loss": 0.8165,
+      "step": 14831
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0693961810499095e-06,
+      "loss": 0.7955,
+      "step": 14832
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.068754206707392e-06,
+      "loss": 0.7535,
+      "step": 14833
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.068112269782322e-06,
+      "loss": 0.9072,
+      "step": 14834
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0674703702871392e-06,
+      "loss": 0.8987,
+      "step": 14835
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0668285082342797e-06,
+      "loss": 0.9536,
+      "step": 14836
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.06618668363618e-06,
+      "loss": 0.9669,
+      "step": 14837
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0655448965052716e-06,
+      "loss": 0.9854,
+      "step": 14838
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0649031468539946e-06,
+      "loss": 1.0882,
+      "step": 14839
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0642614346947797e-06,
+      "loss": 0.978,
+      "step": 14840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0636197600400607e-06,
+      "loss": 1.0564,
+      "step": 14841
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.06297812290227e-06,
+      "loss": 0.9554,
+      "step": 14842
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0623365232938386e-06,
+      "loss": 0.8802,
+      "step": 14843
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0616949612271968e-06,
+      "loss": 0.9163,
+      "step": 14844
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.061053436714776e-06,
+      "loss": 1.078,
+      "step": 14845
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.060411949769005e-06,
+      "loss": 0.9248,
+      "step": 14846
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0597705004023127e-06,
+      "loss": 0.9765,
+      "step": 14847
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0591290886271264e-06,
+      "loss": 1.0424,
+      "step": 14848
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0584877144558724e-06,
+      "loss": 0.9408,
+      "step": 14849
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0578463779009794e-06,
+      "loss": 0.9366,
+      "step": 14850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0572050789748732e-06,
+      "loss": 0.9758,
+      "step": 14851
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0565638176899774e-06,
+      "loss": 0.7953,
+      "step": 14852
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0559225940587174e-06,
+      "loss": 0.9395,
+      "step": 14853
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0552814080935155e-06,
+      "loss": 0.8913,
+      "step": 14854
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.054640259806796e-06,
+      "loss": 1.0375,
+      "step": 14855
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0539991492109786e-06,
+      "loss": 0.8762,
+      "step": 14856
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0533580763184876e-06,
+      "loss": 1.0119,
+      "step": 14857
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0527170411417424e-06,
+      "loss": 0.8897,
+      "step": 14858
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.052076043693162e-06,
+      "loss": 0.9768,
+      "step": 14859
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.051435083985167e-06,
+      "loss": 0.8971,
+      "step": 14860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.050794162030174e-06,
+      "loss": 0.832,
+      "step": 14861
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0501532778406034e-06,
+      "loss": 0.9064,
+      "step": 14862
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0495124314288702e-06,
+      "loss": 0.9262,
+      "step": 14863
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0488716228073913e-06,
+      "loss": 1.0262,
+      "step": 14864
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0482308519885824e-06,
+      "loss": 0.9646,
+      "step": 14865
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0475901189848577e-06,
+      "loss": 1.0017,
+      "step": 14866
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0469494238086317e-06,
+      "loss": 1.0096,
+      "step": 14867
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.046308766472318e-06,
+      "loss": 0.8998,
+      "step": 14868
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0456681469883276e-06,
+      "loss": 0.8548,
+      "step": 14869
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0450275653690743e-06,
+      "loss": 1.024,
+      "step": 14870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.044387021626969e-06,
+      "loss": 1.044,
+      "step": 14871
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0437465157744207e-06,
+      "loss": 0.9538,
+      "step": 14872
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.043106047823838e-06,
+      "loss": 1.1324,
+      "step": 14873
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0424656177876337e-06,
+      "loss": 0.95,
+      "step": 14874
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0418252256782144e-06,
+      "loss": 0.9927,
+      "step": 14875
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0411848715079873e-06,
+      "loss": 0.8532,
+      "step": 14876
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.040544555289358e-06,
+      "loss": 0.8414,
+      "step": 14877
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0399042770347343e-06,
+      "loss": 0.9404,
+      "step": 14878
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0392640367565205e-06,
+      "loss": 1.0245,
+      "step": 14879
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0386238344671207e-06,
+      "loss": 1.043,
+      "step": 14880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.03798367017894e-06,
+      "loss": 0.9943,
+      "step": 14881
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0373435439043803e-06,
+      "loss": 0.8521,
+      "step": 14882
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.036703455655844e-06,
+      "loss": 0.8023,
+      "step": 14883
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.036063405445732e-06,
+      "loss": 0.8782,
+      "step": 14884
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0354233932864473e-06,
+      "loss": 0.9936,
+      "step": 14885
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0347834191903887e-06,
+      "loss": 0.9657,
+      "step": 14886
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0341434831699552e-06,
+      "loss": 0.8811,
+      "step": 14887
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.033503585237545e-06,
+      "loss": 0.8984,
+      "step": 14888
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0328637254055575e-06,
+      "loss": 0.945,
+      "step": 14889
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0322239036863897e-06,
+      "loss": 0.9483,
+      "step": 14890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0315841200924366e-06,
+      "loss": 1.0636,
+      "step": 14891
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.030944374636096e-06,
+      "loss": 0.936,
+      "step": 14892
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0303046673297598e-06,
+      "loss": 0.8839,
+      "step": 14893
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.029664998185825e-06,
+      "loss": 0.9242,
+      "step": 14894
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0290253672166837e-06,
+      "loss": 0.9175,
+      "step": 14895
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0283857744347273e-06,
+      "loss": 0.9886,
+      "step": 14896
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.027746219852351e-06,
+      "loss": 0.9479,
+      "step": 14897
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0271067034819436e-06,
+      "loss": 0.9748,
+      "step": 14898
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0264672253358966e-06,
+      "loss": 1.067,
+      "step": 14899
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0258277854265995e-06,
+      "loss": 0.9923,
+      "step": 14900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0251883837664413e-06,
+      "loss": 1.0135,
+      "step": 14901
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.02454902036781e-06,
+      "loss": 0.8727,
+      "step": 14902
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0239096952430935e-06,
+      "loss": 0.8606,
+      "step": 14903
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.023270408404678e-06,
+      "loss": 1.0419,
+      "step": 14904
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0226311598649504e-06,
+      "loss": 1.1501,
+      "step": 14905
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0219919496362958e-06,
+      "loss": 0.9794,
+      "step": 14906
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.021352777731096e-06,
+      "loss": 1.0335,
+      "step": 14907
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0207136441617398e-06,
+      "loss": 0.9616,
+      "step": 14908
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.020074548940607e-06,
+      "loss": 0.9004,
+      "step": 14909
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0194354920800817e-06,
+      "loss": 0.9503,
+      "step": 14910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0187964735925436e-06,
+      "loss": 0.8442,
+      "step": 14911
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0181574934903745e-06,
+      "loss": 0.9429,
+      "step": 14912
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0175185517859546e-06,
+      "loss": 1.0417,
+      "step": 14913
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.016879648491663e-06,
+      "loss": 0.7866,
+      "step": 14914
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.016240783619878e-06,
+      "loss": 0.8956,
+      "step": 14915
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.015601957182979e-06,
+      "loss": 0.9799,
+      "step": 14916
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0149631691933413e-06,
+      "loss": 1.0557,
+      "step": 14917
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0143244196633414e-06,
+      "loss": 0.9282,
+      "step": 14918
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.013685708605355e-06,
+      "loss": 0.8529,
+      "step": 14919
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0130470360317584e-06,
+      "loss": 0.9381,
+      "step": 14920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0124084019549242e-06,
+      "loss": 0.933,
+      "step": 14921
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.011769806387227e-06,
+      "loss": 0.8201,
+      "step": 14922
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0111312493410384e-06,
+      "loss": 0.8083,
+      "step": 14923
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0104927308287314e-06,
+      "loss": 1.0801,
+      "step": 14924
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0098542508626756e-06,
+      "loss": 0.9721,
+      "step": 14925
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0092158094552425e-06,
+      "loss": 0.7224,
+      "step": 14926
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.008577406618801e-06,
+      "loss": 0.9422,
+      "step": 14927
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0079390423657217e-06,
+      "loss": 0.9506,
+      "step": 14928
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0073007167083707e-06,
+      "loss": 0.9179,
+      "step": 14929
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.006662429659116e-06,
+      "loss": 1.0077,
+      "step": 14930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0060241812303236e-06,
+      "loss": 0.869,
+      "step": 14931
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.005385971434362e-06,
+      "loss": 1.0119,
+      "step": 14932
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0047478002835946e-06,
+      "loss": 0.9067,
+      "step": 14933
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0041096677903857e-06,
+      "loss": 0.8852,
+      "step": 14934
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0034715739670993e-06,
+      "loss": 0.9113,
+      "step": 14935
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0028335188260995e-06,
+      "loss": 0.9223,
+      "step": 14936
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0021955023797454e-06,
+      "loss": 0.8055,
+      "step": 14937
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0015575246404017e-06,
+      "loss": 0.7923,
+      "step": 14938
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0009195856204264e-06,
+      "loss": 0.8818,
+      "step": 14939
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0002816853321816e-06,
+      "loss": 0.9906,
+      "step": 14940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9996438237880255e-06,
+      "loss": 1.0637,
+      "step": 14941
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9990060010003143e-06,
+      "loss": 0.8796,
+      "step": 14942
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.99836821698141e-06,
+      "loss": 0.9577,
+      "step": 14943
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.997730471743667e-06,
+      "loss": 0.8644,
+      "step": 14944
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9970927652994426e-06,
+      "loss": 0.9489,
+      "step": 14945
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.996455097661091e-06,
+      "loss": 0.8229,
+      "step": 14946
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9958174688409682e-06,
+      "loss": 0.9337,
+      "step": 14947
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9951798788514263e-06,
+      "loss": 0.936,
+      "step": 14948
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.99454232770482e-06,
+      "loss": 0.931,
+      "step": 14949
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9939048154135008e-06,
+      "loss": 1.0026,
+      "step": 14950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9932673419898217e-06,
+      "loss": 0.9271,
+      "step": 14951
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.992629907446132e-06,
+      "loss": 0.9425,
+      "step": 14952
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9919925117947824e-06,
+      "loss": 0.7069,
+      "step": 14953
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.991355155048121e-06,
+      "loss": 0.9135,
+      "step": 14954
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9907178372184996e-06,
+      "loss": 1.0348,
+      "step": 14955
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9900805583182645e-06,
+      "loss": 0.9947,
+      "step": 14956
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9894433183597615e-06,
+      "loss": 0.965,
+      "step": 14957
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9888061173553394e-06,
+      "loss": 0.9889,
+      "step": 14958
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.988168955317342e-06,
+      "loss": 0.9077,
+      "step": 14959
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9875318322581147e-06,
+      "loss": 0.7653,
+      "step": 14960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.986894748190001e-06,
+      "loss": 0.9056,
+      "step": 14961
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9862577031253457e-06,
+      "loss": 0.9706,
+      "step": 14962
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.985620697076491e-06,
+      "loss": 1.0996,
+      "step": 14963
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9849837300557782e-06,
+      "loss": 1.0062,
+      "step": 14964
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.984346802075548e-06,
+      "loss": 0.9266,
+      "step": 14965
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9837099131481395e-06,
+      "loss": 0.9692,
+      "step": 14966
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.983073063285896e-06,
+      "loss": 0.8087,
+      "step": 14967
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.982436252501154e-06,
+      "loss": 1.0067,
+      "step": 14968
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9817994808062528e-06,
+      "loss": 1.133,
+      "step": 14969
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9811627482135284e-06,
+      "loss": 0.7809,
+      "step": 14970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9805260547353177e-06,
+      "loss": 0.853,
+      "step": 14971
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9798894003839568e-06,
+      "loss": 0.8415,
+      "step": 14972
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9792527851717805e-06,
+      "loss": 0.7746,
+      "step": 14973
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.978616209111123e-06,
+      "loss": 0.9358,
+      "step": 14974
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9779796722143185e-06,
+      "loss": 0.9722,
+      "step": 14975
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9773431744936986e-06,
+      "loss": 0.9766,
+      "step": 14976
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9767067159615947e-06,
+      "loss": 0.8526,
+      "step": 14977
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.97607029663034e-06,
+      "loss": 0.9792,
+      "step": 14978
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.975433916512265e-06,
+      "loss": 0.9366,
+      "step": 14979
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9747975756196983e-06,
+      "loss": 0.9582,
+      "step": 14980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9741612739649694e-06,
+      "loss": 1.0851,
+      "step": 14981
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9735250115604065e-06,
+      "loss": 0.9092,
+      "step": 14982
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9728887884183355e-06,
+      "loss": 0.8544,
+      "step": 14983
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.972252604551086e-06,
+      "loss": 1.0619,
+      "step": 14984
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.971616459970981e-06,
+      "loss": 0.891,
+      "step": 14985
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.970980354690347e-06,
+      "loss": 0.8766,
+      "step": 14986
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9703442887215085e-06,
+      "loss": 0.964,
+      "step": 14987
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9697082620767897e-06,
+      "loss": 0.9768,
+      "step": 14988
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9690722747685114e-06,
+      "loss": 0.9069,
+      "step": 14989
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9684363268089983e-06,
+      "loss": 0.8458,
+      "step": 14990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.96780041821057e-06,
+      "loss": 0.7842,
+      "step": 14991
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.967164548985548e-06,
+      "loss": 0.9986,
+      "step": 14992
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.966528719146252e-06,
+      "loss": 0.9129,
+      "step": 14993
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.965892928705001e-06,
+      "loss": 0.7406,
+      "step": 14994
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9652571776741122e-06,
+      "loss": 0.9823,
+      "step": 14995
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9646214660659044e-06,
+      "loss": 0.9345,
+      "step": 14996
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9639857938926943e-06,
+      "loss": 0.7315,
+      "step": 14997
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9633501611667975e-06,
+      "loss": 0.9434,
+      "step": 14998
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.962714567900529e-06,
+      "loss": 0.8047,
+      "step": 14999
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.962079014106202e-06,
+      "loss": 0.868,
+      "step": 15000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.961443499796134e-06,
+      "loss": 0.8831,
+      "step": 15001
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9608080249826353e-06,
+      "loss": 0.9773,
+      "step": 15002
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9601725896780186e-06,
+      "loss": 0.7797,
+      "step": 15003
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9595371938945955e-06,
+      "loss": 0.9654,
+      "step": 15004
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9589018376446767e-06,
+      "loss": 0.9966,
+      "step": 15005
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.958266520940571e-06,
+      "loss": 1.0174,
+      "step": 15006
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9576312437945887e-06,
+      "loss": 0.9189,
+      "step": 15007
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.956996006219037e-06,
+      "loss": 0.8094,
+      "step": 15008
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.956360808226225e-06,
+      "loss": 1.0146,
+      "step": 15009
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9557256498284593e-06,
+      "loss": 0.964,
+      "step": 15010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9550905310380453e-06,
+      "loss": 0.878,
+      "step": 15011
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9544554518672865e-06,
+      "loss": 1.0712,
+      "step": 15012
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9538204123284907e-06,
+      "loss": 0.8917,
+      "step": 15013
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9531854124339604e-06,
+      "loss": 1.0148,
+      "step": 15014
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9525504521959985e-06,
+      "loss": 0.8982,
+      "step": 15015
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.951915531626907e-06,
+      "loss": 0.8184,
+      "step": 15016
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.951280650738988e-06,
+      "loss": 0.8818,
+      "step": 15017
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9506458095445408e-06,
+      "loss": 1.1773,
+      "step": 15018
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9500110080558663e-06,
+      "loss": 0.903,
+      "step": 15019
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.949376246285264e-06,
+      "loss": 0.8621,
+      "step": 15020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.948741524245031e-06,
+      "loss": 0.7547,
+      "step": 15021
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.948106841947467e-06,
+      "loss": 0.9974,
+      "step": 15022
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.947472199404866e-06,
+      "loss": 0.9148,
+      "step": 15023
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.946837596629525e-06,
+      "loss": 0.9122,
+      "step": 15024
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.946203033633741e-06,
+      "loss": 0.9626,
+      "step": 15025
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.945568510429807e-06,
+      "loss": 0.9541,
+      "step": 15026
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9449340270300174e-06,
+      "loss": 0.8578,
+      "step": 15027
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.944299583446665e-06,
+      "loss": 0.9854,
+      "step": 15028
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9436651796920423e-06,
+      "loss": 1.0012,
+      "step": 15029
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9430308157784395e-06,
+      "loss": 0.802,
+      "step": 15030
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.942396491718148e-06,
+      "loss": 1.0211,
+      "step": 15031
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9417622075234585e-06,
+      "loss": 0.9044,
+      "step": 15032
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.941127963206659e-06,
+      "loss": 1.0136,
+      "step": 15033
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9404937587800374e-06,
+      "loss": 0.9802,
+      "step": 15034
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9398595942558816e-06,
+      "loss": 1.0251,
+      "step": 15035
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9392254696464794e-06,
+      "loss": 1.1298,
+      "step": 15036
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.938591384964117e-06,
+      "loss": 0.9008,
+      "step": 15037
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9379573402210793e-06,
+      "loss": 1.1751,
+      "step": 15038
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9373233354296493e-06,
+      "loss": 0.9526,
+      "step": 15039
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9366893706021125e-06,
+      "loss": 0.896,
+      "step": 15040
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9360554457507505e-06,
+      "loss": 0.7576,
+      "step": 15041
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9354215608878465e-06,
+      "loss": 0.8808,
+      "step": 15042
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.934787716025682e-06,
+      "loss": 1.0856,
+      "step": 15043
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.934153911176536e-06,
+      "loss": 0.9223,
+      "step": 15044
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.93352014635269e-06,
+      "loss": 0.9342,
+      "step": 15045
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.932886421566422e-06,
+      "loss": 0.9085,
+      "step": 15046
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.93225273683001e-06,
+      "loss": 0.9909,
+      "step": 15047
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.931619092155733e-06,
+      "loss": 0.9245,
+      "step": 15048
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9309854875558674e-06,
+      "loss": 0.9176,
+      "step": 15049
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.930351923042688e-06,
+      "loss": 0.9008,
+      "step": 15050
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.929718398628471e-06,
+      "loss": 1.0593,
+      "step": 15051
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9290849143254908e-06,
+      "loss": 0.7867,
+      "step": 15052
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.92845147014602e-06,
+      "loss": 0.8402,
+      "step": 15053
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.927818066102332e-06,
+      "loss": 0.893,
+      "step": 15054
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9271847022066992e-06,
+      "loss": 0.985,
+      "step": 15055
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.926551378471392e-06,
+      "loss": 0.8443,
+      "step": 15056
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9259180949086817e-06,
+      "loss": 0.8776,
+      "step": 15057
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9252848515308385e-06,
+      "loss": 0.8215,
+      "step": 15058
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9246516483501288e-06,
+      "loss": 0.9017,
+      "step": 15059
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9240184853788235e-06,
+      "loss": 0.8435,
+      "step": 15060
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9233853626291895e-06,
+      "loss": 0.9169,
+      "step": 15061
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.922752280113494e-06,
+      "loss": 0.8362,
+      "step": 15062
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.922119237844001e-06,
+      "loss": 0.808,
+      "step": 15063
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9214862358329765e-06,
+      "loss": 0.9507,
+      "step": 15064
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.920853274092685e-06,
+      "loss": 0.9603,
+      "step": 15065
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9202203526353896e-06,
+      "loss": 0.9154,
+      "step": 15066
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.919587471473353e-06,
+      "loss": 1.1104,
+      "step": 15067
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.918954630618837e-06,
+      "loss": 0.9011,
+      "step": 15068
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9183218300841043e-06,
+      "loss": 1.0401,
+      "step": 15069
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.91768906988141e-06,
+      "loss": 0.9529,
+      "step": 15070
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.917056350023021e-06,
+      "loss": 1.0813,
+      "step": 15071
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.916423670521191e-06,
+      "loss": 0.8087,
+      "step": 15072
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.915791031388182e-06,
+      "loss": 0.9298,
+      "step": 15073
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.915158432636247e-06,
+      "loss": 0.9728,
+      "step": 15074
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9145258742776467e-06,
+      "loss": 0.9802,
+      "step": 15075
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.913893356324632e-06,
+      "loss": 1.0258,
+      "step": 15076
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.913260878789462e-06,
+      "loss": 0.96,
+      "step": 15077
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9126284416843882e-06,
+      "loss": 0.9657,
+      "step": 15078
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.911996045021667e-06,
+      "loss": 0.8509,
+      "step": 15079
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9113636888135453e-06,
+      "loss": 0.9854,
+      "step": 15080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.910731373072281e-06,
+      "loss": 1.0177,
+      "step": 15081
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9100990978101183e-06,
+      "loss": 0.8755,
+      "step": 15082
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9094668630393163e-06,
+      "loss": 1.0807,
+      "step": 15083
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.908834668772117e-06,
+      "loss": 0.948,
+      "step": 15084
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9082025150207738e-06,
+      "loss": 1.0097,
+      "step": 15085
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9075704017975302e-06,
+      "loss": 0.8394,
+      "step": 15086
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.906938329114637e-06,
+      "loss": 0.8979,
+      "step": 15087
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9063062969843374e-06,
+      "loss": 0.7424,
+      "step": 15088
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.905674305418881e-06,
+      "loss": 0.9391,
+      "step": 15089
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.905042354430507e-06,
+      "loss": 1.1932,
+      "step": 15090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.904410444031464e-06,
+      "loss": 1.0408,
+      "step": 15091
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.903778574233992e-06,
+      "loss": 1.0777,
+      "step": 15092
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9031467450503335e-06,
+      "loss": 1.013,
+      "step": 15093
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9025149564927345e-06,
+      "loss": 0.816,
+      "step": 15094
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.90188320857343e-06,
+      "loss": 0.842,
+      "step": 15095
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.901251501304664e-06,
+      "loss": 0.8045,
+      "step": 15096
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9006198346986726e-06,
+      "loss": 0.9765,
+      "step": 15097
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.899988208767697e-06,
+      "loss": 0.9101,
+      "step": 15098
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8993566235239725e-06,
+      "loss": 1.0331,
+      "step": 15099
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8987250789797384e-06,
+      "loss": 0.9175,
+      "step": 15100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8980935751472274e-06,
+      "loss": 0.8574,
+      "step": 15101
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.897462112038678e-06,
+      "loss": 0.7797,
+      "step": 15102
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8968306896663212e-06,
+      "loss": 0.9556,
+      "step": 15103
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8961993080423956e-06,
+      "loss": 0.8481,
+      "step": 15104
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8955679671791265e-06,
+      "loss": 0.9674,
+      "step": 15105
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.894936667088755e-06,
+      "loss": 0.8162,
+      "step": 15106
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8943054077835057e-06,
+      "loss": 1.0242,
+      "step": 15107
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8936741892756137e-06,
+      "loss": 0.8685,
+      "step": 15108
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8930430115773046e-06,
+      "loss": 0.6595,
+      "step": 15109
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.892411874700811e-06,
+      "loss": 1.0439,
+      "step": 15110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8917807786583576e-06,
+      "loss": 0.9059,
+      "step": 15111
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8911497234621757e-06,
+      "loss": 0.8201,
+      "step": 15112
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.890518709124487e-06,
+      "loss": 0.9862,
+      "step": 15113
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8898877356575218e-06,
+      "loss": 0.8533,
+      "step": 15114
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8892568030735006e-06,
+      "loss": 1.1016,
+      "step": 15115
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8886259113846527e-06,
+      "loss": 0.8702,
+      "step": 15116
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8879950606031963e-06,
+      "loss": 0.9907,
+      "step": 15117
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.887364250741357e-06,
+      "loss": 0.9373,
+      "step": 15118
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.886733481811358e-06,
+      "loss": 0.8206,
+      "step": 15119
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8861027538254166e-06,
+      "loss": 0.8814,
+      "step": 15120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.885472066795757e-06,
+      "loss": 0.9429,
+      "step": 15121
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8848414207345938e-06,
+      "loss": 1.0617,
+      "step": 15122
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8842108156541502e-06,
+      "loss": 0.7903,
+      "step": 15123
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8835802515666407e-06,
+      "loss": 1.131,
+      "step": 15124
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8829497284842855e-06,
+      "loss": 1.0527,
+      "step": 15125
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8823192464192975e-06,
+      "loss": 0.9546,
+      "step": 15126
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.881688805383895e-06,
+      "loss": 0.9514,
+      "step": 15127
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.881058405390288e-06,
+      "loss": 0.9831,
+      "step": 15128
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.880428046450697e-06,
+      "loss": 1.0015,
+      "step": 15129
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8797977285773306e-06,
+      "loss": 1.0622,
+      "step": 15130
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8791674517824036e-06,
+      "loss": 0.8947,
+      "step": 15131
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8785372160781243e-06,
+      "loss": 1.1833,
+      "step": 15132
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.877907021476708e-06,
+      "loss": 1.0501,
+      "step": 15133
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8772768679903586e-06,
+      "loss": 1.008,
+      "step": 15134
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.876646755631291e-06,
+      "loss": 0.8971,
+      "step": 15135
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.876016684411709e-06,
+      "loss": 0.9979,
+      "step": 15136
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8753866543438235e-06,
+      "loss": 0.9226,
+      "step": 15137
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8747566654398378e-06,
+      "loss": 0.876,
+      "step": 15138
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8741267177119612e-06,
+      "loss": 0.7977,
+      "step": 15139
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8734968111723937e-06,
+      "loss": 0.9535,
+      "step": 15140
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8728669458333473e-06,
+      "loss": 0.9118,
+      "step": 15141
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8722371217070193e-06,
+      "loss": 0.8565,
+      "step": 15142
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8716073388056163e-06,
+      "loss": 1.0145,
+      "step": 15143
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.870977597141336e-06,
+      "loss": 1.1413,
+      "step": 15144
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.870347896726384e-06,
+      "loss": 0.8809,
+      "step": 15145
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8697182375729566e-06,
+      "loss": 0.9308,
+      "step": 15146
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.869088619693257e-06,
+      "loss": 0.9135,
+      "step": 15147
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8684590430994797e-06,
+      "loss": 1.073,
+      "step": 15148
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.867829507803828e-06,
+      "loss": 0.8457,
+      "step": 15149
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.867200013818493e-06,
+      "loss": 0.9759,
+      "step": 15150
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.866570561155676e-06,
+      "loss": 0.9212,
+      "step": 15151
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.865941149827568e-06,
+      "loss": 1.1096,
+      "step": 15152
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.865311779846367e-06,
+      "loss": 1.0044,
+      "step": 15153
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8646824512242675e-06,
+      "loss": 0.897,
+      "step": 15154
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8640531639734592e-06,
+      "loss": 1.0363,
+      "step": 15155
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8634239181061384e-06,
+      "loss": 0.8908,
+      "step": 15156
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8627947136344926e-06,
+      "loss": 0.9228,
+      "step": 15157
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.862165550570717e-06,
+      "loss": 0.8494,
+      "step": 15158
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.861536428926996e-06,
+      "loss": 0.9414,
+      "step": 15159
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.860907348715524e-06,
+      "loss": 1.0047,
+      "step": 15160
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.860278309948484e-06,
+      "loss": 0.7826,
+      "step": 15161
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8596493126380698e-06,
+      "loss": 1.0267,
+      "step": 15162
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8590203567964602e-06,
+      "loss": 1.151,
+      "step": 15163
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8583914424358494e-06,
+      "loss": 0.8256,
+      "step": 15164
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8577625695684163e-06,
+      "loss": 1.0801,
+      "step": 15165
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8571337382063503e-06,
+      "loss": 1.1465,
+      "step": 15166
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.856504948361829e-06,
+      "loss": 1.0129,
+      "step": 15167
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.855876200047041e-06,
+      "loss": 1.2027,
+      "step": 15168
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8552474932741626e-06,
+      "loss": 0.6976,
+      "step": 15169
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.85461882805538e-06,
+      "loss": 1.0562,
+      "step": 15170
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.853990204402869e-06,
+      "loss": 1.1817,
+      "step": 15171
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.853361622328812e-06,
+      "loss": 0.8748,
+      "step": 15172
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8527330818453857e-06,
+      "loss": 0.8526,
+      "step": 15173
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.85210458296477e-06,
+      "loss": 0.8879,
+      "step": 15174
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8514761256991374e-06,
+      "loss": 0.9704,
+      "step": 15175
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8508477100606712e-06,
+      "loss": 0.7869,
+      "step": 15176
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8502193360615417e-06,
+      "loss": 0.9398,
+      "step": 15177
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8495910037139262e-06,
+      "loss": 0.9038,
+      "step": 15178
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.848962713029996e-06,
+      "loss": 1.0753,
+      "step": 15179
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8483344640219267e-06,
+      "loss": 0.8395,
+      "step": 15180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.847706256701887e-06,
+      "loss": 1.055,
+      "step": 15181
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.847078091082053e-06,
+      "loss": 0.8808,
+      "step": 15182
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8464499671745892e-06,
+      "loss": 1.0538,
+      "step": 15183
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.845821884991672e-06,
+      "loss": 0.968,
+      "step": 15184
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.845193844545465e-06,
+      "loss": 0.9122,
+      "step": 15185
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8445658458481382e-06,
+      "loss": 1.0929,
+      "step": 15186
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8439378889118617e-06,
+      "loss": 0.9511,
+      "step": 15187
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8433099737487966e-06,
+      "loss": 0.9716,
+      "step": 15188
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8426821003711146e-06,
+      "loss": 0.843,
+      "step": 15189
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8420542687909746e-06,
+      "loss": 1.0654,
+      "step": 15190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8414264790205465e-06,
+      "loss": 0.8693,
+      "step": 15191
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8407987310719886e-06,
+      "loss": 0.9585,
+      "step": 15192
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8401710249574664e-06,
+      "loss": 0.9401,
+      "step": 15193
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.83954336068914e-06,
+      "loss": 1.1785,
+      "step": 15194
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.838915738279172e-06,
+      "loss": 0.8955,
+      "step": 15195
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8382881577397204e-06,
+      "loss": 1.0124,
+      "step": 15196
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8376606190829465e-06,
+      "loss": 0.8729,
+      "step": 15197
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8370331223210037e-06,
+      "loss": 0.9124,
+      "step": 15198
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.836405667466058e-06,
+      "loss": 0.9957,
+      "step": 15199
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8357782545302593e-06,
+      "loss": 0.9902,
+      "step": 15200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.835150883525768e-06,
+      "loss": 0.9482,
+      "step": 15201
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8345235544647355e-06,
+      "loss": 0.9035,
+      "step": 15202
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8338962673593196e-06,
+      "loss": 1.0598,
+      "step": 15203
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8332690222216695e-06,
+      "loss": 0.9299,
+      "step": 15204
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8326418190639438e-06,
+      "loss": 1.1552,
+      "step": 15205
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.832014657898289e-06,
+      "loss": 0.8442,
+      "step": 15206
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.83138753873686e-06,
+      "loss": 0.9668,
+      "step": 15207
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.830760461591804e-06,
+      "loss": 0.9997,
+      "step": 15208
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8301334264752733e-06,
+      "loss": 0.9533,
+      "step": 15209
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.829506433399414e-06,
+      "loss": 0.8249,
+      "step": 15210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8288794823763745e-06,
+      "loss": 0.9368,
+      "step": 15211
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8282525734183054e-06,
+      "loss": 1.0629,
+      "step": 15212
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.827625706537347e-06,
+      "loss": 0.8374,
+      "step": 15213
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8269988817456506e-06,
+      "loss": 0.9176,
+      "step": 15214
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8263720990553556e-06,
+      "loss": 0.9894,
+      "step": 15215
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.82574535847861e-06,
+      "loss": 0.9563,
+      "step": 15216
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8251186600275533e-06,
+      "loss": 0.8876,
+      "step": 15217
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.824492003714331e-06,
+      "loss": 0.833,
+      "step": 15218
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8238653895510804e-06,
+      "loss": 0.8778,
+      "step": 15219
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.823238817549947e-06,
+      "loss": 1.0257,
+      "step": 15220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8226122877230634e-06,
+      "loss": 0.765,
+      "step": 15221
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.821985800082576e-06,
+      "loss": 0.9521,
+      "step": 15222
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.821359354640618e-06,
+      "loss": 0.8143,
+      "step": 15223
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8207329514093305e-06,
+      "loss": 0.8543,
+      "step": 15224
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8201065904008464e-06,
+      "loss": 1.0167,
+      "step": 15225
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.819480271627304e-06,
+      "loss": 0.9678,
+      "step": 15226
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.818853995100835e-06,
+      "loss": 1.065,
+      "step": 15227
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.818227760833576e-06,
+      "loss": 0.882,
+      "step": 15228
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.817601568837659e-06,
+      "loss": 0.8461,
+      "step": 15229
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8169754191252175e-06,
+      "loss": 1.0062,
+      "step": 15230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8163493117083795e-06,
+      "loss": 0.9971,
+      "step": 15231
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.815723246599281e-06,
+      "loss": 0.9533,
+      "step": 15232
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8150972238100453e-06,
+      "loss": 0.9207,
+      "step": 15233
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8144712433528094e-06,
+      "loss": 0.9076,
+      "step": 15234
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8138453052396946e-06,
+      "loss": 1.0053,
+      "step": 15235
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8132194094828334e-06,
+      "loss": 0.8857,
+      "step": 15236
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8125935560943483e-06,
+      "loss": 1.0092,
+      "step": 15237
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.811967745086369e-06,
+      "loss": 0.9095,
+      "step": 15238
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.811341976471016e-06,
+      "loss": 1.1428,
+      "step": 15239
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.810716250260418e-06,
+      "loss": 1.0505,
+      "step": 15240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.810090566466694e-06,
+      "loss": 0.906,
+      "step": 15241
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.80946492510197e-06,
+      "loss": 0.908,
+      "step": 15242
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8088393261783646e-06,
+      "loss": 0.7515,
+      "step": 15243
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8082137697080005e-06,
+      "loss": 1.0809,
+      "step": 15244
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8075882557029987e-06,
+      "loss": 0.8856,
+      "step": 15245
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.806962784175476e-06,
+      "loss": 0.9135,
+      "step": 15246
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8063373551375533e-06,
+      "loss": 0.8749,
+      "step": 15247
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8057119686013446e-06,
+      "loss": 0.7608,
+      "step": 15248
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805086624578971e-06,
+      "loss": 1.1483,
+      "step": 15249
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8044613230825446e-06,
+      "loss": 0.9588,
+      "step": 15250
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8038360641241836e-06,
+      "loss": 0.9263,
+      "step": 15251
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8032108477159984e-06,
+      "loss": 1.1374,
+      "step": 15252
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8025856738701075e-06,
+      "loss": 0.8618,
+      "step": 15253
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.801960542598618e-06,
+      "loss": 0.9163,
+      "step": 15254
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8013354539136465e-06,
+      "loss": 1.0655,
+      "step": 15255
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8007104078272982e-06,
+      "loss": 0.7606,
+      "step": 15256
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8000854043516907e-06,
+      "loss": 0.8427,
+      "step": 15257
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7994604434989268e-06,
+      "loss": 0.9168,
+      "step": 15258
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7988355252811207e-06,
+      "loss": 1.0349,
+      "step": 15259
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7982106497103745e-06,
+      "loss": 1.031,
+      "step": 15260
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7975858167987994e-06,
+      "loss": 0.9249,
+      "step": 15261
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.796961026558498e-06,
+      "loss": 0.8163,
+      "step": 15262
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7963362790015787e-06,
+      "loss": 0.8156,
+      "step": 15263
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.795711574140142e-06,
+      "loss": 1.0413,
+      "step": 15264
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7950869119862955e-06,
+      "loss": 1.1178,
+      "step": 15265
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.794462292552138e-06,
+      "loss": 0.8418,
+      "step": 15266
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.793837715849775e-06,
+      "loss": 1.0793,
+      "step": 15267
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7932131818913023e-06,
+      "loss": 1.0039,
+      "step": 15268
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.792588690688828e-06,
+      "loss": 0.9143,
+      "step": 15269
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7919642422544443e-06,
+      "loss": 0.9939,
+      "step": 15270
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.791339836600254e-06,
+      "loss": 0.9568,
+      "step": 15271
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.790715473738352e-06,
+      "loss": 1.0412,
+      "step": 15272
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7900911536808393e-06,
+      "loss": 0.9664,
+      "step": 15273
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7894668764398065e-06,
+      "loss": 0.9102,
+      "step": 15274
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7888426420273534e-06,
+      "loss": 1.1019,
+      "step": 15275
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.788218450455571e-06,
+      "loss": 0.9887,
+      "step": 15276
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.787594301736556e-06,
+      "loss": 0.9161,
+      "step": 15277
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.786970195882398e-06,
+      "loss": 0.9782,
+      "step": 15278
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7863461329051898e-06,
+      "loss": 0.9507,
+      "step": 15279
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.785722112817025e-06,
+      "loss": 0.8688,
+      "step": 15280
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7850981356299913e-06,
+      "loss": 0.9724,
+      "step": 15281
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.78447420135618e-06,
+      "loss": 1.1804,
+      "step": 15282
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7838503100076763e-06,
+      "loss": 0.8807,
+      "step": 15283
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.783226461596572e-06,
+      "loss": 0.8683,
+      "step": 15284
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7826026561349507e-06,
+      "loss": 0.9672,
+      "step": 15285
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7819788936349014e-06,
+      "loss": 0.96,
+      "step": 15286
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7813551741085066e-06,
+      "loss": 0.9343,
+      "step": 15287
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7807314975678536e-06,
+      "loss": 0.9295,
+      "step": 15288
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7801078640250222e-06,
+      "loss": 0.9387,
+      "step": 15289
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7794842734920997e-06,
+      "loss": 0.7122,
+      "step": 15290
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.778860725981162e-06,
+      "loss": 0.8129,
+      "step": 15291
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7782372215042973e-06,
+      "loss": 0.8914,
+      "step": 15292
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7776137600735803e-06,
+      "loss": 0.8873,
+      "step": 15293
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7769903417010954e-06,
+      "loss": 0.9191,
+      "step": 15294
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7763669663989157e-06,
+      "loss": 0.924,
+      "step": 15295
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7757436341791234e-06,
+      "loss": 1.012,
+      "step": 15296
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7751203450537924e-06,
+      "loss": 1.0189,
+      "step": 15297
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7744970990350016e-06,
+      "loss": 1.1364,
+      "step": 15298
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.773873896134823e-06,
+      "loss": 0.8683,
+      "step": 15299
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7732507363653344e-06,
+      "loss": 0.9334,
+      "step": 15300
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.772627619738606e-06,
+      "loss": 1.0859,
+      "step": 15301
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7720045462667134e-06,
+      "loss": 0.8649,
+      "step": 15302
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7713815159617265e-06,
+      "loss": 1.0401,
+      "step": 15303
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7707585288357163e-06,
+      "loss": 0.9704,
+      "step": 15304
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.770135584900756e-06,
+      "loss": 1.0197,
+      "step": 15305
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7695126841689113e-06,
+      "loss": 0.9064,
+      "step": 15306
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7688898266522548e-06,
+      "loss": 0.9768,
+      "step": 15307
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7682670123628486e-06,
+      "loss": 0.8718,
+      "step": 15308
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.767644241312766e-06,
+      "loss": 0.9262,
+      "step": 15309
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7670215135140664e-06,
+      "loss": 0.9053,
+      "step": 15310
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7663988289788214e-06,
+      "loss": 0.9496,
+      "step": 15311
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7657761877190903e-06,
+      "loss": 0.9561,
+      "step": 15312
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.765153589746941e-06,
+      "loss": 0.8609,
+      "step": 15313
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7645310350744296e-06,
+      "loss": 0.975,
+      "step": 15314
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7639085237136265e-06,
+      "loss": 1.0253,
+      "step": 15315
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.763286055676587e-06,
+      "loss": 1.0059,
+      "step": 15316
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.762663630975374e-06,
+      "loss": 0.8359,
+      "step": 15317
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.762041249622044e-06,
+      "loss": 1.0188,
+      "step": 15318
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.761418911628659e-06,
+      "loss": 1.0187,
+      "step": 15319
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7607966170072733e-06,
+      "loss": 0.9643,
+      "step": 15320
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7601743657699465e-06,
+      "loss": 0.9994,
+      "step": 15321
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.759552157928731e-06,
+      "loss": 0.9932,
+      "step": 15322
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.758929993495687e-06,
+      "loss": 1.0376,
+      "step": 15323
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.758307872482864e-06,
+      "loss": 0.8504,
+      "step": 15324
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.757685794902319e-06,
+      "loss": 0.7105,
+      "step": 15325
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.757063760766099e-06,
+      "loss": 0.9688,
+      "step": 15326
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7564417700862645e-06,
+      "loss": 0.9964,
+      "step": 15327
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7558198228748593e-06,
+      "loss": 0.8851,
+      "step": 15328
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.755197919143938e-06,
+      "loss": 0.9575,
+      "step": 15329
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.754576058905546e-06,
+      "loss": 0.9215,
+      "step": 15330
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.753954242171736e-06,
+      "loss": 0.9977,
+      "step": 15331
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7533324689545502e-06,
+      "loss": 0.9523,
+      "step": 15332
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7527107392660404e-06,
+      "loss": 0.7921,
+      "step": 15333
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7520890531182486e-06,
+      "loss": 0.6926,
+      "step": 15334
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7514674105232232e-06,
+      "loss": 1.108,
+      "step": 15335
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7508458114930048e-06,
+      "loss": 0.9159,
+      "step": 15336
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.750224256039638e-06,
+      "loss": 0.9006,
+      "step": 15337
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7496027441751682e-06,
+      "loss": 0.9326,
+      "step": 15338
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.748981275911633e-06,
+      "loss": 0.9126,
+      "step": 15339
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7483598512610775e-06,
+      "loss": 1.0473,
+      "step": 15340
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7477384702355365e-06,
+      "loss": 0.917,
+      "step": 15341
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7471171328470535e-06,
+      "loss": 0.9403,
+      "step": 15342
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7464958391076635e-06,
+      "loss": 0.9282,
+      "step": 15343
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.745874589029407e-06,
+      "loss": 0.9408,
+      "step": 15344
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.745253382624318e-06,
+      "loss": 1.1,
+      "step": 15345
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.744632219904434e-06,
+      "loss": 0.9009,
+      "step": 15346
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7440111008817875e-06,
+      "loss": 0.914,
+      "step": 15347
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7433900255684167e-06,
+      "loss": 0.8432,
+      "step": 15348
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.742768993976348e-06,
+      "loss": 0.9497,
+      "step": 15349
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.742148006117622e-06,
+      "loss": 0.8549,
+      "step": 15350
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.741527062004264e-06,
+      "loss": 0.8567,
+      "step": 15351
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.740906161648309e-06,
+      "loss": 0.8297,
+      "step": 15352
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.740285305061783e-06,
+      "loss": 0.9857,
+      "step": 15353
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7396644922567176e-06,
+      "loss": 0.9896,
+      "step": 15354
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7390437232451384e-06,
+      "loss": 0.8155,
+      "step": 15355
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7384229980390763e-06,
+      "loss": 0.9321,
+      "step": 15356
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.737802316650554e-06,
+      "loss": 1.063,
+      "step": 15357
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7371816790916006e-06,
+      "loss": 0.8819,
+      "step": 15358
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7365610853742365e-06,
+      "loss": 1.0553,
+      "step": 15359
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.73594053551049e-06,
+      "loss": 0.8996,
+      "step": 15360
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7353200295123804e-06,
+      "loss": 0.8541,
+      "step": 15361
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7346995673919315e-06,
+      "loss": 0.9106,
+      "step": 15362
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.734079149161164e-06,
+      "loss": 0.8173,
+      "step": 15363
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7334587748321018e-06,
+      "loss": 0.9453,
+      "step": 15364
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.732838444416759e-06,
+      "loss": 0.7448,
+      "step": 15365
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.732218157927159e-06,
+      "loss": 1.0854,
+      "step": 15366
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7315979153753165e-06,
+      "loss": 1.0164,
+      "step": 15367
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.730977716773251e-06,
+      "loss": 0.9089,
+      "step": 15368
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7303575621329763e-06,
+      "loss": 0.7932,
+      "step": 15369
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.729737451466511e-06,
+      "loss": 1.0552,
+      "step": 15370
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7291173847858647e-06,
+      "loss": 0.8195,
+      "step": 15371
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.728497362103055e-06,
+      "loss": 0.9716,
+      "step": 15372
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.727877383430095e-06,
+      "loss": 1.0002,
+      "step": 15373
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7272574487789933e-06,
+      "loss": 0.9896,
+      "step": 15374
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7266375581617655e-06,
+      "loss": 0.9361,
+      "step": 15375
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7260177115904165e-06,
+      "loss": 1.0263,
+      "step": 15376
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.725397909076961e-06,
+      "loss": 0.9251,
+      "step": 15377
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.724778150633403e-06,
+      "loss": 0.9866,
+      "step": 15378
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7241584362717545e-06,
+      "loss": 0.9161,
+      "step": 15379
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.723538766004018e-06,
+      "loss": 0.8546,
+      "step": 15380
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7229191398422026e-06,
+      "loss": 0.9684,
+      "step": 15381
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7222995577983113e-06,
+      "loss": 0.9817,
+      "step": 15382
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7216800198843513e-06,
+      "loss": 0.906,
+      "step": 15383
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7210605261123202e-06,
+      "loss": 0.835,
+      "step": 15384
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.720441076494228e-06,
+      "loss": 0.8293,
+      "step": 15385
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.719821671042071e-06,
+      "loss": 0.8627,
+      "step": 15386
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.719202309767854e-06,
+      "loss": 0.9479,
+      "step": 15387
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7185829926835728e-06,
+      "loss": 1.0554,
+      "step": 15388
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7179637198012308e-06,
+      "loss": 0.8438,
+      "step": 15389
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.717344491132822e-06,
+      "loss": 0.8541,
+      "step": 15390
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.716725306690348e-06,
+      "loss": 0.9943,
+      "step": 15391
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7161061664858003e-06,
+      "loss": 0.8811,
+      "step": 15392
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7154870705311807e-06,
+      "loss": 0.8643,
+      "step": 15393
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7148680188384786e-06,
+      "loss": 0.9267,
+      "step": 15394
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.714249011419692e-06,
+      "loss": 0.8153,
+      "step": 15395
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7136300482868103e-06,
+      "loss": 0.8483,
+      "step": 15396
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.713011129451828e-06,
+      "loss": 0.961,
+      "step": 15397
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7123922549267385e-06,
+      "loss": 0.9048,
+      "step": 15398
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7117734247235277e-06,
+      "loss": 0.8972,
+      "step": 15399
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.71115463885419e-06,
+      "loss": 1.0702,
+      "step": 15400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.71053589733071e-06,
+      "loss": 1.0616,
+      "step": 15401
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7099172001650797e-06,
+      "loss": 0.8612,
+      "step": 15402
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7092985473692818e-06,
+      "loss": 0.91,
+      "step": 15403
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7086799389553078e-06,
+      "loss": 1.0041,
+      "step": 15404
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.708061374935137e-06,
+      "loss": 1.1451,
+      "step": 15405
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7074428553207598e-06,
+      "loss": 1.2088,
+      "step": 15406
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7068243801241534e-06,
+      "loss": 0.9209,
+      "step": 15407
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.706205949357308e-06,
+      "loss": 0.9109,
+      "step": 15408
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7055875630322002e-06,
+      "loss": 1.0493,
+      "step": 15409
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7049692211608147e-06,
+      "loss": 0.8728,
+      "step": 15410
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7043509237551284e-06,
+      "loss": 0.8592,
+      "step": 15411
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.703732670827124e-06,
+      "loss": 0.9594,
+      "step": 15412
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.703114462388776e-06,
+      "loss": 1.058,
+      "step": 15413
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.702496298452067e-06,
+      "loss": 0.9027,
+      "step": 15414
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.701878179028968e-06,
+      "loss": 0.8865,
+      "step": 15415
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.701260104131461e-06,
+      "loss": 0.9677,
+      "step": 15416
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7006420737715154e-06,
+      "loss": 0.8925,
+      "step": 15417
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7000240879611107e-06,
+      "loss": 0.9854,
+      "step": 15418
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.699406146712213e-06,
+      "loss": 0.9698,
+      "step": 15419
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6987882500368034e-06,
+      "loss": 0.9039,
+      "step": 15420
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6981703979468477e-06,
+      "loss": 0.794,
+      "step": 15421
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6975525904543197e-06,
+      "loss": 0.9168,
+      "step": 15422
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.696934827571187e-06,
+      "loss": 1.0157,
+      "step": 15423
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6963171093094208e-06,
+      "loss": 0.931,
+      "step": 15424
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.695699435680986e-06,
+      "loss": 0.9922,
+      "step": 15425
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6950818066978545e-06,
+      "loss": 1.0469,
+      "step": 15426
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.694464222371988e-06,
+      "loss": 0.946,
+      "step": 15427
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.693846682715356e-06,
+      "loss": 0.9767,
+      "step": 15428
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6932291877399197e-06,
+      "loss": 1.0444,
+      "step": 15429
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6926117374576442e-06,
+      "loss": 0.8339,
+      "step": 15430
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6919943318804955e-06,
+      "loss": 0.8788,
+      "step": 15431
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6913769710204306e-06,
+      "loss": 0.8851,
+      "step": 15432
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6907596548894148e-06,
+      "loss": 0.9739,
+      "step": 15433
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6901423834994055e-06,
+      "loss": 0.8953,
+      "step": 15434
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6895251568623655e-06,
+      "loss": 0.9,
+      "step": 15435
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.688907974990249e-06,
+      "loss": 0.9053,
+      "step": 15436
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.688290837895019e-06,
+      "loss": 0.9554,
+      "step": 15437
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.687673745588627e-06,
+      "loss": 0.973,
+      "step": 15438
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.687056698083034e-06,
+      "loss": 0.9434,
+      "step": 15439
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6864396953901907e-06,
+      "loss": 1.166,
+      "step": 15440
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6858227375220556e-06,
+      "loss": 1.0507,
+      "step": 15441
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6852058244905766e-06,
+      "loss": 0.9577,
+      "step": 15442
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6845889563077125e-06,
+      "loss": 0.8992,
+      "step": 15443
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6839721329854108e-06,
+      "loss": 1.0415,
+      "step": 15444
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6833553545356256e-06,
+      "loss": 0.8349,
+      "step": 15445
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6827386209703026e-06,
+      "loss": 0.8986,
+      "step": 15446
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.682121932301396e-06,
+      "loss": 0.9336,
+      "step": 15447
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6815052885408477e-06,
+      "loss": 0.8759,
+      "step": 15448
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6808886897006114e-06,
+      "loss": 0.9359,
+      "step": 15449
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6802721357926287e-06,
+      "loss": 1.032,
+      "step": 15450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.67965562682885e-06,
+      "loss": 0.8975,
+      "step": 15451
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6790391628212138e-06,
+      "loss": 1.0003,
+      "step": 15452
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.67842274378167e-06,
+      "loss": 0.8713,
+      "step": 15453
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6778063697221567e-06,
+      "loss": 0.923,
+      "step": 15454
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6771900406546182e-06,
+      "loss": 0.9103,
+      "step": 15455
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6765737565909978e-06,
+      "loss": 1.0484,
+      "step": 15456
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6759575175432317e-06,
+      "loss": 0.8316,
+      "step": 15457
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.675341323523264e-06,
+      "loss": 0.9393,
+      "step": 15458
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6747251745430286e-06,
+      "loss": 0.9132,
+      "step": 15459
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6741090706144674e-06,
+      "loss": 0.8601,
+      "step": 15460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6734930117495133e-06,
+      "loss": 1.0354,
+      "step": 15461
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6728769979601044e-06,
+      "loss": 0.8404,
+      "step": 15462
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.672261029258178e-06,
+      "loss": 1.0701,
+      "step": 15463
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6716451056556636e-06,
+      "loss": 1.093,
+      "step": 15464
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6710292271644978e-06,
+      "loss": 0.9798,
+      "step": 15465
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6704133937966142e-06,
+      "loss": 0.8689,
+      "step": 15466
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.669797605563941e-06,
+      "loss": 0.9489,
+      "step": 15467
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6691818624784117e-06,
+      "loss": 0.9167,
+      "step": 15468
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6685661645519536e-06,
+      "loss": 1.0615,
+      "step": 15469
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.667950511796499e-06,
+      "loss": 0.8715,
+      "step": 15470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6673349042239727e-06,
+      "loss": 0.943,
+      "step": 15471
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.666719341846305e-06,
+      "loss": 0.9829,
+      "step": 15472
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.666103824675419e-06,
+      "loss": 0.9062,
+      "step": 15473
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.665488352723244e-06,
+      "loss": 0.8836,
+      "step": 15474
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6648729260017004e-06,
+      "loss": 1.0142,
+      "step": 15475
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6642575445227165e-06,
+      "loss": 1.0464,
+      "step": 15476
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.663642208298209e-06,
+      "loss": 0.8905,
+      "step": 15477
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.663026917340107e-06,
+      "loss": 0.9373,
+      "step": 15478
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.662411671660327e-06,
+      "loss": 0.8511,
+      "step": 15479
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6617964712707923e-06,
+      "loss": 1.0224,
+      "step": 15480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6611813161834187e-06,
+      "loss": 0.9237,
+      "step": 15481
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.660566206410129e-06,
+      "loss": 0.9493,
+      "step": 15482
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6599511419628355e-06,
+      "loss": 0.8585,
+      "step": 15483
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6593361228534603e-06,
+      "loss": 1.1229,
+      "step": 15484
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.658721149093915e-06,
+      "loss": 1.0763,
+      "step": 15485
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6581062206961172e-06,
+      "loss": 0.9654,
+      "step": 15486
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.657491337671979e-06,
+      "loss": 0.7639,
+      "step": 15487
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6568765000334174e-06,
+      "loss": 1.1007,
+      "step": 15488
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.656261707792339e-06,
+      "loss": 0.9543,
+      "step": 15489
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6556469609606596e-06,
+      "loss": 0.9286,
+      "step": 15490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.655032259550289e-06,
+      "loss": 0.8997,
+      "step": 15491
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.654417603573135e-06,
+      "loss": 0.9644,
+      "step": 15492
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6538029930411104e-06,
+      "loss": 1.0087,
+      "step": 15493
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.653188427966118e-06,
+      "loss": 0.7608,
+      "step": 15494
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.65257390836007e-06,
+      "loss": 0.9438,
+      "step": 15495
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6519594342348676e-06,
+      "loss": 0.9231,
+      "step": 15496
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.651345005602421e-06,
+      "loss": 0.9652,
+      "step": 15497
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.65073062247463e-06,
+      "loss": 1.1366,
+      "step": 15498
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6501162848634023e-06,
+      "loss": 1.0487,
+      "step": 15499
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6495019927806343e-06,
+      "loss": 0.9874,
+      "step": 15500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6488877462382356e-06,
+      "loss": 0.9304,
+      "step": 15501
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6482735452481025e-06,
+      "loss": 0.9599,
+      "step": 15502
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.647659389822137e-06,
+      "loss": 0.9782,
+      "step": 15503
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6470452799722357e-06,
+      "loss": 0.8385,
+      "step": 15504
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6464312157103007e-06,
+      "loss": 0.9569,
+      "step": 15505
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6458171970482247e-06,
+      "loss": 0.7268,
+      "step": 15506
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6452032239979086e-06,
+      "loss": 0.9529,
+      "step": 15507
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.644589296571244e-06,
+      "loss": 1.0136,
+      "step": 15508
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6439754147801288e-06,
+      "loss": 1.1897,
+      "step": 15509
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6433615786364543e-06,
+      "loss": 0.7027,
+      "step": 15510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.642747788152116e-06,
+      "loss": 0.8941,
+      "step": 15511
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.642134043339001e-06,
+      "loss": 0.8162,
+      "step": 15512
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6415203442090076e-06,
+      "loss": 0.9563,
+      "step": 15513
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.640906690774021e-06,
+      "loss": 0.8929,
+      "step": 15514
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6402930830459337e-06,
+      "loss": 0.9131,
+      "step": 15515
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6396795210366313e-06,
+      "loss": 0.8715,
+      "step": 15516
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.639066004758004e-06,
+      "loss": 1.1402,
+      "step": 15517
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.638452534221936e-06,
+      "loss": 0.9209,
+      "step": 15518
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.637839109440316e-06,
+      "loss": 0.9038,
+      "step": 15519
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6372257304250258e-06,
+      "loss": 0.9172,
+      "step": 15520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.636612397187953e-06,
+      "loss": 1.003,
+      "step": 15521
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6359991097409767e-06,
+      "loss": 0.9314,
+      "step": 15522
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6353858680959813e-06,
+      "loss": 0.9853,
+      "step": 15523
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6347726722648504e-06,
+      "loss": 0.8499,
+      "step": 15524
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6341595222594606e-06,
+      "loss": 0.7873,
+      "step": 15525
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.633546418091696e-06,
+      "loss": 0.9583,
+      "step": 15526
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.63293335977343e-06,
+      "loss": 0.9461,
+      "step": 15527
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6323203473165447e-06,
+      "loss": 0.9643,
+      "step": 15528
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6317073807329146e-06,
+      "loss": 0.822,
+      "step": 15529
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6310944600344187e-06,
+      "loss": 0.8807,
+      "step": 15530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6304815852329278e-06,
+      "loss": 0.9675,
+      "step": 15531
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6298687563403207e-06,
+      "loss": 0.803,
+      "step": 15532
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6292559733684676e-06,
+      "loss": 0.9511,
+      "step": 15533
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6286432363292437e-06,
+      "loss": 0.893,
+      "step": 15534
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6280305452345144e-06,
+      "loss": 0.9449,
+      "step": 15535
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6274179000961604e-06,
+      "loss": 0.9543,
+      "step": 15536
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6268053009260443e-06,
+      "loss": 0.8728,
+      "step": 15537
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.626192747736038e-06,
+      "loss": 0.8216,
+      "step": 15538
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.625580240538008e-06,
+      "loss": 0.8887,
+      "step": 15539
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6249677793438233e-06,
+      "loss": 0.9801,
+      "step": 15540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.624355364165348e-06,
+      "loss": 0.9399,
+      "step": 15541
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.62374299501445e-06,
+      "loss": 1.0356,
+      "step": 15542
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.623130671902991e-06,
+      "loss": 1.0266,
+      "step": 15543
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.622518394842838e-06,
+      "loss": 0.9531,
+      "step": 15544
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6219061638458495e-06,
+      "loss": 0.9103,
+      "step": 15545
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.621293978923891e-06,
+      "loss": 0.9566,
+      "step": 15546
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.620681840088821e-06,
+      "loss": 0.9587,
+      "step": 15547
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6200697473525006e-06,
+      "loss": 0.8583,
+      "step": 15548
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6194577007267906e-06,
+      "loss": 0.9634,
+      "step": 15549
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6188457002235455e-06,
+      "loss": 0.8059,
+      "step": 15550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6182337458546275e-06,
+      "loss": 0.9224,
+      "step": 15551
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6176218376318883e-06,
+      "loss": 0.8675,
+      "step": 15552
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.617009975567187e-06,
+      "loss": 0.8955,
+      "step": 15553
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6163981596723754e-06,
+      "loss": 0.685,
+      "step": 15554
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6157863899593106e-06,
+      "loss": 0.9781,
+      "step": 15555
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6151746664398415e-06,
+      "loss": 0.9806,
+      "step": 15556
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6145629891258244e-06,
+      "loss": 0.8962,
+      "step": 15557
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6139513580291043e-06,
+      "loss": 0.9001,
+      "step": 15558
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6133397731615396e-06,
+      "loss": 0.8386,
+      "step": 15559
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.612728234534973e-06,
+      "loss": 1.0371,
+      "step": 15560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6121167421612576e-06,
+      "loss": 0.7505,
+      "step": 15561
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.611505296052237e-06,
+      "loss": 0.9376,
+      "step": 15562
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6108938962197603e-06,
+      "loss": 0.9688,
+      "step": 15563
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6102825426756716e-06,
+      "loss": 0.7916,
+      "step": 15564
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6096712354318175e-06,
+      "loss": 0.8601,
+      "step": 15565
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.60905997450004e-06,
+      "loss": 1.0289,
+      "step": 15566
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6084487598921847e-06,
+      "loss": 1.0177,
+      "step": 15567
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6078375916200903e-06,
+      "loss": 0.9833,
+      "step": 15568
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6072264696956017e-06,
+      "loss": 0.9929,
+      "step": 15569
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.606615394130554e-06,
+      "loss": 0.9182,
+      "step": 15570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6060043649367937e-06,
+      "loss": 1.0152,
+      "step": 15571
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.605393382126154e-06,
+      "loss": 0.9367,
+      "step": 15572
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6047824457104766e-06,
+      "loss": 0.8123,
+      "step": 15573
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6041715557015934e-06,
+      "loss": 1.2452,
+      "step": 15574
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6035607121113447e-06,
+      "loss": 0.8552,
+      "step": 15575
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6029499149515624e-06,
+      "loss": 0.9246,
+      "step": 15576
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6023391642340833e-06,
+      "loss": 1.0608,
+      "step": 15577
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.601728459970737e-06,
+      "loss": 0.9327,
+      "step": 15578
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6011178021733596e-06,
+      "loss": 0.9564,
+      "step": 15579
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6005071908537787e-06,
+      "loss": 0.9529,
+      "step": 15580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.599896626023829e-06,
+      "loss": 0.7761,
+      "step": 15581
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5992861076953356e-06,
+      "loss": 0.9178,
+      "step": 15582
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.598675635880129e-06,
+      "loss": 0.9237,
+      "step": 15583
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5980652105900393e-06,
+      "loss": 1.1103,
+      "step": 15584
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5974548318368896e-06,
+      "loss": 0.9427,
+      "step": 15585
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.596844499632509e-06,
+      "loss": 1.0727,
+      "step": 15586
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5962342139887197e-06,
+      "loss": 0.9164,
+      "step": 15587
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5956239749173495e-06,
+      "loss": 0.8594,
+      "step": 15588
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.595013782430217e-06,
+      "loss": 1.0583,
+      "step": 15589
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5944036365391485e-06,
+      "loss": 0.8894,
+      "step": 15590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.593793537255962e-06,
+      "loss": 0.9454,
+      "step": 15591
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.593183484592482e-06,
+      "loss": 1.0552,
+      "step": 15592
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.592573478560523e-06,
+      "loss": 0.9198,
+      "step": 15593
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5919635191719094e-06,
+      "loss": 0.822,
+      "step": 15594
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5913536064384547e-06,
+      "loss": 1.0612,
+      "step": 15595
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5907437403719795e-06,
+      "loss": 0.8691,
+      "step": 15596
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.590133920984296e-06,
+      "loss": 1.0546,
+      "step": 15597
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5895241482872237e-06,
+      "loss": 0.8543,
+      "step": 15598
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.588914422292571e-06,
+      "loss": 0.9617,
+      "step": 15599
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5883047430121577e-06,
+      "loss": 0.864,
+      "step": 15600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5876951104577896e-06,
+      "loss": 1.079,
+      "step": 15601
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5870855246412846e-06,
+      "loss": 0.9915,
+      "step": 15602
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.586475985574448e-06,
+      "loss": 1.1845,
+      "step": 15603
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5858664932690935e-06,
+      "loss": 0.9743,
+      "step": 15604
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5852570477370243e-06,
+      "loss": 1.0808,
+      "step": 15605
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5846476489900556e-06,
+      "loss": 0.8024,
+      "step": 15606
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.584038297039989e-06,
+      "loss": 0.8071,
+      "step": 15607
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.583428991898635e-06,
+      "loss": 1.1873,
+      "step": 15608
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5828197335777938e-06,
+      "loss": 1.0028,
+      "step": 15609
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5822105220892733e-06,
+      "loss": 0.8554,
+      "step": 15610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.581601357444874e-06,
+      "loss": 0.9852,
+      "step": 15611
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5809922396564012e-06,
+      "loss": 1.0603,
+      "step": 15612
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.580383168735654e-06,
+      "loss": 0.9504,
+      "step": 15613
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.579774144694436e-06,
+      "loss": 0.9128,
+      "step": 15614
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5791651675445424e-06,
+      "loss": 0.7937,
+      "step": 15615
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5785562372977746e-06,
+      "loss": 0.7519,
+      "step": 15616
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5779473539659327e-06,
+      "loss": 0.7955,
+      "step": 15617
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.577338517560809e-06,
+      "loss": 0.9596,
+      "step": 15618
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5767297280942046e-06,
+      "loss": 0.8469,
+      "step": 15619
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.576120985577909e-06,
+      "loss": 0.9133,
+      "step": 15620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.575512290023723e-06,
+      "loss": 0.8477,
+      "step": 15621
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5749036414434333e-06,
+      "loss": 0.9274,
+      "step": 15622
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5742950398488375e-06,
+      "loss": 0.9249,
+      "step": 15623
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.573686485251723e-06,
+      "loss": 1.0562,
+      "step": 15624
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.573077977663885e-06,
+      "loss": 1.0448,
+      "step": 15625
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5724695170971083e-06,
+      "loss": 0.9404,
+      "step": 15626
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5718611035631858e-06,
+      "loss": 0.899,
+      "step": 15627
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5712527370739005e-06,
+      "loss": 1.0343,
+      "step": 15628
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.570644417641046e-06,
+      "loss": 0.9571,
+      "step": 15629
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.570036145276402e-06,
+      "loss": 1.055,
+      "step": 15630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.569427919991759e-06,
+      "loss": 0.8489,
+      "step": 15631
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.568819741798898e-06,
+      "loss": 1.0873,
+      "step": 15632
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5682116107096033e-06,
+      "loss": 0.8883,
+      "step": 15633
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.567603526735656e-06,
+      "loss": 0.9048,
+      "step": 15634
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5669954898888406e-06,
+      "loss": 0.912,
+      "step": 15635
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5663875001809335e-06,
+      "loss": 0.9205,
+      "step": 15636
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.565779557623718e-06,
+      "loss": 0.8496,
+      "step": 15637
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5651716622289703e-06,
+      "loss": 0.9946,
+      "step": 15638
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5645638140084705e-06,
+      "loss": 0.8358,
+      "step": 15639
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5639560129739926e-06,
+      "loss": 0.9643,
+      "step": 15640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.563348259137315e-06,
+      "loss": 0.9015,
+      "step": 15641
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5627405525102134e-06,
+      "loss": 0.8891,
+      "step": 15642
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.562132893104459e-06,
+      "loss": 0.9471,
+      "step": 15643
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5615252809318287e-06,
+      "loss": 0.8999,
+      "step": 15644
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.560917716004091e-06,
+      "loss": 0.9323,
+      "step": 15645
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5603101983330214e-06,
+      "loss": 1.0387,
+      "step": 15646
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.559702727930386e-06,
+      "loss": 0.9932,
+      "step": 15647
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.559095304807959e-06,
+      "loss": 0.9246,
+      "step": 15648
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5584879289775044e-06,
+      "loss": 0.8167,
+      "step": 15649
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.557880600450795e-06,
+      "loss": 0.9592,
+      "step": 15650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5572733192395904e-06,
+      "loss": 0.7914,
+      "step": 15651
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.556666085355666e-06,
+      "loss": 0.7446,
+      "step": 15652
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.556058898810779e-06,
+      "loss": 0.8886,
+      "step": 15653
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5554517596166984e-06,
+      "loss": 0.932,
+      "step": 15654
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5548446677851835e-06,
+      "loss": 0.8347,
+      "step": 15655
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.554237623328001e-06,
+      "loss": 0.9592,
+      "step": 15656
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5536306262569076e-06,
+      "loss": 0.9639,
+      "step": 15657
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.553023676583668e-06,
+      "loss": 0.9509,
+      "step": 15658
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5524167743200374e-06,
+      "loss": 0.8254,
+      "step": 15659
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5518099194777786e-06,
+      "loss": 1.0285,
+      "step": 15660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5512031120686453e-06,
+      "loss": 0.7234,
+      "step": 15661
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5505963521043986e-06,
+      "loss": 0.9675,
+      "step": 15662
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.549989639596787e-06,
+      "loss": 1.0185,
+      "step": 15663
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.549382974557575e-06,
+      "loss": 0.9333,
+      "step": 15664
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5487763569985102e-06,
+      "loss": 0.9843,
+      "step": 15665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5481697869313495e-06,
+      "loss": 1.1271,
+      "step": 15666
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5475632643678407e-06,
+      "loss": 0.9375,
+      "step": 15667
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.546956789319739e-06,
+      "loss": 0.885,
+      "step": 15668
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5463503617987916e-06,
+      "loss": 1.0313,
+      "step": 15669
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5457439818167505e-06,
+      "loss": 0.8392,
+      "step": 15670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.545137649385362e-06,
+      "loss": 1.0852,
+      "step": 15671
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5445313645163765e-06,
+      "loss": 1.0372,
+      "step": 15672
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.543925127221537e-06,
+      "loss": 0.8682,
+      "step": 15673
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5433189375125907e-06,
+      "loss": 0.9862,
+      "step": 15674
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.542712795401284e-06,
+      "loss": 0.9277,
+      "step": 15675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5421067008993588e-06,
+      "loss": 0.8751,
+      "step": 15676
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5415006540185604e-06,
+      "loss": 0.9067,
+      "step": 15677
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.540894654770627e-06,
+      "loss": 0.9094,
+      "step": 15678
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5402887031673037e-06,
+      "loss": 0.9356,
+      "step": 15679
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5396827992203264e-06,
+      "loss": 0.7673,
+      "step": 15680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.53907694294144e-06,
+      "loss": 0.9552,
+      "step": 15681
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5384711343423763e-06,
+      "loss": 0.9282,
+      "step": 15682
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.537865373434879e-06,
+      "loss": 0.9357,
+      "step": 15683
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.537259660230679e-06,
+      "loss": 0.8813,
+      "step": 15684
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5366539947415163e-06,
+      "loss": 0.9542,
+      "step": 15685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5360483769791198e-06,
+      "loss": 0.8683,
+      "step": 15686
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5354428069552316e-06,
+      "loss": 1.1097,
+      "step": 15687
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5348372846815773e-06,
+      "loss": 1.0529,
+      "step": 15688
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.534231810169894e-06,
+      "loss": 0.9777,
+      "step": 15689
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5336263834319075e-06,
+      "loss": 0.8616,
+      "step": 15690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.533021004479353e-06,
+      "loss": 0.9166,
+      "step": 15691
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5324156733239546e-06,
+      "loss": 0.9115,
+      "step": 15692
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.531810389977445e-06,
+      "loss": 1.0019,
+      "step": 15693
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5312051544515463e-06,
+      "loss": 1.0159,
+      "step": 15694
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5305999667579908e-06,
+      "loss": 1.0376,
+      "step": 15695
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5299948269084983e-06,
+      "loss": 0.8613,
+      "step": 15696
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.529389734914798e-06,
+      "loss": 0.9249,
+      "step": 15697
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5287846907886067e-06,
+      "loss": 0.9834,
+      "step": 15698
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.528179694541656e-06,
+      "loss": 1.0044,
+      "step": 15699
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5275747461856605e-06,
+      "loss": 0.899,
+      "step": 15700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5269698457323454e-06,
+      "loss": 0.921,
+      "step": 15701
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5263649931934268e-06,
+      "loss": 0.6749,
+      "step": 15702
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5257601885806276e-06,
+      "loss": 0.8169,
+      "step": 15703
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5251554319056614e-06,
+      "loss": 0.8896,
+      "step": 15704
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.524550723180249e-06,
+      "loss": 1.0221,
+      "step": 15705
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.523946062416103e-06,
+      "loss": 0.827,
+      "step": 15706
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.523341449624942e-06,
+      "loss": 0.966,
+      "step": 15707
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5227368848184773e-06,
+      "loss": 1.0084,
+      "step": 15708
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.522132368008423e-06,
+      "loss": 1.0874,
+      "step": 15709
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5215278992064944e-06,
+      "loss": 0.8149,
+      "step": 15710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5209234784243984e-06,
+      "loss": 0.9735,
+      "step": 15711
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5203191056738494e-06,
+      "loss": 0.9267,
+      "step": 15712
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5197147809665536e-06,
+      "loss": 0.9765,
+      "step": 15713
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5191105043142235e-06,
+      "loss": 1.1107,
+      "step": 15714
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5185062757285624e-06,
+      "loss": 0.8577,
+      "step": 15715
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5179020952212807e-06,
+      "loss": 1.006,
+      "step": 15716
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5172979628040816e-06,
+      "loss": 0.9164,
+      "step": 15717
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.516693878488673e-06,
+      "loss": 0.8113,
+      "step": 15718
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.516089842286755e-06,
+      "loss": 1.0095,
+      "step": 15719
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5154858542100354e-06,
+      "loss": 1.0006,
+      "step": 15720
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5148819142702095e-06,
+      "loss": 0.8678,
+      "step": 15721
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5142780224789864e-06,
+      "loss": 0.9097,
+      "step": 15722
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.513674178848061e-06,
+      "loss": 0.8988,
+      "step": 15723
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5130703833891366e-06,
+      "loss": 0.9861,
+      "step": 15724
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.512466636113908e-06,
+      "loss": 1.2799,
+      "step": 15725
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5118629370340755e-06,
+      "loss": 0.8106,
+      "step": 15726
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5112592861613317e-06,
+      "loss": 0.955,
+      "step": 15727
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5106556835073768e-06,
+      "loss": 0.9127,
+      "step": 15728
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.510052129083902e-06,
+      "loss": 1.0442,
+      "step": 15729
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.509448622902604e-06,
+      "loss": 0.9186,
+      "step": 15730
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.508845164975172e-06,
+      "loss": 0.8947,
+      "step": 15731
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.508241755313302e-06,
+      "loss": 0.9086,
+      "step": 15732
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5076383939286803e-06,
+      "loss": 0.8882,
+      "step": 15733
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.507035080833e-06,
+      "loss": 0.7906,
+      "step": 15734
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.506431816037951e-06,
+      "loss": 1.1025,
+      "step": 15735
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5058285995552177e-06,
+      "loss": 0.8481,
+      "step": 15736
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.505225431396492e-06,
+      "loss": 0.8472,
+      "step": 15737
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.504622311573456e-06,
+      "loss": 0.8517,
+      "step": 15738
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.504019240097798e-06,
+      "loss": 0.944,
+      "step": 15739
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.503416216981199e-06,
+      "loss": 0.9939,
+      "step": 15740
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.502813242235347e-06,
+      "loss": 1.019,
+      "step": 15741
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.50221031587192e-06,
+      "loss": 1.0741,
+      "step": 15742
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.501607437902603e-06,
+      "loss": 1.025,
+      "step": 15743
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5010046083390716e-06,
+      "loss": 0.9892,
+      "step": 15744
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5004018271930123e-06,
+      "loss": 0.7965,
+      "step": 15745
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4997990944760995e-06,
+      "loss": 1.0027,
+      "step": 15746
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4991964102000137e-06,
+      "loss": 1.0774,
+      "step": 15747
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4985937743764284e-06,
+      "loss": 0.9478,
+      "step": 15748
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.497991187017023e-06,
+      "loss": 1.0057,
+      "step": 15749
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.497388648133469e-06,
+      "loss": 1.0016,
+      "step": 15750
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4967861577374443e-06,
+      "loss": 0.8773,
+      "step": 15751
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4961837158406178e-06,
+      "loss": 1.1357,
+      "step": 15752
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.495581322454666e-06,
+      "loss": 0.9258,
+      "step": 15753
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.494978977591256e-06,
+      "loss": 0.941,
+      "step": 15754
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4943766812620616e-06,
+      "loss": 1.1623,
+      "step": 15755
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4937744334787473e-06,
+      "loss": 0.997,
+      "step": 15756
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4931722342529884e-06,
+      "loss": 0.9765,
+      "step": 15757
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.492570083596447e-06,
+      "loss": 0.929,
+      "step": 15758
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.491967981520793e-06,
+      "loss": 0.8473,
+      "step": 15759
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4913659280376888e-06,
+      "loss": 0.8578,
+      "step": 15760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.490763923158803e-06,
+      "loss": 1.0634,
+      "step": 15761
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4901619668957947e-06,
+      "loss": 0.9053,
+      "step": 15762
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4895600592603313e-06,
+      "loss": 0.9767,
+      "step": 15763
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.48895820026407e-06,
+      "loss": 0.8782,
+      "step": 15764
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4883563899186757e-06,
+      "loss": 0.9437,
+      "step": 15765
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.487754628235805e-06,
+      "loss": 0.9163,
+      "step": 15766
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4871529152271184e-06,
+      "loss": 0.9222,
+      "step": 15767
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.486551250904276e-06,
+      "loss": 1.0772,
+      "step": 15768
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.485949635278931e-06,
+      "loss": 0.962,
+      "step": 15769
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4853480683627436e-06,
+      "loss": 0.8408,
+      "step": 15770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.484746550167365e-06,
+      "loss": 0.8161,
+      "step": 15771
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4841450807044533e-06,
+      "loss": 1.0389,
+      "step": 15772
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4835436599856576e-06,
+      "loss": 1.0036,
+      "step": 15773
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4829422880226358e-06,
+      "loss": 0.6576,
+      "step": 15774
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.482340964827033e-06,
+      "loss": 1.0586,
+      "step": 15775
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4817396904105057e-06,
+      "loss": 0.9147,
+      "step": 15776
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.481138464784698e-06,
+      "loss": 1.0568,
+      "step": 15777
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.480537287961264e-06,
+      "loss": 1.0962,
+      "step": 15778
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4799361599518433e-06,
+      "loss": 0.7673,
+      "step": 15779
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4793350807680928e-06,
+      "loss": 0.9629,
+      "step": 15780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.478734050421651e-06,
+      "loss": 0.861,
+      "step": 15781
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4781330689241673e-06,
+      "loss": 0.7685,
+      "step": 15782
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.477532136287281e-06,
+      "loss": 0.9432,
+      "step": 15783
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4769312525226403e-06,
+      "loss": 0.9188,
+      "step": 15784
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.476330417641881e-06,
+      "loss": 0.7871,
+      "step": 15785
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4757296316566503e-06,
+      "loss": 0.8248,
+      "step": 15786
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.475128894578583e-06,
+      "loss": 1.0355,
+      "step": 15787
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4745282064193226e-06,
+      "loss": 1.131,
+      "step": 15788
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4739275671905035e-06,
+      "loss": 0.8554,
+      "step": 15789
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4733269769037665e-06,
+      "loss": 1.0662,
+      "step": 15790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4727264355707426e-06,
+      "loss": 0.9858,
+      "step": 15791
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4721259432030742e-06,
+      "loss": 0.9946,
+      "step": 15792
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4715254998123904e-06,
+      "loss": 0.9668,
+      "step": 15793
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.470925105410329e-06,
+      "loss": 0.9726,
+      "step": 15794
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4703247600085173e-06,
+      "loss": 1.0223,
+      "step": 15795
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4697244636185917e-06,
+      "loss": 1.0977,
+      "step": 15796
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.469124216252179e-06,
+      "loss": 0.9066,
+      "step": 15797
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.468524017920912e-06,
+      "loss": 0.9492,
+      "step": 15798
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4679238686364164e-06,
+      "loss": 1.0296,
+      "step": 15799
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.467323768410323e-06,
+      "loss": 0.9321,
+      "step": 15800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4667237172542556e-06,
+      "loss": 0.9812,
+      "step": 15801
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4661237151798413e-06,
+      "loss": 0.8572,
+      "step": 15802
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.465523762198707e-06,
+      "loss": 1.1008,
+      "step": 15803
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4649238583224733e-06,
+      "loss": 0.8321,
+      "step": 15804
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4643240035627668e-06,
+      "loss": 1.0683,
+      "step": 15805
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4637241979312054e-06,
+      "loss": 0.8044,
+      "step": 15806
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.463124441439414e-06,
+      "loss": 0.8963,
+      "step": 15807
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4625247340990095e-06,
+      "loss": 0.8449,
+      "step": 15808
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4619250759216147e-06,
+      "loss": 0.8288,
+      "step": 15809
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4613254669188434e-06,
+      "loss": 0.8281,
+      "step": 15810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4607259071023174e-06,
+      "loss": 0.9866,
+      "step": 15811
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.46012639648365e-06,
+      "loss": 0.9169,
+      "step": 15812
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4595269350744583e-06,
+      "loss": 0.8962,
+      "step": 15813
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.458927522886352e-06,
+      "loss": 0.8667,
+      "step": 15814
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.458328159930953e-06,
+      "loss": 0.8153,
+      "step": 15815
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4577288462198673e-06,
+      "loss": 0.9554,
+      "step": 15816
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4571295817647104e-06,
+      "loss": 0.7651,
+      "step": 15817
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4565303665770894e-06,
+      "loss": 1.031,
+      "step": 15818
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.455931200668617e-06,
+      "loss": 0.9205,
+      "step": 15819
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4553320840508986e-06,
+      "loss": 0.9445,
+      "step": 15820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4547330167355464e-06,
+      "loss": 0.8663,
+      "step": 15821
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4541339987341622e-06,
+      "loss": 0.8307,
+      "step": 15822
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4535350300583565e-06,
+      "loss": 1.0014,
+      "step": 15823
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4529361107197296e-06,
+      "loss": 0.9007,
+      "step": 15824
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.45233724072989e-06,
+      "loss": 0.9307,
+      "step": 15825
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4517384201004363e-06,
+      "loss": 0.9431,
+      "step": 15826
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4511396488429727e-06,
+      "loss": 0.8748,
+      "step": 15827
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4505409269691028e-06,
+      "loss": 0.9333,
+      "step": 15828
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4499422544904217e-06,
+      "loss": 1.0083,
+      "step": 15829
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4493436314185325e-06,
+      "loss": 0.9336,
+      "step": 15830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4487450577650302e-06,
+      "loss": 0.8695,
+      "step": 15831
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4481465335415162e-06,
+      "loss": 0.8693,
+      "step": 15832
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.447548058759582e-06,
+      "loss": 0.9649,
+      "step": 15833
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.446949633430826e-06,
+      "loss": 0.888,
+      "step": 15834
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4463512575668413e-06,
+      "loss": 1.1162,
+      "step": 15835
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.445752931179223e-06,
+      "loss": 0.9873,
+      "step": 15836
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4451546542795585e-06,
+      "loss": 1.0227,
+      "step": 15837
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4445564268794468e-06,
+      "loss": 0.9284,
+      "step": 15838
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.443958248990473e-06,
+      "loss": 0.9515,
+      "step": 15839
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4433601206242298e-06,
+      "loss": 1.1062,
+      "step": 15840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4427620417923027e-06,
+      "loss": 0.9541,
+      "step": 15841
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.442164012506283e-06,
+      "loss": 0.7948,
+      "step": 15842
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.441566032777753e-06,
+      "loss": 1.0766,
+      "step": 15843
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.440968102618303e-06,
+      "loss": 0.9602,
+      "step": 15844
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.440370222039514e-06,
+      "loss": 0.9809,
+      "step": 15845
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4397723910529724e-06,
+      "loss": 0.991,
+      "step": 15846
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4391746096702584e-06,
+      "loss": 0.968,
+      "step": 15847
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.438576877902957e-06,
+      "loss": 0.8407,
+      "step": 15848
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4379791957626446e-06,
+      "loss": 0.9515,
+      "step": 15849
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4373815632609076e-06,
+      "loss": 0.9236,
+      "step": 15850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.43678398040932e-06,
+      "loss": 0.9753,
+      "step": 15851
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4361864472194626e-06,
+      "loss": 0.9865,
+      "step": 15852
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.435588963702909e-06,
+      "loss": 1.1219,
+      "step": 15853
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.43499152987124e-06,
+      "loss": 1.0907,
+      "step": 15854
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.434394145736026e-06,
+      "loss": 0.9431,
+      "step": 15855
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4337968113088454e-06,
+      "loss": 0.7849,
+      "step": 15856
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.433199526601268e-06,
+      "loss": 1.0089,
+      "step": 15857
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.432602291624869e-06,
+      "loss": 0.8159,
+      "step": 15858
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.432005106391217e-06,
+      "loss": 0.8938,
+      "step": 15859
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4314079709118827e-06,
+      "loss": 0.8936,
+      "step": 15860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4308108851984387e-06,
+      "loss": 1.0112,
+      "step": 15861
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4302138492624487e-06,
+      "loss": 1.073,
+      "step": 15862
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4296168631154848e-06,
+      "loss": 0.9937,
+      "step": 15863
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.429019926769109e-06,
+      "loss": 0.9806,
+      "step": 15864
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4284230402348908e-06,
+      "loss": 0.8937,
+      "step": 15865
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.427826203524391e-06,
+      "loss": 1.1023,
+      "step": 15866
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4272294166491773e-06,
+      "loss": 0.9462,
+      "step": 15867
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4266326796208076e-06,
+      "loss": 0.9893,
+      "step": 15868
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.426035992450848e-06,
+      "loss": 0.7645,
+      "step": 15869
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4254393551508555e-06,
+      "loss": 1.0053,
+      "step": 15870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4248427677323933e-06,
+      "loss": 0.9626,
+      "step": 15871
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4242462302070146e-06,
+      "loss": 1.0366,
+      "step": 15872
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4236497425862842e-06,
+      "loss": 1.1024,
+      "step": 15873
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4230533048817547e-06,
+      "loss": 1.0127,
+      "step": 15874
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.422456917104984e-06,
+      "loss": 0.8395,
+      "step": 15875
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.421860579267524e-06,
+      "loss": 0.9484,
+      "step": 15876
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4212642913809325e-06,
+      "loss": 0.8572,
+      "step": 15877
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4206680534567583e-06,
+      "loss": 1.0173,
+      "step": 15878
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4200718655065573e-06,
+      "loss": 0.8673,
+      "step": 15879
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4194757275418775e-06,
+      "loss": 1.0438,
+      "step": 15880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4188796395742716e-06,
+      "loss": 0.8911,
+      "step": 15881
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.418283601615285e-06,
+      "loss": 0.9267,
+      "step": 15882
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4176876136764706e-06,
+      "loss": 0.9219,
+      "step": 15883
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4170916757693703e-06,
+      "loss": 0.9784,
+      "step": 15884
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4164957879055334e-06,
+      "loss": 0.8532,
+      "step": 15885
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.415899950096506e-06,
+      "loss": 0.7656,
+      "step": 15886
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4153041623538294e-06,
+      "loss": 0.9062,
+      "step": 15887
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4147084246890478e-06,
+      "loss": 0.9233,
+      "step": 15888
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4141127371137064e-06,
+      "loss": 0.9675,
+      "step": 15889
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.413517099639342e-06,
+      "loss": 0.8953,
+      "step": 15890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.412921512277499e-06,
+      "loss": 0.864,
+      "step": 15891
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.412325975039713e-06,
+      "loss": 1.106,
+      "step": 15892
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4117304879375264e-06,
+      "loss": 1.0967,
+      "step": 15893
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4111350509824726e-06,
+      "loss": 1.0686,
+      "step": 15894
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4105396641860897e-06,
+      "loss": 1.146,
+      "step": 15895
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4099443275599154e-06,
+      "loss": 0.8877,
+      "step": 15896
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.40934904111548e-06,
+      "loss": 1.0784,
+      "step": 15897
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4087538048643216e-06,
+      "loss": 1.0235,
+      "step": 15898
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4081586188179684e-06,
+      "loss": 0.9239,
+      "step": 15899
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.407563482987956e-06,
+      "loss": 0.8621,
+      "step": 15900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.406968397385811e-06,
+      "loss": 0.836,
+      "step": 15901
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4063733620230673e-06,
+      "loss": 1.0328,
+      "step": 15902
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4057783769112493e-06,
+      "loss": 0.9753,
+      "step": 15903
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4051834420618886e-06,
+      "loss": 0.9336,
+      "step": 15904
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.404588557486508e-06,
+      "loss": 0.9043,
+      "step": 15905
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4039937231966374e-06,
+      "loss": 1.0547,
+      "step": 15906
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.403398939203796e-06,
+      "loss": 0.8713,
+      "step": 15907
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.402804205519515e-06,
+      "loss": 0.8983,
+      "step": 15908
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.402209522155311e-06,
+      "loss": 0.9801,
+      "step": 15909
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4016148891227102e-06,
+      "loss": 0.9537,
+      "step": 15910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.40102030643323e-06,
+      "loss": 0.9417,
+      "step": 15911
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.400425774098393e-06,
+      "loss": 0.8498,
+      "step": 15912
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.399831292129716e-06,
+      "loss": 1.0524,
+      "step": 15913
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3992368605387194e-06,
+      "loss": 0.8515,
+      "step": 15914
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.398642479336917e-06,
+      "loss": 0.9897,
+      "step": 15915
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3980481485358283e-06,
+      "loss": 1.024,
+      "step": 15916
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.397453868146964e-06,
+      "loss": 0.9053,
+      "step": 15917
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3968596381818434e-06,
+      "loss": 0.8813,
+      "step": 15918
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.396265458651974e-06,
+      "loss": 1.0886,
+      "step": 15919
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.395671329568871e-06,
+      "loss": 0.812,
+      "step": 15920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3950772509440474e-06,
+      "loss": 0.9181,
+      "step": 15921
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.394483222789009e-06,
+      "loss": 0.8288,
+      "step": 15922
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3938892451152697e-06,
+      "loss": 0.8498,
+      "step": 15923
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3932953179343327e-06,
+      "loss": 0.9026,
+      "step": 15924
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3927014412577094e-06,
+      "loss": 1.0473,
+      "step": 15925
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3921076150969024e-06,
+      "loss": 0.8871,
+      "step": 15926
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.391513839463421e-06,
+      "loss": 1.0495,
+      "step": 15927
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.390920114368765e-06,
+      "loss": 1.0252,
+      "step": 15928
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.390326439824442e-06,
+      "loss": 0.8751,
+      "step": 15929
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3897328158419485e-06,
+      "loss": 0.9838,
+      "step": 15930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.389139242432793e-06,
+      "loss": 0.9918,
+      "step": 15931
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3885457196084712e-06,
+      "loss": 0.6512,
+      "step": 15932
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.387952247380485e-06,
+      "loss": 0.8114,
+      "step": 15933
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3873588257603305e-06,
+      "loss": 0.9614,
+      "step": 15934
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.386765454759507e-06,
+      "loss": 0.9179,
+      "step": 15935
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.386172134389508e-06,
+      "loss": 0.971,
+      "step": 15936
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.385578864661834e-06,
+      "loss": 1.0933,
+      "step": 15937
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.384985645587974e-06,
+      "loss": 0.8414,
+      "step": 15938
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3843924771794254e-06,
+      "loss": 0.9428,
+      "step": 15939
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3837993594476775e-06,
+      "loss": 1.0298,
+      "step": 15940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.383206292404226e-06,
+      "loss": 1.0112,
+      "step": 15941
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3826132760605545e-06,
+      "loss": 0.9636,
+      "step": 15942
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3820203104281616e-06,
+      "loss": 0.7572,
+      "step": 15943
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.381427395518529e-06,
+      "loss": 0.9711,
+      "step": 15944
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3808345313431484e-06,
+      "loss": 0.9545,
+      "step": 15945
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3802417179135027e-06,
+      "loss": 0.988,
+      "step": 15946
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.379648955241082e-06,
+      "loss": 0.8425,
+      "step": 15947
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.379056243337366e-06,
+      "loss": 1.0257,
+      "step": 15948
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3784635822138424e-06,
+      "loss": 1.0029,
+      "step": 15949
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.377870971881991e-06,
+      "loss": 0.9063,
+      "step": 15950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.377278412353296e-06,
+      "loss": 0.9528,
+      "step": 15951
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3766859036392353e-06,
+      "loss": 0.8819,
+      "step": 15952
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3760934457512897e-06,
+      "loss": 0.7896,
+      "step": 15953
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3755010387009406e-06,
+      "loss": 0.8493,
+      "step": 15954
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.374908682499662e-06,
+      "loss": 0.9887,
+      "step": 15955
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3743163771589333e-06,
+      "loss": 0.953,
+      "step": 15956
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.373724122690228e-06,
+      "loss": 0.764,
+      "step": 15957
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3731319191050232e-06,
+      "loss": 0.9595,
+      "step": 15958
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.372539766414791e-06,
+      "loss": 0.8826,
+      "step": 15959
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.371947664631006e-06,
+      "loss": 0.875,
+      "step": 15960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.371355613765137e-06,
+      "loss": 0.989,
+      "step": 15961
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3707636138286577e-06,
+      "loss": 0.9745,
+      "step": 15962
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3701716648330365e-06,
+      "loss": 0.8788,
+      "step": 15963
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3695797667897437e-06,
+      "loss": 0.9884,
+      "step": 15964
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3689879197102426e-06,
+      "loss": 0.9849,
+      "step": 15965
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3683961236060077e-06,
+      "loss": 1.0232,
+      "step": 15966
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.367804378488499e-06,
+      "loss": 1.0918,
+      "step": 15967
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3672126843691855e-06,
+      "loss": 0.9879,
+      "step": 15968
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.366621041259527e-06,
+      "loss": 0.9604,
+      "step": 15969
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.36602944917099e-06,
+      "loss": 0.7792,
+      "step": 15970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3654379081150336e-06,
+      "loss": 0.9218,
+      "step": 15971
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.364846418103122e-06,
+      "loss": 0.8155,
+      "step": 15972
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.364254979146711e-06,
+      "loss": 0.947,
+      "step": 15973
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3636635912572643e-06,
+      "loss": 0.8612,
+      "step": 15974
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.363072254446235e-06,
+      "loss": 0.886,
+      "step": 15975
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.362480968725085e-06,
+      "loss": 0.9001,
+      "step": 15976
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3618897341052655e-06,
+      "loss": 0.8446,
+      "step": 15977
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3612985505982346e-06,
+      "loss": 1.0275,
+      "step": 15978
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.360707418215447e-06,
+      "loss": 0.9463,
+      "step": 15979
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.360116336968353e-06,
+      "loss": 0.9844,
+      "step": 15980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3595253068684084e-06,
+      "loss": 0.8231,
+      "step": 15981
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3589343279270594e-06,
+      "loss": 1.0469,
+      "step": 15982
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3583434001557607e-06,
+      "loss": 0.9848,
+      "step": 15983
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3577525235659577e-06,
+      "loss": 1.0851,
+      "step": 15984
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.357161698169102e-06,
+      "loss": 0.9459,
+      "step": 15985
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.356570923976637e-06,
+      "loss": 0.9236,
+      "step": 15986
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.355980201000011e-06,
+      "loss": 1.0353,
+      "step": 15987
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.355389529250669e-06,
+      "loss": 0.8769,
+      "step": 15988
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3547989087400562e-06,
+      "loss": 1.0187,
+      "step": 15989
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3542083394796126e-06,
+      "loss": 1.0866,
+      "step": 15990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3536178214807848e-06,
+      "loss": 0.9986,
+      "step": 15991
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3530273547550093e-06,
+      "loss": 0.8187,
+      "step": 15992
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3524369393137304e-06,
+      "loss": 0.7633,
+      "step": 15993
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3518465751683827e-06,
+      "loss": 0.9215,
+      "step": 15994
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.351256262330409e-06,
+      "loss": 0.9567,
+      "step": 15995
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.350666000811243e-06,
+      "loss": 0.982,
+      "step": 15996
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.350075790622324e-06,
+      "loss": 0.8577,
+      "step": 15997
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.349485631775083e-06,
+      "loss": 0.8642,
+      "step": 15998
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.348895524280959e-06,
+      "loss": 0.992,
+      "step": 15999
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.348305468151379e-06,
+      "loss": 1.0149,
+      "step": 16000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3477154633977826e-06,
+      "loss": 0.8957,
+      "step": 16001
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.347125510031596e-06,
+      "loss": 0.976,
+      "step": 16002
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3465356080642524e-06,
+      "loss": 0.9394,
+      "step": 16003
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.345945757507178e-06,
+      "loss": 0.8989,
+      "step": 16004
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3453559583718044e-06,
+      "loss": 0.8713,
+      "step": 16005
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3447662106695546e-06,
+      "loss": 0.9676,
+      "step": 16006
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.344176514411859e-06,
+      "loss": 0.9604,
+      "step": 16007
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3435868696101394e-06,
+      "loss": 0.9222,
+      "step": 16008
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3429972762758237e-06,
+      "loss": 1.0805,
+      "step": 16009
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.342407734420331e-06,
+      "loss": 0.9368,
+      "step": 16010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3418182440550875e-06,
+      "loss": 1.0663,
+      "step": 16011
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3412288051915107e-06,
+      "loss": 0.806,
+      "step": 16012
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3406394178410226e-06,
+      "loss": 0.8899,
+      "step": 16013
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.340050082015044e-06,
+      "loss": 0.8775,
+      "step": 16014
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.33946079772499e-06,
+      "loss": 0.9066,
+      "step": 16015
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.338871564982282e-06,
+      "loss": 0.8454,
+      "step": 16016
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3382823837983314e-06,
+      "loss": 0.861,
+      "step": 16017
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3376932541845574e-06,
+      "loss": 0.8402,
+      "step": 16018
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3371041761523714e-06,
+      "loss": 1.1589,
+      "step": 16019
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.33651514971319e-06,
+      "loss": 0.9168,
+      "step": 16020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3359261748784213e-06,
+      "loss": 1.0065,
+      "step": 16021
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3353372516594804e-06,
+      "loss": 1.0173,
+      "step": 16022
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3347483800677715e-06,
+      "loss": 0.8697,
+      "step": 16023
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3341595601147127e-06,
+      "loss": 0.9345,
+      "step": 16024
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3335707918117058e-06,
+      "loss": 1.0214,
+      "step": 16025
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3329820751701615e-06,
+      "loss": 0.832,
+      "step": 16026
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.332393410201483e-06,
+      "loss": 0.8163,
+      "step": 16027
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3318047969170782e-06,
+      "loss": 1.1227,
+      "step": 16028
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.331216235328349e-06,
+      "loss": 0.9649,
+      "step": 16029
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3306277254467013e-06,
+      "loss": 1.0109,
+      "step": 16030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.330039267283535e-06,
+      "loss": 0.8439,
+      "step": 16031
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.329450860850253e-06,
+      "loss": 0.8831,
+      "step": 16032
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3288625061582533e-06,
+      "loss": 0.9372,
+      "step": 16033
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.328274203218939e-06,
+      "loss": 0.8696,
+      "step": 16034
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.327685952043701e-06,
+      "loss": 1.0756,
+      "step": 16035
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3270977526439462e-06,
+      "loss": 0.9826,
+      "step": 16036
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.326509605031064e-06,
+      "loss": 0.8637,
+      "step": 16037
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.325921509216454e-06,
+      "loss": 0.837,
+      "step": 16038
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.325333465211506e-06,
+      "loss": 0.9151,
+      "step": 16039
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.324745473027617e-06,
+      "loss": 0.9167,
+      "step": 16040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.324157532676176e-06,
+      "loss": 0.9392,
+      "step": 16041
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.323569644168578e-06,
+      "loss": 0.9111,
+      "step": 16042
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3229818075162088e-06,
+      "loss": 1.0129,
+      "step": 16043
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3223940227304614e-06,
+      "loss": 1.0438,
+      "step": 16044
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.321806289822721e-06,
+      "loss": 0.8793,
+      "step": 16045
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3212186088043764e-06,
+      "loss": 0.7375,
+      "step": 16046
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3206309796868147e-06,
+      "loss": 1.0043,
+      "step": 16047
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3200434024814187e-06,
+      "loss": 0.9703,
+      "step": 16048
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.319455877199576e-06,
+      "loss": 0.8733,
+      "step": 16049
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3188684038526655e-06,
+      "loss": 1.0675,
+      "step": 16050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.318280982452073e-06,
+      "loss": 0.8271,
+      "step": 16051
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3176936130091766e-06,
+      "loss": 0.7977,
+      "step": 16052
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3171062955353603e-06,
+      "loss": 0.841,
+      "step": 16053
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.316519030041998e-06,
+      "loss": 1.0701,
+      "step": 16054
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.315931816540473e-06,
+      "loss": 0.9099,
+      "step": 16055
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.315344655042159e-06,
+      "loss": 0.8976,
+      "step": 16056
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3147575455584347e-06,
+      "loss": 0.932,
+      "step": 16057
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3141704881006706e-06,
+      "loss": 0.9214,
+      "step": 16058
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3135834826802473e-06,
+      "loss": 0.9962,
+      "step": 16059
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3129965293085327e-06,
+      "loss": 0.7917,
+      "step": 16060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3124096279969027e-06,
+      "loss": 0.7002,
+      "step": 16061
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3118227787567247e-06,
+      "loss": 1.0207,
+      "step": 16062
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3112359815993723e-06,
+      "loss": 0.8328,
+      "step": 16063
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3106492365362115e-06,
+      "loss": 0.9823,
+      "step": 16064
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.310062543578613e-06,
+      "loss": 0.9765,
+      "step": 16065
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.309475902737941e-06,
+      "loss": 0.8421,
+      "step": 16066
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3088893140255647e-06,
+      "loss": 1.0773,
+      "step": 16067
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.308302777452846e-06,
+      "loss": 0.8764,
+      "step": 16068
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.307716293031152e-06,
+      "loss": 1.0722,
+      "step": 16069
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3071298607718425e-06,
+      "loss": 0.9535,
+      "step": 16070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.306543480686281e-06,
+      "loss": 0.9439,
+      "step": 16071
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3059571527858306e-06,
+      "loss": 0.9535,
+      "step": 16072
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.305370877081847e-06,
+      "loss": 0.7606,
+      "step": 16073
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3047846535856933e-06,
+      "loss": 0.8067,
+      "step": 16074
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3041984823087237e-06,
+      "loss": 1.0216,
+      "step": 16075
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3036123632622993e-06,
+      "loss": 0.9629,
+      "step": 16076
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.303026296457771e-06,
+      "loss": 0.9798,
+      "step": 16077
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.302440281906499e-06,
+      "loss": 0.702,
+      "step": 16078
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.301854319619833e-06,
+      "loss": 0.9254,
+      "step": 16079
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.301268409609129e-06,
+      "loss": 1.0035,
+      "step": 16080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3006825518857332e-06,
+      "loss": 1.0193,
+      "step": 16081
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3000967464610045e-06,
+      "loss": 0.9186,
+      "step": 16082
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.299510993346288e-06,
+      "loss": 0.9782,
+      "step": 16083
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.298925292552934e-06,
+      "loss": 0.907,
+      "step": 16084
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.298339644092289e-06,
+      "loss": 0.8673,
+      "step": 16085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.297754047975702e-06,
+      "loss": 0.7908,
+      "step": 16086
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2971685042145152e-06,
+      "loss": 1.0317,
+      "step": 16087
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2965830128200778e-06,
+      "loss": 1.0333,
+      "step": 16088
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2959975738037293e-06,
+      "loss": 0.9349,
+      "step": 16089
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.295412187176816e-06,
+      "loss": 0.8533,
+      "step": 16090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2948268529506768e-06,
+      "loss": 0.8114,
+      "step": 16091
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.294241571136655e-06,
+      "loss": 0.8385,
+      "step": 16092
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2936563417460853e-06,
+      "loss": 0.8037,
+      "step": 16093
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.293071164790314e-06,
+      "loss": 0.9812,
+      "step": 16094
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.292486040280673e-06,
+      "loss": 0.9045,
+      "step": 16095
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2919009682285025e-06,
+      "loss": 1.057,
+      "step": 16096
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.291315948645135e-06,
+      "loss": 0.8658,
+      "step": 16097
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2907309815419084e-06,
+      "loss": 0.9576,
+      "step": 16098
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2901460669301518e-06,
+      "loss": 0.9538,
+      "step": 16099
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.289561204821203e-06,
+      "loss": 0.846,
+      "step": 16100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2889763952263887e-06,
+      "loss": 0.8914,
+      "step": 16101
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.288391638157044e-06,
+      "loss": 0.9323,
+      "step": 16102
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2878069336244945e-06,
+      "loss": 1.1351,
+      "step": 16103
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2872222816400724e-06,
+      "loss": 1.0153,
+      "step": 16104
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.286637682215101e-06,
+      "loss": 1.0143,
+      "step": 16105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.286053135360909e-06,
+      "loss": 0.9408,
+      "step": 16106
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2854686410888237e-06,
+      "loss": 0.9835,
+      "step": 16107
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2848841994101663e-06,
+      "loss": 0.8028,
+      "step": 16108
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.284299810336263e-06,
+      "loss": 0.9539,
+      "step": 16109
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2837154738784333e-06,
+      "loss": 0.7795,
+      "step": 16110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.283131190048002e-06,
+      "loss": 0.9515,
+      "step": 16111
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2825469588562856e-06,
+      "loss": 1.2918,
+      "step": 16112
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.281962780314607e-06,
+      "loss": 0.9865,
+      "step": 16113
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2813786544342816e-06,
+      "loss": 1.0719,
+      "step": 16114
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2807945812266296e-06,
+      "loss": 0.8637,
+      "step": 16115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2802105607029624e-06,
+      "loss": 1.0346,
+      "step": 16116
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2796265928746027e-06,
+      "loss": 1.0077,
+      "step": 16117
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.279042677752858e-06,
+      "loss": 0.8314,
+      "step": 16118
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2784588153490463e-06,
+      "loss": 0.859,
+      "step": 16119
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.277875005674476e-06,
+      "loss": 0.9936,
+      "step": 16120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.277291248740462e-06,
+      "loss": 0.8585,
+      "step": 16121
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2767075445583103e-06,
+      "loss": 0.9616,
+      "step": 16122
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.276123893139334e-06,
+      "loss": 0.9076,
+      "step": 16123
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2755402944948374e-06,
+      "loss": 0.9295,
+      "step": 16124
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.274956748636132e-06,
+      "loss": 0.919,
+      "step": 16125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.274373255574519e-06,
+      "loss": 1.0001,
+      "step": 16126
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2737898153213077e-06,
+      "loss": 1.078,
+      "step": 16127
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2732064278877975e-06,
+      "loss": 1.0625,
+      "step": 16128
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.272623093285297e-06,
+      "loss": 0.8809,
+      "step": 16129
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2720398115251042e-06,
+      "loss": 0.9405,
+      "step": 16130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2714565826185225e-06,
+      "loss": 0.979,
+      "step": 16131
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.270873406576849e-06,
+      "loss": 0.9256,
+      "step": 16132
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2702902834113862e-06,
+      "loss": 0.9779,
+      "step": 16133
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.269707213133428e-06,
+      "loss": 0.9264,
+      "step": 16134
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2691241957542746e-06,
+      "loss": 1.1566,
+      "step": 16135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2685412312852194e-06,
+      "loss": 1.021,
+      "step": 16136
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2679583197375597e-06,
+      "loss": 0.944,
+      "step": 16137
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2673754611225863e-06,
+      "loss": 0.9445,
+      "step": 16138
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.266792655451593e-06,
+      "loss": 0.9614,
+      "step": 16139
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2662099027358736e-06,
+      "loss": 0.9188,
+      "step": 16140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.265627202986716e-06,
+      "loss": 0.9367,
+      "step": 16141
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.265044556215413e-06,
+      "loss": 0.9854,
+      "step": 16142
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.264461962433249e-06,
+      "loss": 0.8528,
+      "step": 16143
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.263879421651516e-06,
+      "loss": 1.0423,
+      "step": 16144
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2632969338814964e-06,
+      "loss": 0.8259,
+      "step": 16145
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.26271449913448e-06,
+      "loss": 0.912,
+      "step": 16146
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2621321174217477e-06,
+      "loss": 0.875,
+      "step": 16147
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.261549788754586e-06,
+      "loss": 0.9731,
+      "step": 16148
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2609675131442745e-06,
+      "loss": 0.9114,
+      "step": 16149
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.260385290602098e-06,
+      "loss": 1.0809,
+      "step": 16150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2598031211393307e-06,
+      "loss": 0.8996,
+      "step": 16151
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.259221004767261e-06,
+      "loss": 0.9054,
+      "step": 16152
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.25863894149716e-06,
+      "loss": 0.7577,
+      "step": 16153
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2580569313403096e-06,
+      "loss": 0.7684,
+      "step": 16154
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2574749743079825e-06,
+      "loss": 0.8935,
+      "step": 16155
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2568930704114573e-06,
+      "loss": 1.0383,
+      "step": 16156
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.256311219662005e-06,
+      "loss": 0.9281,
+      "step": 16157
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2557294220709027e-06,
+      "loss": 0.7146,
+      "step": 16158
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2551476776494175e-06,
+      "loss": 0.9896,
+      "step": 16159
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.254565986408826e-06,
+      "loss": 0.8848,
+      "step": 16160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2539843483603934e-06,
+      "loss": 1.0529,
+      "step": 16161
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2534027635153936e-06,
+      "loss": 0.9663,
+      "step": 16162
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.25282123188509e-06,
+      "loss": 1.0603,
+      "step": 16163
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.252239753480752e-06,
+      "loss": 1.0153,
+      "step": 16164
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2516583283136474e-06,
+      "loss": 0.8366,
+      "step": 16165
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.251076956395037e-06,
+      "loss": 0.8694,
+      "step": 16166
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.250495637736189e-06,
+      "loss": 1.1506,
+      "step": 16167
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.249914372348363e-06,
+      "loss": 0.869,
+      "step": 16168
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.249333160242824e-06,
+      "loss": 0.9887,
+      "step": 16169
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2487520014308283e-06,
+      "loss": 0.857,
+      "step": 16170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.248170895923641e-06,
+      "loss": 0.9818,
+      "step": 16171
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2475898437325166e-06,
+      "loss": 0.8926,
+      "step": 16172
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.247008844868716e-06,
+      "loss": 0.9528,
+      "step": 16173
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2464278993434912e-06,
+      "loss": 1.0651,
+      "step": 16174
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.245847007168105e-06,
+      "loss": 0.953,
+      "step": 16175
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.245266168353806e-06,
+      "loss": 0.935,
+      "step": 16176
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2446853829118526e-06,
+      "loss": 0.9855,
+      "step": 16177
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2441046508534932e-06,
+      "loss": 0.9451,
+      "step": 16178
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.243523972189982e-06,
+      "loss": 0.8421,
+      "step": 16179
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2429433469325672e-06,
+      "loss": 0.8395,
+      "step": 16180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2423627750925016e-06,
+      "loss": 0.9528,
+      "step": 16181
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.24178225668103e-06,
+      "loss": 1.0312,
+      "step": 16182
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2412017917094024e-06,
+      "loss": 0.9387,
+      "step": 16183
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2406213801888636e-06,
+      "loss": 1.0684,
+      "step": 16184
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.240041022130661e-06,
+      "loss": 1.0485,
+      "step": 16185
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2394607175460335e-06,
+      "loss": 0.8954,
+      "step": 16186
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2388804664462326e-06,
+      "loss": 1.0474,
+      "step": 16187
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2383002688424945e-06,
+      "loss": 1.0662,
+      "step": 16188
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.237720124746064e-06,
+      "loss": 0.773,
+      "step": 16189
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.237140034168177e-06,
+      "loss": 0.8781,
+      "step": 16190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.236559997120078e-06,
+      "loss": 0.8802,
+      "step": 16191
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2359800136129995e-06,
+      "loss": 1.0495,
+      "step": 16192
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2354000836581834e-06,
+      "loss": 0.7368,
+      "step": 16193
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.234820207266862e-06,
+      "loss": 0.9135,
+      "step": 16194
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2342403844502735e-06,
+      "loss": 0.997,
+      "step": 16195
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.233660615219648e-06,
+      "loss": 0.808,
+      "step": 16196
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2330808995862206e-06,
+      "loss": 1.0936,
+      "step": 16197
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2325012375612253e-06,
+      "loss": 0.9579,
+      "step": 16198
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2319216291558894e-06,
+      "loss": 0.9291,
+      "step": 16199
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2313420743814453e-06,
+      "loss": 0.9121,
+      "step": 16200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2307625732491183e-06,
+      "loss": 0.8392,
+      "step": 16201
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2301831257701405e-06,
+      "loss": 0.7802,
+      "step": 16202
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2296037319557347e-06,
+      "loss": 0.7898,
+      "step": 16203
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2290243918171295e-06,
+      "loss": 0.9851,
+      "step": 16204
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2284451053655465e-06,
+      "loss": 1.0146,
+      "step": 16205
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2278658726122127e-06,
+      "loss": 0.8328,
+      "step": 16206
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2272866935683474e-06,
+      "loss": 1.0246,
+      "step": 16207
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.226707568245175e-06,
+      "loss": 1.0244,
+      "step": 16208
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2261284966539105e-06,
+      "loss": 1.0152,
+      "step": 16209
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2255494788057806e-06,
+      "loss": 0.9004,
+      "step": 16210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.224970514711999e-06,
+      "loss": 0.9842,
+      "step": 16211
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2243916043837854e-06,
+      "loss": 0.8669,
+      "step": 16212
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2238127478323534e-06,
+      "loss": 1.0279,
+      "step": 16213
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2232339450689217e-06,
+      "loss": 0.8876,
+      "step": 16214
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2226551961047005e-06,
+      "loss": 0.9663,
+      "step": 16215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2220765009509065e-06,
+      "loss": 0.8294,
+      "step": 16216
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2214978596187485e-06,
+      "loss": 0.8192,
+      "step": 16217
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2209192721194408e-06,
+      "loss": 0.8349,
+      "step": 16218
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.22034073846419e-06,
+      "loss": 0.77,
+      "step": 16219
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2197622586642085e-06,
+      "loss": 1.0005,
+      "step": 16220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2191838327307e-06,
+      "loss": 0.7869,
+      "step": 16221
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.218605460674876e-06,
+      "loss": 0.8966,
+      "step": 16222
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2180271425079396e-06,
+      "loss": 0.8527,
+      "step": 16223
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.217448878241098e-06,
+      "loss": 0.8961,
+      "step": 16224
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2168706678855512e-06,
+      "loss": 0.8906,
+      "step": 16225
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2162925114525052e-06,
+      "loss": 0.9262,
+      "step": 16226
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.215714408953159e-06,
+      "loss": 0.9088,
+      "step": 16227
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2151363603987166e-06,
+      "loss": 0.9122,
+      "step": 16228
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2145583658003733e-06,
+      "loss": 0.8344,
+      "step": 16229
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.213980425169332e-06,
+      "loss": 0.9902,
+      "step": 16230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2134025385167858e-06,
+      "loss": 0.9664,
+      "step": 16231
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2128247058539333e-06,
+      "loss": 0.8998,
+      "step": 16232
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.212246927191972e-06,
+      "loss": 0.9302,
+      "step": 16233
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2116692025420916e-06,
+      "loss": 0.8818,
+      "step": 16234
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2110915319154908e-06,
+      "loss": 1.014,
+      "step": 16235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.210513915323356e-06,
+      "loss": 1.0531,
+      "step": 16236
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.209936352776884e-06,
+      "loss": 0.8895,
+      "step": 16237
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.20935884428726e-06,
+      "loss": 0.9653,
+      "step": 16238
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2087813898656775e-06,
+      "loss": 0.8582,
+      "step": 16239
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2082039895233205e-06,
+      "loss": 0.9718,
+      "step": 16240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2076266432713796e-06,
+      "loss": 1.0259,
+      "step": 16241
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2070493511210374e-06,
+      "loss": 0.9571,
+      "step": 16242
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2064721130834822e-06,
+      "loss": 0.9256,
+      "step": 16243
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.205894929169893e-06,
+      "loss": 1.0063,
+      "step": 16244
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.205317799391458e-06,
+      "loss": 0.9144,
+      "step": 16245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.204740723759356e-06,
+      "loss": 0.9509,
+      "step": 16246
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.20416370228477e-06,
+      "loss": 0.8087,
+      "step": 16247
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2035867349788765e-06,
+      "loss": 1.0797,
+      "step": 16248
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2030098218528574e-06,
+      "loss": 0.9289,
+      "step": 16249
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2024329629178866e-06,
+      "loss": 0.8652,
+      "step": 16250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.201856158185144e-06,
+      "loss": 0.9794,
+      "step": 16251
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.201279407665802e-06,
+      "loss": 0.9921,
+      "step": 16252
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2007027113710387e-06,
+      "loss": 1.0867,
+      "step": 16253
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2001260693120236e-06,
+      "loss": 0.9817,
+      "step": 16254
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1995494814999325e-06,
+      "loss": 0.9848,
+      "step": 16255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.198972947945933e-06,
+      "loss": 1.0313,
+      "step": 16256
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1983964686611976e-06,
+      "loss": 1.0765,
+      "step": 16257
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1978200436568964e-06,
+      "loss": 1.0103,
+      "step": 16258
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1972436729441943e-06,
+      "loss": 0.9668,
+      "step": 16259
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.196667356534263e-06,
+      "loss": 0.6715,
+      "step": 16260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.196091094438263e-06,
+      "loss": 0.7426,
+      "step": 16261
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.195514886667364e-06,
+      "loss": 1.039,
+      "step": 16262
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1949387332327267e-06,
+      "loss": 1.0052,
+      "step": 16263
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1943626341455163e-06,
+      "loss": 0.9501,
+      "step": 16264
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.193786589416892e-06,
+      "loss": 1.145,
+      "step": 16265
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1932105990580176e-06,
+      "loss": 0.8404,
+      "step": 16266
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1926346630800476e-06,
+      "loss": 0.9709,
+      "step": 16267
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.192058781494148e-06,
+      "loss": 0.9032,
+      "step": 16268
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.19148295431147e-06,
+      "loss": 1.0099,
+      "step": 16269
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1909071815431753e-06,
+      "loss": 1.0712,
+      "step": 16270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1903314632004147e-06,
+      "loss": 1.0337,
+      "step": 16271
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1897557992943465e-06,
+      "loss": 0.9506,
+      "step": 16272
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1891801898361204e-06,
+      "loss": 1.0781,
+      "step": 16273
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.188604634836892e-06,
+      "loss": 0.9653,
+      "step": 16274
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1880291343078097e-06,
+      "loss": 0.8986,
+      "step": 16275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.187453688260027e-06,
+      "loss": 0.9577,
+      "step": 16276
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1868782967046893e-06,
+      "loss": 0.9121,
+      "step": 16277
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1863029596529476e-06,
+      "loss": 0.9999,
+      "step": 16278
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.185727677115945e-06,
+      "loss": 1.0986,
+      "step": 16279
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.185152449104833e-06,
+      "loss": 0.9194,
+      "step": 16280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1845772756307525e-06,
+      "loss": 0.9228,
+      "step": 16281
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.184002156704851e-06,
+      "loss": 0.9586,
+      "step": 16282
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.183427092338266e-06,
+      "loss": 1.0826,
+      "step": 16283
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1828520825421446e-06,
+      "loss": 0.9127,
+      "step": 16284
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1822771273276234e-06,
+      "loss": 0.8032,
+      "step": 16285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1817022267058453e-06,
+      "loss": 0.8088,
+      "step": 16286
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1811273806879453e-06,
+      "loss": 0.947,
+      "step": 16287
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.180552589285065e-06,
+      "loss": 0.9152,
+      "step": 16288
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1799778525083367e-06,
+      "loss": 1.087,
+      "step": 16289
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1794031703688978e-06,
+      "loss": 0.8792,
+      "step": 16290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.178828542877884e-06,
+      "loss": 0.9057,
+      "step": 16291
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.178253970046426e-06,
+      "loss": 0.9869,
+      "step": 16292
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1776794518856594e-06,
+      "loss": 0.9085,
+      "step": 16293
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1771049884067107e-06,
+      "loss": 0.8905,
+      "step": 16294
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.176530579620715e-06,
+      "loss": 0.8841,
+      "step": 16295
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1759562255387967e-06,
+      "loss": 0.9428,
+      "step": 16296
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1753819261720877e-06,
+      "loss": 0.7543,
+      "step": 16297
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.174807681531711e-06,
+      "loss": 0.741,
+      "step": 16298
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1742334916287974e-06,
+      "loss": 0.7737,
+      "step": 16299
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1736593564744663e-06,
+      "loss": 0.996,
+      "step": 16300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1730852760798455e-06,
+      "loss": 1.0412,
+      "step": 16301
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.172511250456053e-06,
+      "loss": 0.968,
+      "step": 16302
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.171937279614218e-06,
+      "loss": 1.0835,
+      "step": 16303
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.171363363565454e-06,
+      "loss": 0.9077,
+      "step": 16304
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.170789502320885e-06,
+      "loss": 0.8834,
+      "step": 16305
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1702156958916265e-06,
+      "loss": 0.9497,
+      "step": 16306
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1696419442887985e-06,
+      "loss": 0.9086,
+      "step": 16307
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1690682475235136e-06,
+      "loss": 0.7638,
+      "step": 16308
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1684946056068914e-06,
+      "loss": 0.9423,
+      "step": 16309
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1679210185500423e-06,
+      "loss": 0.8734,
+      "step": 16310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.167347486364082e-06,
+      "loss": 1.0503,
+      "step": 16311
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.166774009060121e-06,
+      "loss": 0.8153,
+      "step": 16312
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.166200586649272e-06,
+      "loss": 0.9863,
+      "step": 16313
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.165627219142642e-06,
+      "loss": 0.9852,
+      "step": 16314
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1650539065513415e-06,
+      "loss": 1.0456,
+      "step": 16315
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1644806488864787e-06,
+      "loss": 0.8315,
+      "step": 16316
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.163907446159162e-06,
+      "loss": 1.0526,
+      "step": 16317
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1633342983804934e-06,
+      "loss": 0.9643,
+      "step": 16318
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1627612055615808e-06,
+      "loss": 0.859,
+      "step": 16319
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1621881677135247e-06,
+      "loss": 0.877,
+      "step": 16320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1616151848474317e-06,
+      "loss": 0.8696,
+      "step": 16321
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1610422569743976e-06,
+      "loss": 0.9951,
+      "step": 16322
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.160469384105528e-06,
+      "loss": 0.8567,
+      "step": 16323
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1598965662519185e-06,
+      "loss": 0.9927,
+      "step": 16324
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1593238034246683e-06,
+      "loss": 0.9451,
+      "step": 16325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.158751095634877e-06,
+      "loss": 1.0639,
+      "step": 16326
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1581784428936374e-06,
+      "loss": 1.0708,
+      "step": 16327
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1576058452120473e-06,
+      "loss": 0.8601,
+      "step": 16328
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.157033302601197e-06,
+      "loss": 0.7989,
+      "step": 16329
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.156460815072184e-06,
+      "loss": 1.0402,
+      "step": 16330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1558883826360955e-06,
+      "loss": 0.9359,
+      "step": 16331
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1553160053040262e-06,
+      "loss": 0.9794,
+      "step": 16332
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1547436830870623e-06,
+      "loss": 1.046,
+      "step": 16333
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.154171415996296e-06,
+      "loss": 0.9265,
+      "step": 16334
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1535992040428113e-06,
+      "loss": 0.9134,
+      "step": 16335
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.153027047237698e-06,
+      "loss": 0.7806,
+      "step": 16336
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1524549455920363e-06,
+      "loss": 1.0355,
+      "step": 16337
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1518828991169177e-06,
+      "loss": 0.8878,
+      "step": 16338
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1513109078234197e-06,
+      "loss": 0.8986,
+      "step": 16339
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.150738971722629e-06,
+      "loss": 1.0459,
+      "step": 16340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.150167090825622e-06,
+      "loss": 1.0698,
+      "step": 16341
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.149595265143484e-06,
+      "loss": 0.8591,
+      "step": 16342
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1490234946872884e-06,
+      "loss": 0.9497,
+      "step": 16343
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.148451779468118e-06,
+      "loss": 0.8761,
+      "step": 16344
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1478801194970455e-06,
+      "loss": 1.1799,
+      "step": 16345
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1473085147851508e-06,
+      "loss": 0.8319,
+      "step": 16346
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1467369653435043e-06,
+      "loss": 0.8854,
+      "step": 16347
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.146165471183184e-06,
+      "loss": 0.9973,
+      "step": 16348
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.145594032315259e-06,
+      "loss": 0.8886,
+      "step": 16349
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1450226487508017e-06,
+      "loss": 0.8055,
+      "step": 16350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.144451320500885e-06,
+      "loss": 0.8739,
+      "step": 16351
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1438800475765737e-06,
+      "loss": 0.9821,
+      "step": 16352
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.143308829988941e-06,
+      "loss": 1.0457,
+      "step": 16353
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1427376677490496e-06,
+      "loss": 0.7763,
+      "step": 16354
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1421665608679697e-06,
+      "loss": 0.877,
+      "step": 16355
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.141595509356763e-06,
+      "loss": 0.9522,
+      "step": 16356
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.141024513226496e-06,
+      "loss": 0.9359,
+      "step": 16357
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1404535724882296e-06,
+      "loss": 1.0154,
+      "step": 16358
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1398826871530275e-06,
+      "loss": 1.0271,
+      "step": 16359
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1393118572319465e-06,
+      "loss": 0.8763,
+      "step": 16360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1387410827360527e-06,
+      "loss": 1.2509,
+      "step": 16361
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1381703636764e-06,
+      "loss": 0.8961,
+      "step": 16362
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1375997000640487e-06,
+      "loss": 0.8658,
+      "step": 16363
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1370290919100524e-06,
+      "loss": 0.8171,
+      "step": 16364
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1364585392254694e-06,
+      "loss": 0.8491,
+      "step": 16365
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1358880420213514e-06,
+      "loss": 1.0871,
+      "step": 16366
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.135317600308754e-06,
+      "loss": 0.9479,
+      "step": 16367
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.134747214098727e-06,
+      "loss": 0.8897,
+      "step": 16368
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1341768834023245e-06,
+      "loss": 1.0618,
+      "step": 16369
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.133606608230593e-06,
+      "loss": 0.9708,
+      "step": 16370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.133036388594585e-06,
+      "loss": 1.0987,
+      "step": 16371
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.132466224505343e-06,
+      "loss": 0.9116,
+      "step": 16372
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.131896115973921e-06,
+      "loss": 0.9374,
+      "step": 16373
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.131326063011359e-06,
+      "loss": 0.8314,
+      "step": 16374
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.130756065628706e-06,
+      "loss": 0.9687,
+      "step": 16375
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.130186123837002e-06,
+      "loss": 1.0311,
+      "step": 16376
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1296162376472922e-06,
+      "loss": 0.9132,
+      "step": 16377
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.129046407070615e-06,
+      "loss": 0.9831,
+      "step": 16378
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1284766321180145e-06,
+      "loss": 1.067,
+      "step": 16379
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1279069128005254e-06,
+      "loss": 0.7014,
+      "step": 16380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1273372491291906e-06,
+      "loss": 0.9885,
+      "step": 16381
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1267676411150427e-06,
+      "loss": 1.0077,
+      "step": 16382
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1261980887691207e-06,
+      "loss": 0.9529,
+      "step": 16383
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1256285921024605e-06,
+      "loss": 0.6954,
+      "step": 16384
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1250591511260922e-06,
+      "loss": 1.0598,
+      "step": 16385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1244897658510527e-06,
+      "loss": 0.8016,
+      "step": 16386
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.12392043628837e-06,
+      "loss": 0.8373,
+      "step": 16387
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.123351162449078e-06,
+      "loss": 1.0304,
+      "step": 16388
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.122781944344203e-06,
+      "loss": 0.7551,
+      "step": 16389
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.122212781984777e-06,
+      "loss": 0.9348,
+      "step": 16390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1216436753818237e-06,
+      "loss": 1.0,
+      "step": 16391
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1210746245463733e-06,
+      "loss": 1.0395,
+      "step": 16392
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1205056294894464e-06,
+      "loss": 0.9696,
+      "step": 16393
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1199366902220727e-06,
+      "loss": 0.9505,
+      "step": 16394
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1193678067552682e-06,
+      "loss": 0.9192,
+      "step": 16395
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1187989791000623e-06,
+      "loss": 0.9727,
+      "step": 16396
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.118230207267471e-06,
+      "loss": 1.1296,
+      "step": 16397
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1176614912685177e-06,
+      "loss": 0.9427,
+      "step": 16398
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1170928311142177e-06,
+      "loss": 0.9907,
+      "step": 16399
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.116524226815591e-06,
+      "loss": 0.8086,
+      "step": 16400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1159556783836522e-06,
+      "loss": 0.9451,
+      "step": 16401
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1153871858294194e-06,
+      "loss": 0.8306,
+      "step": 16402
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.114818749163904e-06,
+      "loss": 0.9843,
+      "step": 16403
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1142503683981226e-06,
+      "loss": 0.9078,
+      "step": 16404
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.113682043543084e-06,
+      "loss": 1.0138,
+      "step": 16405
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.113113774609803e-06,
+      "loss": 0.9709,
+      "step": 16406
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1125455616092853e-06,
+      "loss": 0.9116,
+      "step": 16407
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1119774045525434e-06,
+      "loss": 0.9904,
+      "step": 16408
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.111409303450585e-06,
+      "loss": 0.8534,
+      "step": 16409
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.110841258314415e-06,
+      "loss": 0.934,
+      "step": 16410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1102732691550416e-06,
+      "loss": 0.9306,
+      "step": 16411
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.109705335983466e-06,
+      "loss": 1.0442,
+      "step": 16412
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1091374588106944e-06,
+      "loss": 1.0184,
+      "step": 16413
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1085696376477306e-06,
+      "loss": 0.9062,
+      "step": 16414
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1080018725055724e-06,
+      "loss": 0.8992,
+      "step": 16415
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1074341633952238e-06,
+      "loss": 0.9374,
+      "step": 16416
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1068665103276797e-06,
+      "loss": 0.8121,
+      "step": 16417
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1062989133139416e-06,
+      "loss": 0.9572,
+      "step": 16418
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1057313723650067e-06,
+      "loss": 1.1464,
+      "step": 16419
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1051638874918684e-06,
+      "loss": 0.908,
+      "step": 16420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1045964587055253e-06,
+      "loss": 1.0208,
+      "step": 16421
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1040290860169664e-06,
+      "loss": 0.9289,
+      "step": 16422
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1034617694371897e-06,
+      "loss": 1.1105,
+      "step": 16423
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.102894508977182e-06,
+      "loss": 0.9901,
+      "step": 16424
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1023273046479376e-06,
+      "loss": 0.8545,
+      "step": 16425
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1017601564604427e-06,
+      "loss": 1.043,
+      "step": 16426
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1011930644256897e-06,
+      "loss": 1.0794,
+      "step": 16427
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1006260285546607e-06,
+      "loss": 0.8605,
+      "step": 16428
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1000590488583473e-06,
+      "loss": 0.8154,
+      "step": 16429
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.099492125347728e-06,
+      "loss": 0.8414,
+      "step": 16430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0989252580337944e-06,
+      "loss": 0.7865,
+      "step": 16431
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0983584469275243e-06,
+      "loss": 0.8925,
+      "step": 16432
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.097791692039903e-06,
+      "loss": 0.9185,
+      "step": 16433
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.097224993381907e-06,
+      "loss": 0.9427,
+      "step": 16434
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0966583509645207e-06,
+      "loss": 0.8421,
+      "step": 16435
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0960917647987185e-06,
+      "loss": 0.9521,
+      "step": 16436
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0955252348954806e-06,
+      "loss": 0.9203,
+      "step": 16437
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0949587612657812e-06,
+      "loss": 1.0771,
+      "step": 16438
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0943923439205987e-06,
+      "loss": 0.9983,
+      "step": 16439
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0938259828709034e-06,
+      "loss": 0.9484,
+      "step": 16440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0932596781276724e-06,
+      "loss": 0.9178,
+      "step": 16441
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.092693429701874e-06,
+      "loss": 0.9807,
+      "step": 16442
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.092127237604481e-06,
+      "loss": 0.9994,
+      "step": 16443
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.091561101846465e-06,
+      "loss": 0.7107,
+      "step": 16444
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.090995022438791e-06,
+      "loss": 0.9698,
+      "step": 16445
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0904289993924297e-06,
+      "loss": 0.8207,
+      "step": 16446
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0898630327183456e-06,
+      "loss": 0.9723,
+      "step": 16447
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0892971224275064e-06,
+      "loss": 0.9764,
+      "step": 16448
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.088731268530874e-06,
+      "loss": 1.0137,
+      "step": 16449
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0881654710394145e-06,
+      "loss": 1.0691,
+      "step": 16450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.087599729964086e-06,
+      "loss": 0.9159,
+      "step": 16451
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0870340453158546e-06,
+      "loss": 0.9259,
+      "step": 16452
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.086468417105674e-06,
+      "loss": 1.0621,
+      "step": 16453
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0859028453445106e-06,
+      "loss": 0.9732,
+      "step": 16454
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0853373300433167e-06,
+      "loss": 0.8895,
+      "step": 16455
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0847718712130526e-06,
+      "loss": 0.8329,
+      "step": 16456
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0842064688646703e-06,
+      "loss": 0.8813,
+      "step": 16457
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0836411230091285e-06,
+      "loss": 1.1492,
+      "step": 16458
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.083075833657376e-06,
+      "loss": 0.9187,
+      "step": 16459
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.08251060082037e-06,
+      "loss": 1.0264,
+      "step": 16460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.081945424509057e-06,
+      "loss": 1.0835,
+      "step": 16461
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0813803047343918e-06,
+      "loss": 0.8324,
+      "step": 16462
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0808152415073187e-06,
+      "loss": 1.0144,
+      "step": 16463
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.08025023483879e-06,
+      "loss": 0.9909,
+      "step": 16464
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0796852847397475e-06,
+      "loss": 1.0174,
+      "step": 16465
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0791203912211435e-06,
+      "loss": 0.963,
+      "step": 16466
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0785555542939173e-06,
+      "loss": 0.8818,
+      "step": 16467
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.077990773969017e-06,
+      "loss": 1.0177,
+      "step": 16468
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0774260502573808e-06,
+      "loss": 1.0344,
+      "step": 16469
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.076861383169953e-06,
+      "loss": 0.955,
+      "step": 16470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.076296772717671e-06,
+      "loss": 0.7775,
+      "step": 16471
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0757322189114787e-06,
+      "loss": 0.9191,
+      "step": 16472
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.075167721762309e-06,
+      "loss": 0.8486,
+      "step": 16473
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.074603281281103e-06,
+      "loss": 1.0503,
+      "step": 16474
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.074038897478793e-06,
+      "loss": 0.979,
+      "step": 16475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.073474570366315e-06,
+      "loss": 0.8631,
+      "step": 16476
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.072910299954606e-06,
+      "loss": 0.9969,
+      "step": 16477
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0723460862545937e-06,
+      "loss": 0.8794,
+      "step": 16478
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.071781929277214e-06,
+      "loss": 1.0135,
+      "step": 16479
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0712178290333934e-06,
+      "loss": 0.9655,
+      "step": 16480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0706537855340648e-06,
+      "loss": 0.8854,
+      "step": 16481
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0700897987901525e-06,
+      "loss": 1.021,
+      "step": 16482
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0695258688125876e-06,
+      "loss": 0.9443,
+      "step": 16483
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0689619956122925e-06,
+      "loss": 1.0241,
+      "step": 16484
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0683981792001956e-06,
+      "loss": 1.0285,
+      "step": 16485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0678344195872174e-06,
+      "loss": 0.8751,
+      "step": 16486
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.067270716784284e-06,
+      "loss": 0.8375,
+      "step": 16487
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.066707070802312e-06,
+      "loss": 0.8977,
+      "step": 16488
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.066143481652228e-06,
+      "loss": 1.1251,
+      "step": 16489
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0655799493449467e-06,
+      "loss": 0.9333,
+      "step": 16490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0650164738913904e-06,
+      "loss": 0.9668,
+      "step": 16491
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.064453055302472e-06,
+      "loss": 0.9002,
+      "step": 16492
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0638896935891123e-06,
+      "loss": 1.0679,
+      "step": 16493
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.063326388762221e-06,
+      "loss": 1.132,
+      "step": 16494
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0627631408327178e-06,
+      "loss": 1.0083,
+      "step": 16495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0621999498115097e-06,
+      "loss": 0.9561,
+      "step": 16496
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.061636815709513e-06,
+      "loss": 0.9657,
+      "step": 16497
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.061073738537635e-06,
+      "loss": 0.9813,
+      "step": 16498
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.060510718306788e-06,
+      "loss": 0.9949,
+      "step": 16499
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.059947755027878e-06,
+      "loss": 1.0474,
+      "step": 16500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.059384848711813e-06,
+      "loss": 0.8022,
+      "step": 16501
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0588219993695007e-06,
+      "loss": 0.9948,
+      "step": 16502
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.058259207011844e-06,
+      "loss": 1.0557,
+      "step": 16503
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0576964716497495e-06,
+      "loss": 0.9123,
+      "step": 16504
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.057133793294116e-06,
+      "loss": 1.0413,
+      "step": 16505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.05657117195585e-06,
+      "loss": 0.74,
+      "step": 16506
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0560086076458485e-06,
+      "loss": 0.9477,
+      "step": 16507
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.055446100375014e-06,
+      "loss": 0.8624,
+      "step": 16508
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0548836501542407e-06,
+      "loss": 0.9935,
+      "step": 16509
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0543212569944315e-06,
+      "loss": 0.9959,
+      "step": 16510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.053758920906476e-06,
+      "loss": 0.8002,
+      "step": 16511
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.053196641901276e-06,
+      "loss": 0.9032,
+      "step": 16512
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0526344199897212e-06,
+      "loss": 0.9985,
+      "step": 16513
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0520722551827076e-06,
+      "loss": 0.8725,
+      "step": 16514
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0515101474911235e-06,
+      "loss": 0.9466,
+      "step": 16515
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.050948096925864e-06,
+      "loss": 0.9802,
+      "step": 16516
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.050386103497814e-06,
+      "loss": 0.8468,
+      "step": 16517
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0498241672178653e-06,
+      "loss": 0.9461,
+      "step": 16518
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.049262288096903e-06,
+      "loss": 0.8215,
+      "step": 16519
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.048700466145816e-06,
+      "loss": 0.8572,
+      "step": 16520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0481387013754868e-06,
+      "loss": 0.82,
+      "step": 16521
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.047576993796802e-06,
+      "loss": 0.9961,
+      "step": 16522
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.04701534342064e-06,
+      "loss": 1.0346,
+      "step": 16523
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0464537502578885e-06,
+      "loss": 0.9286,
+      "step": 16524
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0458922143194247e-06,
+      "loss": 0.8889,
+      "step": 16525
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0453307356161304e-06,
+      "loss": 0.8838,
+      "step": 16526
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0447693141588815e-06,
+      "loss": 1.0549,
+      "step": 16527
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0442079499585586e-06,
+      "loss": 0.9338,
+      "step": 16528
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0436466430260345e-06,
+      "loss": 0.9419,
+      "step": 16529
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0430853933721874e-06,
+      "loss": 0.9714,
+      "step": 16530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0425242010078884e-06,
+      "loss": 0.9397,
+      "step": 16531
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0419630659440143e-06,
+      "loss": 1.1053,
+      "step": 16532
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0414019881914325e-06,
+      "loss": 1.0018,
+      "step": 16533
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.040840967761018e-06,
+      "loss": 1.0619,
+      "step": 16534
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.040280004663637e-06,
+      "loss": 1.0679,
+      "step": 16535
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.039719098910159e-06,
+      "loss": 1.063,
+      "step": 16536
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0391582505114544e-06,
+      "loss": 1.0585,
+      "step": 16537
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0385974594783848e-06,
+      "loss": 0.9054,
+      "step": 16538
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0380367258218197e-06,
+      "loss": 0.9911,
+      "step": 16539
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.037476049552619e-06,
+      "loss": 1.043,
+      "step": 16540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0369154306816495e-06,
+      "loss": 1.0044,
+      "step": 16541
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0363548692197705e-06,
+      "loss": 0.9033,
+      "step": 16542
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.035794365177845e-06,
+      "loss": 0.8433,
+      "step": 16543
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.035233918566729e-06,
+      "loss": 0.9472,
+      "step": 16544
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0346735293972858e-06,
+      "loss": 0.9748,
+      "step": 16545
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.034113197680366e-06,
+      "loss": 0.9129,
+      "step": 16546
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0335529234268346e-06,
+      "loss": 0.8529,
+      "step": 16547
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.03299270664754e-06,
+      "loss": 0.7863,
+      "step": 16548
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.032432547353341e-06,
+      "loss": 0.9238,
+      "step": 16549
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.031872445555086e-06,
+      "loss": 1.0934,
+      "step": 16550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.031312401263631e-06,
+      "loss": 0.9025,
+      "step": 16551
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.030752414489823e-06,
+      "loss": 0.7669,
+      "step": 16552
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0301924852445153e-06,
+      "loss": 0.9486,
+      "step": 16553
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.029632613538553e-06,
+      "loss": 0.8542,
+      "step": 16554
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0290727993827864e-06,
+      "loss": 0.7627,
+      "step": 16555
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0285130427880585e-06,
+      "loss": 0.9202,
+      "step": 16556
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0279533437652186e-06,
+      "loss": 1.0074,
+      "step": 16557
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0273937023251045e-06,
+      "loss": 0.8193,
+      "step": 16558
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0268341184785674e-06,
+      "loss": 0.8627,
+      "step": 16559
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0262745922364418e-06,
+      "loss": 1.0496,
+      "step": 16560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.025715123609574e-06,
+      "loss": 1.0994,
+      "step": 16561
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0251557126087994e-06,
+      "loss": 0.9811,
+      "step": 16562
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0245963592449595e-06,
+      "loss": 0.9846,
+      "step": 16563
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0240370635288885e-06,
+      "loss": 1.0256,
+      "step": 16564
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.023477825471426e-06,
+      "loss": 0.8332,
+      "step": 16565
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.022918645083404e-06,
+      "loss": 0.947,
+      "step": 16566
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0223595223756593e-06,
+      "loss": 1.0162,
+      "step": 16567
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0218004573590222e-06,
+      "loss": 0.7337,
+      "step": 16568
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.021241450044325e-06,
+      "loss": 0.9318,
+      "step": 16569
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0206825004424015e-06,
+      "loss": 1.0442,
+      "step": 16570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.020123608564076e-06,
+      "loss": 0.7426,
+      "step": 16571
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0195647744201826e-06,
+      "loss": 0.9204,
+      "step": 16572
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0190059980215426e-06,
+      "loss": 0.9914,
+      "step": 16573
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0184472793789876e-06,
+      "loss": 0.9293,
+      "step": 16574
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0178886185033385e-06,
+      "loss": 0.8514,
+      "step": 16575
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0173300154054224e-06,
+      "loss": 0.8972,
+      "step": 16576
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0167714700960594e-06,
+      "loss": 0.9647,
+      "step": 16577
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0162129825860745e-06,
+      "loss": 0.9875,
+      "step": 16578
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0156545528862835e-06,
+      "loss": 1.0699,
+      "step": 16579
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.015096181007511e-06,
+      "loss": 0.9124,
+      "step": 16580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.014537866960569e-06,
+      "loss": 0.797,
+      "step": 16581
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0139796107562827e-06,
+      "loss": 1.0282,
+      "step": 16582
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0134214124054614e-06,
+      "loss": 0.9584,
+      "step": 16583
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.012863271918925e-06,
+      "loss": 0.9104,
+      "step": 16584
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0123051893074834e-06,
+      "loss": 0.9088,
+      "step": 16585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0117471645819533e-06,
+      "loss": 1.0454,
+      "step": 16586
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.011189197753142e-06,
+      "loss": 1.0795,
+      "step": 16587
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.010631288831863e-06,
+      "loss": 0.9724,
+      "step": 16588
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.010073437828923e-06,
+      "loss": 0.8778,
+      "step": 16589
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0095156447551345e-06,
+      "loss": 0.9078,
+      "step": 16590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0089579096213e-06,
+      "loss": 0.9422,
+      "step": 16591
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0084002324382295e-06,
+      "loss": 0.9404,
+      "step": 16592
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.007842613216724e-06,
+      "loss": 0.9655,
+      "step": 16593
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0072850519675885e-06,
+      "loss": 0.8478,
+      "step": 16594
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0067275487016285e-06,
+      "loss": 1.1796,
+      "step": 16595
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0061701034296414e-06,
+      "loss": 0.964,
+      "step": 16596
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0056127161624316e-06,
+      "loss": 0.8564,
+      "step": 16597
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.005055386910794e-06,
+      "loss": 1.0104,
+      "step": 16598
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.004498115685531e-06,
+      "loss": 0.8935,
+      "step": 16599
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0039409024974353e-06,
+      "loss": 0.9652,
+      "step": 16600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.003383747357307e-06,
+      "loss": 1.0301,
+      "step": 16601
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.002826650275937e-06,
+      "loss": 0.9376,
+      "step": 16602
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0022696112641225e-06,
+      "loss": 0.9027,
+      "step": 16603
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0017126303326507e-06,
+      "loss": 0.8553,
+      "step": 16604
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0011557074923205e-06,
+      "loss": 0.9458,
+      "step": 16605
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0005988427539163e-06,
+      "loss": 0.8564,
+      "step": 16606
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.000042036128231e-06,
+      "loss": 0.8779,
+      "step": 16607
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9994852876260486e-06,
+      "loss": 0.6433,
+      "step": 16608
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9989285972581595e-06,
+      "loss": 0.9296,
+      "step": 16609
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.998371965035347e-06,
+      "loss": 1.0249,
+      "step": 16610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.997815390968399e-06,
+      "loss": 1.0369,
+      "step": 16611
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.997258875068094e-06,
+      "loss": 0.9689,
+      "step": 16612
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9967024173452194e-06,
+      "loss": 0.8853,
+      "step": 16613
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9961460178105523e-06,
+      "loss": 1.0298,
+      "step": 16614
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9955896764748766e-06,
+      "loss": 0.8567,
+      "step": 16615
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9950333933489664e-06,
+      "loss": 0.9744,
+      "step": 16616
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9944771684436054e-06,
+      "loss": 1.0106,
+      "step": 16617
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9939210017695658e-06,
+      "loss": 0.854,
+      "step": 16618
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9933648933376263e-06,
+      "loss": 1.0561,
+      "step": 16619
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.992808843158559e-06,
+      "loss": 0.9988,
+      "step": 16620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.99225285124314e-06,
+      "loss": 0.9888,
+      "step": 16621
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.991696917602137e-06,
+      "loss": 0.8252,
+      "step": 16622
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.991141042246327e-06,
+      "loss": 1.0059,
+      "step": 16623
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9905852251864736e-06,
+      "loss": 0.9718,
+      "step": 16624
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9900294664333513e-06,
+      "loss": 0.9631,
+      "step": 16625
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9894737659977237e-06,
+      "loss": 1.1278,
+      "step": 16626
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9889181238903603e-06,
+      "loss": 0.8835,
+      "step": 16627
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9883625401220235e-06,
+      "loss": 0.8003,
+      "step": 16628
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.98780701470348e-06,
+      "loss": 0.9356,
+      "step": 16629
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9872515476454934e-06,
+      "loss": 1.1518,
+      "step": 16630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.986696138958823e-06,
+      "loss": 1.0614,
+      "step": 16631
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9861407886542335e-06,
+      "loss": 0.9513,
+      "step": 16632
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9855854967424803e-06,
+      "loss": 0.9733,
+      "step": 16633
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.985030263234326e-06,
+      "loss": 0.8573,
+      "step": 16634
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.984475088140525e-06,
+      "loss": 0.8674,
+      "step": 16635
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.983919971471837e-06,
+      "loss": 0.9206,
+      "step": 16636
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9833649132390137e-06,
+      "loss": 1.0194,
+      "step": 16637
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.982809913452812e-06,
+      "loss": 1.1393,
+      "step": 16638
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.982254972123981e-06,
+      "loss": 0.885,
+      "step": 16639
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.981700089263279e-06,
+      "loss": 1.0035,
+      "step": 16640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9811452648814507e-06,
+      "loss": 1.0941,
+      "step": 16641
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9805904989892504e-06,
+      "loss": 0.8397,
+      "step": 16642
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.980035791597422e-06,
+      "loss": 0.9149,
+      "step": 16643
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.979481142716717e-06,
+      "loss": 0.9379,
+      "step": 16644
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.978926552357878e-06,
+      "loss": 0.9991,
+      "step": 16645
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9783720205316535e-06,
+      "loss": 0.7843,
+      "step": 16646
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9778175472487843e-06,
+      "loss": 0.9158,
+      "step": 16647
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9772631325200158e-06,
+      "loss": 1.1015,
+      "step": 16648
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9767087763560873e-06,
+      "loss": 0.9544,
+      "step": 16649
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.976154478767742e-06,
+      "loss": 0.9565,
+      "step": 16650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.975600239765714e-06,
+      "loss": 0.9705,
+      "step": 16651
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9750460593607484e-06,
+      "loss": 0.89,
+      "step": 16652
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9744919375635786e-06,
+      "loss": 0.9339,
+      "step": 16653
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.973937874384942e-06,
+      "loss": 0.8942,
+      "step": 16654
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.973383869835571e-06,
+      "loss": 0.9078,
+      "step": 16655
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9728299239262032e-06,
+      "loss": 0.8979,
+      "step": 16656
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.972276036667567e-06,
+      "loss": 0.91,
+      "step": 16657
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9717222080703975e-06,
+      "loss": 1.0019,
+      "step": 16658
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9711684381454215e-06,
+      "loss": 0.8486,
+      "step": 16659
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9706147269033713e-06,
+      "loss": 0.9261,
+      "step": 16660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9700610743549725e-06,
+      "loss": 0.9718,
+      "step": 16661
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9695074805109527e-06,
+      "loss": 1.0033,
+      "step": 16662
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.96895394538204e-06,
+      "loss": 0.8692,
+      "step": 16663
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9684004689789554e-06,
+      "loss": 0.9105,
+      "step": 16664
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9678470513124253e-06,
+      "loss": 1.0708,
+      "step": 16665
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9672936923931696e-06,
+      "loss": 0.9682,
+      "step": 16666
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9667403922319123e-06,
+      "loss": 0.9364,
+      "step": 16667
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.96618715083937e-06,
+      "loss": 1.0728,
+      "step": 16668
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9656339682262648e-06,
+      "loss": 1.0238,
+      "step": 16669
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.965080844403312e-06,
+      "loss": 0.8787,
+      "step": 16670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9645277793812305e-06,
+      "loss": 0.9443,
+      "step": 16671
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9639747731707335e-06,
+      "loss": 0.8292,
+      "step": 16672
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.963421825782538e-06,
+      "loss": 0.8845,
+      "step": 16673
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.962868937227353e-06,
+      "loss": 0.8951,
+      "step": 16674
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9623161075158963e-06,
+      "loss": 0.9397,
+      "step": 16675
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9617633366588747e-06,
+      "loss": 1.176,
+      "step": 16676
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9612106246670003e-06,
+      "loss": 0.8882,
+      "step": 16677
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9606579715509795e-06,
+      "loss": 1.0349,
+      "step": 16678
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9601053773215227e-06,
+      "loss": 0.8879,
+      "step": 16679
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.959552841989333e-06,
+      "loss": 0.7591,
+      "step": 16680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.959000365565119e-06,
+      "loss": 1.1489,
+      "step": 16681
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9584479480595808e-06,
+      "loss": 0.8271,
+      "step": 16682
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.957895589483426e-06,
+      "loss": 0.6466,
+      "step": 16683
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.957343289847353e-06,
+      "loss": 0.8933,
+      "step": 16684
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.956791049162064e-06,
+      "loss": 0.8346,
+      "step": 16685
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.956238867438257e-06,
+      "loss": 0.8948,
+      "step": 16686
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.955686744686631e-06,
+      "loss": 0.8322,
+      "step": 16687
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.955134680917886e-06,
+      "loss": 0.9223,
+      "step": 16688
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9545826761427146e-06,
+      "loss": 0.7884,
+      "step": 16689
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9540307303718137e-06,
+      "loss": 0.7757,
+      "step": 16690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.953478843615875e-06,
+      "loss": 1.1099,
+      "step": 16691
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.952927015885595e-06,
+      "loss": 1.084,
+      "step": 16692
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9523752471916605e-06,
+      "loss": 1.0492,
+      "step": 16693
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9518235375447665e-06,
+      "loss": 0.909,
+      "step": 16694
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.951271886955598e-06,
+      "loss": 0.9533,
+      "step": 16695
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.950720295434847e-06,
+      "loss": 0.9361,
+      "step": 16696
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.950168762993195e-06,
+      "loss": 0.8903,
+      "step": 16697
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9496172896413347e-06,
+      "loss": 0.8526,
+      "step": 16698
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.949065875389946e-06,
+      "loss": 0.86,
+      "step": 16699
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9485145202497153e-06,
+      "loss": 0.8536,
+      "step": 16700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9479632242313224e-06,
+      "loss": 1.0077,
+      "step": 16701
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.947411987345451e-06,
+      "loss": 0.9997,
+      "step": 16702
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.946860809602778e-06,
+      "loss": 0.9727,
+      "step": 16703
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9463096910139866e-06,
+      "loss": 0.9599,
+      "step": 16704
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.94575863158975e-06,
+      "loss": 0.9358,
+      "step": 16705
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9452076313407494e-06,
+      "loss": 1.0653,
+      "step": 16706
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9446566902776553e-06,
+      "loss": 0.7864,
+      "step": 16707
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9441058084111473e-06,
+      "loss": 1.0224,
+      "step": 16708
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9435549857518925e-06,
+      "loss": 0.7868,
+      "step": 16709
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.94300422231057e-06,
+      "loss": 0.8658,
+      "step": 16710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9424535180978448e-06,
+      "loss": 0.7603,
+      "step": 16711
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9419028731243916e-06,
+      "loss": 0.8112,
+      "step": 16712
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.941352287400875e-06,
+      "loss": 1.0156,
+      "step": 16713
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9408017609379655e-06,
+      "loss": 0.914,
+      "step": 16714
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.940251293746326e-06,
+      "loss": 0.8604,
+      "step": 16715
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.939700885836625e-06,
+      "loss": 0.8743,
+      "step": 16716
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.939150537219524e-06,
+      "loss": 0.8692,
+      "step": 16717
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.938600247905689e-06,
+      "loss": 0.9404,
+      "step": 16718
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9380500179057776e-06,
+      "loss": 0.9123,
+      "step": 16719
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9374998472304523e-06,
+      "loss": 0.9408,
+      "step": 16720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.936949735890375e-06,
+      "loss": 0.8539,
+      "step": 16721
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9363996838961994e-06,
+      "loss": 0.9438,
+      "step": 16722
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9358496912585874e-06,
+      "loss": 0.869,
+      "step": 16723
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9352997579881903e-06,
+      "loss": 0.9687,
+      "step": 16724
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9347498840956673e-06,
+      "loss": 0.707,
+      "step": 16725
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9342000695916684e-06,
+      "loss": 0.8898,
+      "step": 16726
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.933650314486849e-06,
+      "loss": 1.0166,
+      "step": 16727
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9331006187918577e-06,
+      "loss": 0.9438,
+      "step": 16728
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9325509825173476e-06,
+      "loss": 1.0078,
+      "step": 16729
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9320014056739657e-06,
+      "loss": 0.9514,
+      "step": 16730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.931451888272361e-06,
+      "loss": 0.9629,
+      "step": 16731
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.930902430323177e-06,
+      "loss": 0.9018,
+      "step": 16732
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.930353031837066e-06,
+      "loss": 1.0737,
+      "step": 16733
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9298036928246666e-06,
+      "loss": 1.0269,
+      "step": 16734
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9292544132966263e-06,
+      "loss": 0.8715,
+      "step": 16735
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.928705193263583e-06,
+      "loss": 0.8928,
+      "step": 16736
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.928156032736182e-06,
+      "loss": 0.9421,
+      "step": 16737
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.927606931725059e-06,
+      "loss": 1.0823,
+      "step": 16738
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9270578902408564e-06,
+      "loss": 0.9451,
+      "step": 16739
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.926508908294209e-06,
+      "loss": 1.1228,
+      "step": 16740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9259599858957554e-06,
+      "loss": 1.0104,
+      "step": 16741
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9254111230561282e-06,
+      "loss": 1.0389,
+      "step": 16742
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.924862319785965e-06,
+      "loss": 0.8627,
+      "step": 16743
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9243135760958935e-06,
+      "loss": 0.8645,
+      "step": 16744
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.923764891996552e-06,
+      "loss": 0.8515,
+      "step": 16745
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.923216267498566e-06,
+      "loss": 1.025,
+      "step": 16746
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.922667702612569e-06,
+      "loss": 1.1419,
+      "step": 16747
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9221191973491854e-06,
+      "loss": 1.0032,
+      "step": 16748
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9215707517190457e-06,
+      "loss": 0.9855,
+      "step": 16749
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9210223657327725e-06,
+      "loss": 0.824,
+      "step": 16750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9204740394009947e-06,
+      "loss": 0.9913,
+      "step": 16751
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9199257727343317e-06,
+      "loss": 1.0605,
+      "step": 16752
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.91937756574341e-06,
+      "loss": 0.9142,
+      "step": 16753
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9188294184388473e-06,
+      "loss": 1.1408,
+      "step": 16754
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.918281330831266e-06,
+      "loss": 1.0388,
+      "step": 16755
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9177333029312866e-06,
+      "loss": 0.814,
+      "step": 16756
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9171853347495234e-06,
+      "loss": 0.9609,
+      "step": 16757
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.916637426296597e-06,
+      "loss": 1.0341,
+      "step": 16758
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.916089577583119e-06,
+      "loss": 1.0825,
+      "step": 16759
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9155417886197076e-06,
+      "loss": 0.9852,
+      "step": 16760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.914994059416973e-06,
+      "loss": 0.8952,
+      "step": 16761
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9144463899855303e-06,
+      "loss": 0.9349,
+      "step": 16762
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9138987803359873e-06,
+      "loss": 1.0068,
+      "step": 16763
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9133512304789566e-06,
+      "loss": 1.0419,
+      "step": 16764
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9128037404250442e-06,
+      "loss": 0.7447,
+      "step": 16765
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9122563101848607e-06,
+      "loss": 1.0035,
+      "step": 16766
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9117089397690074e-06,
+      "loss": 0.8112,
+      "step": 16767
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9111616291880964e-06,
+      "loss": 0.8882,
+      "step": 16768
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.910614378452726e-06,
+      "loss": 0.9096,
+      "step": 16769
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9100671875735034e-06,
+      "loss": 0.8056,
+      "step": 16770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.909520056561026e-06,
+      "loss": 0.9035,
+      "step": 16771
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9089729854258988e-06,
+      "loss": 0.8284,
+      "step": 16772
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.908425974178717e-06,
+      "loss": 1.045,
+      "step": 16773
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.907879022830082e-06,
+      "loss": 0.9871,
+      "step": 16774
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.907332131390588e-06,
+      "loss": 0.8154,
+      "step": 16775
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9067852998708342e-06,
+      "loss": 1.0248,
+      "step": 16776
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9062385282814117e-06,
+      "loss": 0.9098,
+      "step": 16777
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9056918166329174e-06,
+      "loss": 0.8735,
+      "step": 16778
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9051451649359408e-06,
+      "loss": 0.9754,
+      "step": 16779
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9045985732010746e-06,
+      "loss": 0.9343,
+      "step": 16780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9040520414389108e-06,
+      "loss": 0.9481,
+      "step": 16781
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9035055696600346e-06,
+      "loss": 1.0088,
+      "step": 16782
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9029591578750373e-06,
+      "loss": 1.1582,
+      "step": 16783
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9024128060945018e-06,
+      "loss": 0.9853,
+      "step": 16784
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9018665143290177e-06,
+      "loss": 1.0318,
+      "step": 16785
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9013202825891652e-06,
+      "loss": 0.9796,
+      "step": 16786
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.900774110885531e-06,
+      "loss": 0.9132,
+      "step": 16787
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9002279992286937e-06,
+      "loss": 0.9644,
+      "step": 16788
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8996819476292371e-06,
+      "loss": 0.9024,
+      "step": 16789
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8991359560977367e-06,
+      "loss": 0.9995,
+      "step": 16790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8985900246447769e-06,
+      "loss": 1.0658,
+      "step": 16791
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8980441532809302e-06,
+      "loss": 1.1008,
+      "step": 16792
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8974983420167759e-06,
+      "loss": 1.0107,
+      "step": 16793
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.896952590862886e-06,
+      "loss": 0.7557,
+      "step": 16794
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8964068998298374e-06,
+      "loss": 0.9531,
+      "step": 16795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.895861268928199e-06,
+      "loss": 1.0073,
+      "step": 16796
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8953156981685467e-06,
+      "loss": 0.8929,
+      "step": 16797
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8947701875614467e-06,
+      "loss": 1.0096,
+      "step": 16798
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8942247371174717e-06,
+      "loss": 1.07,
+      "step": 16799
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8936793468471854e-06,
+      "loss": 1.2209,
+      "step": 16800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8931340167611596e-06,
+      "loss": 0.8372,
+      "step": 16801
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8925887468699538e-06,
+      "loss": 0.9496,
+      "step": 16802
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8920435371841394e-06,
+      "loss": 1.1058,
+      "step": 16803
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8914983877142751e-06,
+      "loss": 0.9797,
+      "step": 16804
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8909532984709262e-06,
+      "loss": 1.1871,
+      "step": 16805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8904082694646496e-06,
+      "loss": 0.8437,
+      "step": 16806
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8898633007060095e-06,
+      "loss": 0.8891,
+      "step": 16807
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8893183922055608e-06,
+      "loss": 1.0298,
+      "step": 16808
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8887735439738648e-06,
+      "loss": 0.9937,
+      "step": 16809
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.888228756021473e-06,
+      "loss": 0.8506,
+      "step": 16810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8876840283589453e-06,
+      "loss": 0.9835,
+      "step": 16811
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.887139360996832e-06,
+      "loss": 0.9328,
+      "step": 16812
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8865947539456875e-06,
+      "loss": 1.1431,
+      "step": 16813
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8860502072160647e-06,
+      "loss": 1.0149,
+      "step": 16814
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8855057208185112e-06,
+      "loss": 0.876,
+      "step": 16815
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8849612947635798e-06,
+      "loss": 0.8861,
+      "step": 16816
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8844169290618148e-06,
+      "loss": 1.0726,
+      "step": 16817
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.883872623723767e-06,
+      "loss": 1.003,
+      "step": 16818
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8833283787599781e-06,
+      "loss": 0.9052,
+      "step": 16819
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.882784194180997e-06,
+      "loss": 1.0534,
+      "step": 16820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.882240069997363e-06,
+      "loss": 0.8274,
+      "step": 16821
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8816960062196228e-06,
+      "loss": 1.0068,
+      "step": 16822
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.881152002858313e-06,
+      "loss": 0.9376,
+      "step": 16823
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8806080599239774e-06,
+      "loss": 1.0058,
+      "step": 16824
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8800641774271494e-06,
+      "loss": 0.8833,
+      "step": 16825
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8795203553783736e-06,
+      "loss": 0.8216,
+      "step": 16826
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8789765937881816e-06,
+      "loss": 0.806,
+      "step": 16827
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.878432892667112e-06,
+      "loss": 0.9338,
+      "step": 16828
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8778892520256952e-06,
+      "loss": 1.0509,
+      "step": 16829
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8773456718744677e-06,
+      "loss": 1.1143,
+      "step": 16830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8768021522239576e-06,
+      "loss": 0.9051,
+      "step": 16831
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8762586930846988e-06,
+      "loss": 0.8762,
+      "step": 16832
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.875715294467218e-06,
+      "loss": 0.9813,
+      "step": 16833
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8751719563820464e-06,
+      "loss": 1.0781,
+      "step": 16834
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8746286788397072e-06,
+      "loss": 1.0135,
+      "step": 16835
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8740854618507303e-06,
+      "loss": 0.9004,
+      "step": 16836
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8735423054256368e-06,
+      "loss": 0.9108,
+      "step": 16837
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.872999209574952e-06,
+      "loss": 0.9451,
+      "step": 16838
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8724561743091995e-06,
+      "loss": 0.9857,
+      "step": 16839
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8719131996388978e-06,
+      "loss": 0.981,
+      "step": 16840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8713702855745686e-06,
+      "loss": 0.9238,
+      "step": 16841
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8708274321267317e-06,
+      "loss": 0.9423,
+      "step": 16842
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8702846393059021e-06,
+      "loss": 0.823,
+      "step": 16843
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8697419071225993e-06,
+      "loss": 0.924,
+      "step": 16844
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.869199235587335e-06,
+      "loss": 0.9925,
+      "step": 16845
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.868656624710627e-06,
+      "loss": 0.9113,
+      "step": 16846
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8681140745029853e-06,
+      "loss": 0.9096,
+      "step": 16847
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8675715849749226e-06,
+      "loss": 0.9794,
+      "step": 16848
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8670291561369519e-06,
+      "loss": 0.789,
+      "step": 16849
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8664867879995786e-06,
+      "loss": 0.9708,
+      "step": 16850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.865944480573315e-06,
+      "loss": 1.0619,
+      "step": 16851
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.865402233868664e-06,
+      "loss": 0.7569,
+      "step": 16852
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.864860047896136e-06,
+      "loss": 0.9158,
+      "step": 16853
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.864317922666231e-06,
+      "loss": 0.9365,
+      "step": 16854
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.863775858189457e-06,
+      "loss": 0.9852,
+      "step": 16855
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8632338544763129e-06,
+      "loss": 1.0177,
+      "step": 16856
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8626919115373032e-06,
+      "loss": 1.1098,
+      "step": 16857
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8621500293829237e-06,
+      "loss": 0.9402,
+      "step": 16858
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8616082080236776e-06,
+      "loss": 1.0453,
+      "step": 16859
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.861066447470057e-06,
+      "loss": 1.0208,
+      "step": 16860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8605247477325654e-06,
+      "loss": 1.0075,
+      "step": 16861
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.859983108821693e-06,
+      "loss": 0.9479,
+      "step": 16862
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8594415307479375e-06,
+      "loss": 0.9741,
+      "step": 16863
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8589000135217882e-06,
+      "loss": 0.9298,
+      "step": 16864
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8583585571537398e-06,
+      "loss": 0.8863,
+      "step": 16865
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8578171616542806e-06,
+      "loss": 0.8895,
+      "step": 16866
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8572758270339026e-06,
+      "loss": 0.937,
+      "step": 16867
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.856734553303091e-06,
+      "loss": 1.0555,
+      "step": 16868
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.856193340472336e-06,
+      "loss": 0.9026,
+      "step": 16869
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.85565218855212e-06,
+      "loss": 0.9812,
+      "step": 16870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8551110975529318e-06,
+      "loss": 1.0599,
+      "step": 16871
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8545700674852507e-06,
+      "loss": 1.063,
+      "step": 16872
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.854029098359561e-06,
+      "loss": 0.8333,
+      "step": 16873
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8534881901863456e-06,
+      "loss": 0.9306,
+      "step": 16874
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8529473429760815e-06,
+      "loss": 0.9461,
+      "step": 16875
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8524065567392496e-06,
+      "loss": 0.8553,
+      "step": 16876
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8518658314863253e-06,
+      "loss": 1.0247,
+      "step": 16877
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8513251672277882e-06,
+      "loss": 0.8926,
+      "step": 16878
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.85078456397411e-06,
+      "loss": 0.8289,
+      "step": 16879
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8502440217357686e-06,
+      "loss": 0.871,
+      "step": 16880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8497035405232328e-06,
+      "loss": 0.9454,
+      "step": 16881
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8491631203469778e-06,
+      "loss": 0.971,
+      "step": 16882
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8486227612174696e-06,
+      "loss": 0.8295,
+      "step": 16883
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8480824631451837e-06,
+      "loss": 1.0788,
+      "step": 16884
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8475422261405835e-06,
+      "loss": 0.7231,
+      "step": 16885
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.847002050214139e-06,
+      "loss": 0.9278,
+      "step": 16886
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.846461935376313e-06,
+      "loss": 0.832,
+      "step": 16887
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8459218816375734e-06,
+      "loss": 0.9249,
+      "step": 16888
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8453818890083803e-06,
+      "loss": 0.9831,
+      "step": 16889
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8448419574991994e-06,
+      "loss": 0.8039,
+      "step": 16890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8443020871204876e-06,
+      "loss": 0.8562,
+      "step": 16891
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8437622778827092e-06,
+      "loss": 0.7915,
+      "step": 16892
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.843222529796319e-06,
+      "loss": 0.9643,
+      "step": 16893
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8426828428717775e-06,
+      "loss": 0.8168,
+      "step": 16894
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8421432171195363e-06,
+      "loss": 0.9803,
+      "step": 16895
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.841603652550058e-06,
+      "loss": 0.9616,
+      "step": 16896
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.84106414917379e-06,
+      "loss": 0.9932,
+      "step": 16897
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.84052470700119e-06,
+      "loss": 1.069,
+      "step": 16898
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8399853260427047e-06,
+      "loss": 1.0025,
+      "step": 16899
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8394460063087894e-06,
+      "loss": 1.0238,
+      "step": 16900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8389067478098888e-06,
+      "loss": 1.2375,
+      "step": 16901
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8383675505564541e-06,
+      "loss": 0.7831,
+      "step": 16902
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8378284145589299e-06,
+      "loss": 0.8415,
+      "step": 16903
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8372893398277642e-06,
+      "loss": 0.9307,
+      "step": 16904
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8367503263733983e-06,
+      "loss": 1.0767,
+      "step": 16905
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8362113742062776e-06,
+      "loss": 0.851,
+      "step": 16906
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.835672483336845e-06,
+      "loss": 0.8927,
+      "step": 16907
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8351336537755394e-06,
+      "loss": 1.0902,
+      "step": 16908
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.834594885532802e-06,
+      "loss": 1.0122,
+      "step": 16909
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8340561786190697e-06,
+      "loss": 0.9223,
+      "step": 16910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8335175330447825e-06,
+      "loss": 0.7879,
+      "step": 16911
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8329789488203732e-06,
+      "loss": 1.0231,
+      "step": 16912
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8324404259562795e-06,
+      "loss": 0.9226,
+      "step": 16913
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8319019644629326e-06,
+      "loss": 1.0369,
+      "step": 16914
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8313635643507683e-06,
+      "loss": 1.0637,
+      "step": 16915
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8308252256302146e-06,
+      "loss": 0.8848,
+      "step": 16916
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8302869483117047e-06,
+      "loss": 0.9832,
+      "step": 16917
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.829748732405663e-06,
+      "loss": 1.0149,
+      "step": 16918
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8292105779225234e-06,
+      "loss": 0.9127,
+      "step": 16919
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8286724848727083e-06,
+      "loss": 0.9447,
+      "step": 16920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8281344532666462e-06,
+      "loss": 0.869,
+      "step": 16921
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8275964831147574e-06,
+      "loss": 1.0591,
+      "step": 16922
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8270585744274693e-06,
+      "loss": 0.7806,
+      "step": 16923
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8265207272151996e-06,
+      "loss": 0.8716,
+      "step": 16924
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8259829414883728e-06,
+      "loss": 0.8923,
+      "step": 16925
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.825445217257405e-06,
+      "loss": 0.9648,
+      "step": 16926
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8249075545327172e-06,
+      "loss": 0.9115,
+      "step": 16927
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8243699533247239e-06,
+      "loss": 0.8653,
+      "step": 16928
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.823832413643844e-06,
+      "loss": 0.9644,
+      "step": 16929
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8232949355004892e-06,
+      "loss": 0.7768,
+      "step": 16930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.822757518905074e-06,
+      "loss": 0.9776,
+      "step": 16931
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8222201638680132e-06,
+      "loss": 0.9388,
+      "step": 16932
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8216828703997142e-06,
+      "loss": 1.0133,
+      "step": 16933
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8211456385105902e-06,
+      "loss": 0.9985,
+      "step": 16934
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8206084682110465e-06,
+      "loss": 0.9492,
+      "step": 16935
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8200713595114943e-06,
+      "loss": 0.9807,
+      "step": 16936
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8195343124223364e-06,
+      "loss": 0.944,
+      "step": 16937
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.81899732695398e-06,
+      "loss": 0.8467,
+      "step": 16938
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8184604031168307e-06,
+      "loss": 0.892,
+      "step": 16939
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.817923540921287e-06,
+      "loss": 1.0597,
+      "step": 16940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8173867403777533e-06,
+      "loss": 1.1809,
+      "step": 16941
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8168500014966316e-06,
+      "loss": 0.9865,
+      "step": 16942
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8163133242883168e-06,
+      "loss": 0.9686,
+      "step": 16943
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.815776708763211e-06,
+      "loss": 1.0856,
+      "step": 16944
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8152401549317078e-06,
+      "loss": 0.9704,
+      "step": 16945
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8147036628042054e-06,
+      "loss": 0.9558,
+      "step": 16946
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8141672323910957e-06,
+      "loss": 0.8245,
+      "step": 16947
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8136308637027745e-06,
+      "loss": 0.9775,
+      "step": 16948
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8130945567496311e-06,
+      "loss": 0.972,
+      "step": 16949
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8125583115420597e-06,
+      "loss": 1.0558,
+      "step": 16950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.812022128090446e-06,
+      "loss": 1.1226,
+      "step": 16951
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.811486006405182e-06,
+      "loss": 0.7889,
+      "step": 16952
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8109499464966502e-06,
+      "loss": 1.0345,
+      "step": 16953
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8104139483752426e-06,
+      "loss": 0.8996,
+      "step": 16954
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8098780120513398e-06,
+      "loss": 1.0129,
+      "step": 16955
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8093421375353293e-06,
+      "loss": 0.9669,
+      "step": 16956
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8088063248375888e-06,
+      "loss": 1.0027,
+      "step": 16957
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8082705739685036e-06,
+      "loss": 0.8228,
+      "step": 16958
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8077348849384502e-06,
+      "loss": 0.9932,
+      "step": 16959
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8071992577578111e-06,
+      "loss": 0.9735,
+      "step": 16960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8066636924369602e-06,
+      "loss": 0.8977,
+      "step": 16961
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.806128188986278e-06,
+      "loss": 1.2245,
+      "step": 16962
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.805592747416136e-06,
+      "loss": 0.8486,
+      "step": 16963
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8050573677369116e-06,
+      "loss": 0.887,
+      "step": 16964
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.804522049958974e-06,
+      "loss": 0.9675,
+      "step": 16965
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8039867940926965e-06,
+      "loss": 1.0187,
+      "step": 16966
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.803451600148452e-06,
+      "loss": 0.8438,
+      "step": 16967
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8029164681366062e-06,
+      "loss": 1.0117,
+      "step": 16968
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8023813980675297e-06,
+      "loss": 0.7597,
+      "step": 16969
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.801846389951587e-06,
+      "loss": 1.0975,
+      "step": 16970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.801311443799147e-06,
+      "loss": 1.0422,
+      "step": 16971
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.80077655962057e-06,
+      "loss": 1.0728,
+      "step": 16972
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8002417374262232e-06,
+      "loss": 1.0317,
+      "step": 16973
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7997069772264652e-06,
+      "loss": 1.0268,
+      "step": 16974
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7991722790316606e-06,
+      "loss": 0.9197,
+      "step": 16975
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7986376428521634e-06,
+      "loss": 1.1603,
+      "step": 16976
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7981030686983392e-06,
+      "loss": 0.9203,
+      "step": 16977
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7975685565805395e-06,
+      "loss": 1.0476,
+      "step": 16978
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7970341065091246e-06,
+      "loss": 0.9404,
+      "step": 16979
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.796499718494446e-06,
+      "loss": 0.8554,
+      "step": 16980
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.79596539254686e-06,
+      "loss": 1.0094,
+      "step": 16981
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7954311286767163e-06,
+      "loss": 0.8851,
+      "step": 16982
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.794896926894369e-06,
+      "loss": 1.1072,
+      "step": 16983
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.794362787210165e-06,
+      "loss": 1.05,
+      "step": 16984
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7938287096344564e-06,
+      "loss": 0.9875,
+      "step": 16985
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7932946941775882e-06,
+      "loss": 0.9338,
+      "step": 16986
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7927607408499092e-06,
+      "loss": 0.9645,
+      "step": 16987
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7922268496617596e-06,
+      "loss": 1.0533,
+      "step": 16988
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7916930206234912e-06,
+      "loss": 0.9449,
+      "step": 16989
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7911592537454402e-06,
+      "loss": 0.856,
+      "step": 16990
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.790625549037953e-06,
+      "loss": 1.0116,
+      "step": 16991
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7900919065113664e-06,
+      "loss": 0.8547,
+      "step": 16992
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7895583261760223e-06,
+      "loss": 0.8087,
+      "step": 16993
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7890248080422557e-06,
+      "loss": 1.0137,
+      "step": 16994
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7884913521204067e-06,
+      "loss": 0.9731,
+      "step": 16995
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.787957958420808e-06,
+      "loss": 0.9075,
+      "step": 16996
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7874246269537966e-06,
+      "loss": 0.8622,
+      "step": 16997
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7868913577297027e-06,
+      "loss": 0.9456,
+      "step": 16998
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7863581507588595e-06,
+      "loss": 1.0143,
+      "step": 16999
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7858250060516002e-06,
+      "loss": 0.8907,
+      "step": 17000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.785291923618251e-06,
+      "loss": 0.9636,
+      "step": 17001
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7847589034691427e-06,
+      "loss": 0.8722,
+      "step": 17002
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7842259456145995e-06,
+      "loss": 0.9699,
+      "step": 17003
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7836930500649513e-06,
+      "loss": 0.933,
+      "step": 17004
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7831602168305184e-06,
+      "loss": 0.9555,
+      "step": 17005
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7826274459216293e-06,
+      "loss": 0.883,
+      "step": 17006
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7820947373486008e-06,
+      "loss": 0.8952,
+      "step": 17007
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7815620911217591e-06,
+      "loss": 0.9958,
+      "step": 17008
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7810295072514199e-06,
+      "loss": 0.9511,
+      "step": 17009
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.780496985747906e-06,
+      "loss": 0.8688,
+      "step": 17010
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7799645266215288e-06,
+      "loss": 0.9985,
+      "step": 17011
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7794321298826118e-06,
+      "loss": 0.9183,
+      "step": 17012
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7788997955414656e-06,
+      "loss": 0.9527,
+      "step": 17013
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7783675236084064e-06,
+      "loss": 0.9569,
+      "step": 17014
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7778353140937438e-06,
+      "loss": 0.915,
+      "step": 17015
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7773031670077934e-06,
+      "loss": 0.8635,
+      "step": 17016
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7767710823608608e-06,
+      "loss": 0.8305,
+      "step": 17017
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7762390601632595e-06,
+      "loss": 0.8035,
+      "step": 17018
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7757071004252928e-06,
+      "loss": 0.8643,
+      "step": 17019
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7751752031572721e-06,
+      "loss": 0.9529,
+      "step": 17020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7746433683694985e-06,
+      "loss": 0.6916,
+      "step": 17021
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7741115960722792e-06,
+      "loss": 0.8383,
+      "step": 17022
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7735798862759146e-06,
+      "loss": 0.8772,
+      "step": 17023
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7730482389907078e-06,
+      "loss": 0.883,
+      "step": 17024
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7725166542269612e-06,
+      "loss": 0.9298,
+      "step": 17025
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7719851319949705e-06,
+      "loss": 0.8842,
+      "step": 17026
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7714536723050373e-06,
+      "loss": 0.9877,
+      "step": 17027
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7709222751674549e-06,
+      "loss": 0.9281,
+      "step": 17028
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7703909405925228e-06,
+      "loss": 0.7602,
+      "step": 17029
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7698596685905323e-06,
+      "loss": 0.9598,
+      "step": 17030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7693284591717797e-06,
+      "loss": 0.7701,
+      "step": 17031
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7687973123465528e-06,
+      "loss": 0.9125,
+      "step": 17032
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7682662281251478e-06,
+      "loss": 0.8852,
+      "step": 17033
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7677352065178476e-06,
+      "loss": 0.8892,
+      "step": 17034
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7672042475349482e-06,
+      "loss": 0.7435,
+      "step": 17035
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7666733511867311e-06,
+      "loss": 0.9356,
+      "step": 17036
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7661425174834863e-06,
+      "loss": 0.9687,
+      "step": 17037
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.765611746435495e-06,
+      "loss": 0.8891,
+      "step": 17038
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7650810380530443e-06,
+      "loss": 1.0532,
+      "step": 17039
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7645503923464124e-06,
+      "loss": 0.8711,
+      "step": 17040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7640198093258848e-06,
+      "loss": 0.9014,
+      "step": 17041
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.763489289001738e-06,
+      "loss": 0.8353,
+      "step": 17042
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7629588313842533e-06,
+      "loss": 0.8934,
+      "step": 17043
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7624284364837052e-06,
+      "loss": 0.8258,
+      "step": 17044
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.761898104310374e-06,
+      "loss": 0.806,
+      "step": 17045
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7613678348745283e-06,
+      "loss": 0.9672,
+      "step": 17046
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7608376281864502e-06,
+      "loss": 0.8923,
+      "step": 17047
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7603074842564061e-06,
+      "loss": 1.0103,
+      "step": 17048
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7597774030946718e-06,
+      "loss": 0.6483,
+      "step": 17049
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7592473847115131e-06,
+      "loss": 1.0263,
+      "step": 17050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7587174291172033e-06,
+      "loss": 0.9868,
+      "step": 17051
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7581875363220064e-06,
+      "loss": 0.7922,
+      "step": 17052
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.757657706336192e-06,
+      "loss": 0.9481,
+      "step": 17053
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7571279391700225e-06,
+      "loss": 1.1409,
+      "step": 17054
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7565982348337652e-06,
+      "loss": 1.0311,
+      "step": 17055
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7560685933376797e-06,
+      "loss": 0.969,
+      "step": 17056
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7555390146920315e-06,
+      "loss": 0.8603,
+      "step": 17057
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7550094989070764e-06,
+      "loss": 0.925,
+      "step": 17058
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7544800459930767e-06,
+      "loss": 1.0096,
+      "step": 17059
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7539506559602915e-06,
+      "loss": 1.0364,
+      "step": 17060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7534213288189744e-06,
+      "loss": 0.8537,
+      "step": 17061
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7528920645793845e-06,
+      "loss": 0.9651,
+      "step": 17062
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7523628632517719e-06,
+      "loss": 1.0433,
+      "step": 17063
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7518337248463945e-06,
+      "loss": 1.0636,
+      "step": 17064
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7513046493734997e-06,
+      "loss": 0.8645,
+      "step": 17065
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7507756368433426e-06,
+      "loss": 1.0485,
+      "step": 17066
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7502466872661683e-06,
+      "loss": 1.0666,
+      "step": 17067
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7497178006522293e-06,
+      "loss": 0.9766,
+      "step": 17068
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.749188977011767e-06,
+      "loss": 0.8183,
+      "step": 17069
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7486602163550343e-06,
+      "loss": 0.9489,
+      "step": 17070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.748131518692271e-06,
+      "loss": 0.9847,
+      "step": 17071
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7476028840337234e-06,
+      "loss": 1.0512,
+      "step": 17072
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.74707431238963e-06,
+      "loss": 0.8976,
+      "step": 17073
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.746545803770236e-06,
+      "loss": 0.9361,
+      "step": 17074
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7460173581857776e-06,
+      "loss": 0.9861,
+      "step": 17075
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.745488975646496e-06,
+      "loss": 1.0332,
+      "step": 17076
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.744960656162626e-06,
+      "loss": 0.8591,
+      "step": 17077
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7444323997444063e-06,
+      "loss": 0.8476,
+      "step": 17078
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7439042064020683e-06,
+      "loss": 0.9037,
+      "step": 17079
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7433760761458501e-06,
+      "loss": 0.9812,
+      "step": 17080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7428480089859778e-06,
+      "loss": 0.817,
+      "step": 17081
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7423200049326904e-06,
+      "loss": 0.9191,
+      "step": 17082
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.741792063996212e-06,
+      "loss": 1.0157,
+      "step": 17083
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7412641861867746e-06,
+      "loss": 0.9034,
+      "step": 17084
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.740736371514603e-06,
+      "loss": 0.9791,
+      "step": 17085
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7402086199899265e-06,
+      "loss": 1.0439,
+      "step": 17086
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7396809316229673e-06,
+      "loss": 0.9061,
+      "step": 17087
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7391533064239519e-06,
+      "loss": 0.9788,
+      "step": 17088
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7386257444030996e-06,
+      "loss": 1.0204,
+      "step": 17089
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7380982455706353e-06,
+      "loss": 0.9646,
+      "step": 17090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.737570809936776e-06,
+      "loss": 0.9595,
+      "step": 17091
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7370434375117423e-06,
+      "loss": 0.9017,
+      "step": 17092
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7365161283057536e-06,
+      "loss": 0.9944,
+      "step": 17093
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7359888823290228e-06,
+      "loss": 0.9989,
+      "step": 17094
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.735461699591769e-06,
+      "loss": 1.0167,
+      "step": 17095
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7349345801042022e-06,
+      "loss": 0.901,
+      "step": 17096
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.734407523876539e-06,
+      "loss": 0.9194,
+      "step": 17097
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7338805309189877e-06,
+      "loss": 0.7285,
+      "step": 17098
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7333536012417618e-06,
+      "loss": 0.9996,
+      "step": 17099
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7328267348550676e-06,
+      "loss": 0.8023,
+      "step": 17100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7322999317691152e-06,
+      "loss": 1.0241,
+      "step": 17101
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7317731919941093e-06,
+      "loss": 0.9744,
+      "step": 17102
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.731246515540258e-06,
+      "loss": 1.0599,
+      "step": 17103
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7307199024177607e-06,
+      "loss": 0.9261,
+      "step": 17104
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7301933526368269e-06,
+      "loss": 0.9,
+      "step": 17105
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7296668662076537e-06,
+      "loss": 1.0234,
+      "step": 17106
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7291404431404446e-06,
+      "loss": 0.9689,
+      "step": 17107
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7286140834453958e-06,
+      "loss": 0.8193,
+      "step": 17108
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.728087787132709e-06,
+      "loss": 0.9203,
+      "step": 17109
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7275615542125774e-06,
+      "loss": 1.0076,
+      "step": 17110
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7270353846951997e-06,
+      "loss": 1.0243,
+      "step": 17111
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7265092785907673e-06,
+      "loss": 0.9134,
+      "step": 17112
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.725983235909477e-06,
+      "loss": 1.1107,
+      "step": 17113
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7254572566615168e-06,
+      "loss": 0.8492,
+      "step": 17114
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7249313408570817e-06,
+      "loss": 0.7697,
+      "step": 17115
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.724405488506357e-06,
+      "loss": 0.8996,
+      "step": 17116
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7238796996195329e-06,
+      "loss": 1.0384,
+      "step": 17117
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7233539742067984e-06,
+      "loss": 0.9122,
+      "step": 17118
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7228283122783357e-06,
+      "loss": 0.7424,
+      "step": 17119
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7223027138443322e-06,
+      "loss": 0.8995,
+      "step": 17120
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.721777178914969e-06,
+      "loss": 0.8374,
+      "step": 17121
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7212517075004309e-06,
+      "loss": 0.9864,
+      "step": 17122
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7207262996108953e-06,
+      "loss": 0.9344,
+      "step": 17123
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7202009552565458e-06,
+      "loss": 0.8526,
+      "step": 17124
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7196756744475568e-06,
+      "loss": 1.0147,
+      "step": 17125
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7191504571941092e-06,
+      "loss": 1.1722,
+      "step": 17126
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7186253035063738e-06,
+      "loss": 0.9299,
+      "step": 17127
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7181002133945324e-06,
+      "loss": 0.9998,
+      "step": 17128
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7175751868687535e-06,
+      "loss": 1.0036,
+      "step": 17129
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7170502239392124e-06,
+      "loss": 0.8216,
+      "step": 17130
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7165253246160762e-06,
+      "loss": 0.9522,
+      "step": 17131
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7160004889095195e-06,
+      "loss": 0.8496,
+      "step": 17132
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715475716829706e-06,
+      "loss": 1.0007,
+      "step": 17133
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7149510083868076e-06,
+      "loss": 0.9755,
+      "step": 17134
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7144263635909864e-06,
+      "loss": 0.9663,
+      "step": 17135
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7139017824524106e-06,
+      "loss": 0.8496,
+      "step": 17136
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7133772649812407e-06,
+      "loss": 0.8311,
+      "step": 17137
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.712852811187643e-06,
+      "loss": 0.9395,
+      "step": 17138
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7123284210817725e-06,
+      "loss": 0.9034,
+      "step": 17139
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7118040946737968e-06,
+      "loss": 0.7177,
+      "step": 17140
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7112798319738687e-06,
+      "loss": 1.0466,
+      "step": 17141
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7107556329921498e-06,
+      "loss": 0.822,
+      "step": 17142
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.710231497738793e-06,
+      "loss": 0.9732,
+      "step": 17143
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.709707426223956e-06,
+      "loss": 1.0446,
+      "step": 17144
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7091834184577894e-06,
+      "loss": 1.0483,
+      "step": 17145
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7086594744504498e-06,
+      "loss": 0.8902,
+      "step": 17146
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7081355942120843e-06,
+      "loss": 0.9373,
+      "step": 17147
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7076117777528467e-06,
+      "loss": 0.8738,
+      "step": 17148
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7070880250828826e-06,
+      "loss": 0.9924,
+      "step": 17149
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7065643362123407e-06,
+      "loss": 1.0173,
+      "step": 17150
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7060407111513695e-06,
+      "loss": 1.019,
+      "step": 17151
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7055171499101103e-06,
+      "loss": 0.9375,
+      "step": 17152
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.704993652498711e-06,
+      "loss": 0.7538,
+      "step": 17153
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.704470218927311e-06,
+      "loss": 0.8183,
+      "step": 17154
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.703946849206054e-06,
+      "loss": 0.7694,
+      "step": 17155
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.703423543345078e-06,
+      "loss": 0.9809,
+      "step": 17156
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7029003013545241e-06,
+      "loss": 0.9647,
+      "step": 17157
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7023771232445273e-06,
+      "loss": 0.9517,
+      "step": 17158
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7018540090252283e-06,
+      "loss": 0.9695,
+      "step": 17159
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7013309587067573e-06,
+      "loss": 0.8204,
+      "step": 17160
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7008079722992527e-06,
+      "loss": 0.8011,
+      "step": 17161
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7002850498128415e-06,
+      "loss": 0.9823,
+      "step": 17162
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6997621912576623e-06,
+      "loss": 0.7242,
+      "step": 17163
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.699239396643841e-06,
+      "loss": 0.8524,
+      "step": 17164
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6987166659815085e-06,
+      "loss": 1.0492,
+      "step": 17165
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6981939992807906e-06,
+      "loss": 1.2169,
+      "step": 17166
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6976713965518166e-06,
+      "loss": 0.821,
+      "step": 17167
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6971488578047085e-06,
+      "loss": 0.9531,
+      "step": 17168
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6966263830495939e-06,
+      "loss": 0.9185,
+      "step": 17169
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6961039722965917e-06,
+      "loss": 0.8068,
+      "step": 17170
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6955816255558277e-06,
+      "loss": 0.8123,
+      "step": 17171
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6950593428374185e-06,
+      "loss": 0.9451,
+      "step": 17172
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6945371241514857e-06,
+      "loss": 0.8497,
+      "step": 17173
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6940149695081437e-06,
+      "loss": 0.8432,
+      "step": 17174
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6934928789175154e-06,
+      "loss": 0.882,
+      "step": 17175
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6929708523897104e-06,
+      "loss": 0.9429,
+      "step": 17176
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6924488899348462e-06,
+      "loss": 0.7819,
+      "step": 17177
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6919269915630332e-06,
+      "loss": 0.9316,
+      "step": 17178
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6914051572843864e-06,
+      "loss": 0.8828,
+      "step": 17179
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6908833871090118e-06,
+      "loss": 0.8891,
+      "step": 17180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.690361681047023e-06,
+      "loss": 0.7147,
+      "step": 17181
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6898400391085246e-06,
+      "loss": 1.0912,
+      "step": 17182
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6893184613036255e-06,
+      "loss": 0.958,
+      "step": 17183
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6887969476424293e-06,
+      "loss": 0.957,
+      "step": 17184
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6882754981350402e-06,
+      "loss": 0.9346,
+      "step": 17185
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6877541127915647e-06,
+      "loss": 1.0165,
+      "step": 17186
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6872327916221e-06,
+      "loss": 1.0053,
+      "step": 17187
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6867115346367496e-06,
+      "loss": 0.9503,
+      "step": 17188
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6861903418456104e-06,
+      "loss": 1.0629,
+      "step": 17189
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6856692132587832e-06,
+      "loss": 0.9821,
+      "step": 17190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6851481488863619e-06,
+      "loss": 0.9365,
+      "step": 17191
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6846271487384447e-06,
+      "loss": 1.3251,
+      "step": 17192
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6841062128251223e-06,
+      "loss": 0.7903,
+      "step": 17193
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6835853411564922e-06,
+      "loss": 0.7605,
+      "step": 17194
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6830645337426417e-06,
+      "loss": 1.0372,
+      "step": 17195
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6825437905936653e-06,
+      "loss": 0.9222,
+      "step": 17196
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6820231117196473e-06,
+      "loss": 1.1012,
+      "step": 17197
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6815024971306821e-06,
+      "loss": 1.2284,
+      "step": 17198
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6809819468368516e-06,
+      "loss": 0.8667,
+      "step": 17199
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6804614608482444e-06,
+      "loss": 0.9187,
+      "step": 17200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6799410391749416e-06,
+      "loss": 0.9871,
+      "step": 17201
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6794206818270304e-06,
+      "loss": 1.0516,
+      "step": 17202
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6789003888145876e-06,
+      "loss": 0.9382,
+      "step": 17203
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.678380160147699e-06,
+      "loss": 0.8453,
+      "step": 17204
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6778599958364394e-06,
+      "loss": 0.9319,
+      "step": 17205
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6773398958908904e-06,
+      "loss": 0.8681,
+      "step": 17206
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6768198603211255e-06,
+      "loss": 0.9455,
+      "step": 17207
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.676299889137224e-06,
+      "loss": 0.899,
+      "step": 17208
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.675779982349256e-06,
+      "loss": 1.0215,
+      "step": 17209
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6752601399672968e-06,
+      "loss": 0.95,
+      "step": 17210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6747403620014202e-06,
+      "loss": 0.754,
+      "step": 17211
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.674220648461693e-06,
+      "loss": 1.0273,
+      "step": 17212
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6737009993581883e-06,
+      "loss": 0.939,
+      "step": 17213
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.67318141470097e-06,
+      "loss": 0.9684,
+      "step": 17214
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6726618945001094e-06,
+      "loss": 1.0692,
+      "step": 17215
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6721424387656676e-06,
+      "loss": 0.9888,
+      "step": 17216
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6716230475077133e-06,
+      "loss": 0.9148,
+      "step": 17217
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6711037207363051e-06,
+      "loss": 1.0621,
+      "step": 17218
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6705844584615094e-06,
+      "loss": 0.9451,
+      "step": 17219
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6700652606933815e-06,
+      "loss": 0.8043,
+      "step": 17220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6695461274419867e-06,
+      "loss": 0.8188,
+      "step": 17221
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6690270587173784e-06,
+      "loss": 0.8157,
+      "step": 17222
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6685080545296173e-06,
+      "loss": 0.7972,
+      "step": 17223
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.667989114888755e-06,
+      "loss": 0.9262,
+      "step": 17224
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6674702398048497e-06,
+      "loss": 0.9124,
+      "step": 17225
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6669514292879507e-06,
+      "loss": 0.9178,
+      "step": 17226
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.666432683348113e-06,
+      "loss": 0.8945,
+      "step": 17227
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6659140019953851e-06,
+      "loss": 0.9764,
+      "step": 17228
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.665395385239818e-06,
+      "loss": 1.1015,
+      "step": 17229
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6648768330914578e-06,
+      "loss": 0.8482,
+      "step": 17230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.664358345560354e-06,
+      "loss": 0.9469,
+      "step": 17231
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6638399226565471e-06,
+      "loss": 0.7258,
+      "step": 17232
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.663321564390088e-06,
+      "loss": 1.0171,
+      "step": 17233
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6628032707710151e-06,
+      "loss": 0.8873,
+      "step": 17234
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6622850418093739e-06,
+      "loss": 0.8208,
+      "step": 17235
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6617668775152014e-06,
+      "loss": 0.9446,
+      "step": 17236
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6612487778985392e-06,
+      "loss": 0.8086,
+      "step": 17237
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6607307429694237e-06,
+      "loss": 0.9279,
+      "step": 17238
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6602127727378936e-06,
+      "loss": 0.9149,
+      "step": 17239
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6596948672139824e-06,
+      "loss": 0.9116,
+      "step": 17240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6591770264077267e-06,
+      "loss": 1.0856,
+      "step": 17241
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.658659250329157e-06,
+      "loss": 0.9573,
+      "step": 17242
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.658141538988306e-06,
+      "loss": 1.0226,
+      "step": 17243
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6576238923952066e-06,
+      "loss": 0.8569,
+      "step": 17244
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.657106310559884e-06,
+      "loss": 1.0089,
+      "step": 17245
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6565887934923701e-06,
+      "loss": 0.835,
+      "step": 17246
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6560713412026884e-06,
+      "loss": 0.8899,
+      "step": 17247
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6555539537008681e-06,
+      "loss": 0.9406,
+      "step": 17248
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6550366309969291e-06,
+      "loss": 1.0456,
+      "step": 17249
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6545193731008984e-06,
+      "loss": 0.9356,
+      "step": 17250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.654002180022794e-06,
+      "loss": 0.9337,
+      "step": 17251
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6534850517726404e-06,
+      "loss": 0.9626,
+      "step": 17252
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6529679883604528e-06,
+      "loss": 0.9453,
+      "step": 17253
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6524509897962532e-06,
+      "loss": 1.0764,
+      "step": 17254
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6519340560900526e-06,
+      "loss": 0.9713,
+      "step": 17255
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.651417187251874e-06,
+      "loss": 0.8645,
+      "step": 17256
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.650900383291726e-06,
+      "loss": 0.955,
+      "step": 17257
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6503836442196247e-06,
+      "loss": 0.9615,
+      "step": 17258
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6498669700455794e-06,
+      "loss": 0.8725,
+      "step": 17259
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.649350360779603e-06,
+      "loss": 0.7011,
+      "step": 17260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6488338164317014e-06,
+      "loss": 0.8981,
+      "step": 17261
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.648317337011886e-06,
+      "loss": 1.0091,
+      "step": 17262
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6478009225301605e-06,
+      "loss": 0.7982,
+      "step": 17263
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6472845729965337e-06,
+      "loss": 0.9351,
+      "step": 17264
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6467682884210057e-06,
+      "loss": 1.0045,
+      "step": 17265
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6462520688135836e-06,
+      "loss": 0.8935,
+      "step": 17266
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6457359141842634e-06,
+      "loss": 0.9308,
+      "step": 17267
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6452198245430517e-06,
+      "loss": 0.9188,
+      "step": 17268
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6447037998999437e-06,
+      "loss": 1.1224,
+      "step": 17269
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6441878402649403e-06,
+      "loss": 1.1491,
+      "step": 17270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6436719456480344e-06,
+      "loss": 0.9652,
+      "step": 17271
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6431561160592246e-06,
+      "loss": 0.8447,
+      "step": 17272
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6426403515085026e-06,
+      "loss": 0.9181,
+      "step": 17273
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6421246520058638e-06,
+      "loss": 0.8478,
+      "step": 17274
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6416090175612958e-06,
+      "loss": 0.8705,
+      "step": 17275
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6410934481847934e-06,
+      "loss": 0.9525,
+      "step": 17276
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6405779438863418e-06,
+      "loss": 0.831,
+      "step": 17277
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6400625046759305e-06,
+      "loss": 0.9187,
+      "step": 17278
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.639547130563548e-06,
+      "loss": 0.8832,
+      "step": 17279
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6390318215591756e-06,
+      "loss": 0.9079,
+      "step": 17280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6385165776728008e-06,
+      "loss": 0.8568,
+      "step": 17281
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6380013989144034e-06,
+      "loss": 0.765,
+      "step": 17282
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6374862852939678e-06,
+      "loss": 0.9439,
+      "step": 17283
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6369712368214708e-06,
+      "loss": 0.8217,
+      "step": 17284
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6364562535068952e-06,
+      "loss": 0.8591,
+      "step": 17285
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6359413353602143e-06,
+      "loss": 0.9055,
+      "step": 17286
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6354264823914094e-06,
+      "loss": 1.0258,
+      "step": 17287
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6349116946104514e-06,
+      "loss": 1.0356,
+      "step": 17288
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6343969720273179e-06,
+      "loss": 0.9376,
+      "step": 17289
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6338823146519756e-06,
+      "loss": 0.8999,
+      "step": 17290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.633367722494404e-06,
+      "loss": 0.978,
+      "step": 17291
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6328531955645672e-06,
+      "loss": 1.0855,
+      "step": 17292
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6323387338724372e-06,
+      "loss": 0.9485,
+      "step": 17293
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6318243374279785e-06,
+      "loss": 0.9328,
+      "step": 17294
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6313100062411613e-06,
+      "loss": 0.9935,
+      "step": 17295
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6307957403219472e-06,
+      "loss": 0.9258,
+      "step": 17296
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.630281539680303e-06,
+      "loss": 0.8928,
+      "step": 17297
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6297674043261875e-06,
+      "loss": 0.8844,
+      "step": 17298
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.629253334269566e-06,
+      "loss": 0.895,
+      "step": 17299
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6287393295203946e-06,
+      "loss": 1.0205,
+      "step": 17300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6282253900886358e-06,
+      "loss": 1.0186,
+      "step": 17301
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6277115159842432e-06,
+      "loss": 0.8883,
+      "step": 17302
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6271977072171758e-06,
+      "loss": 1.0315,
+      "step": 17303
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6266839637973885e-06,
+      "loss": 0.7794,
+      "step": 17304
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6261702857348334e-06,
+      "loss": 1.0868,
+      "step": 17305
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6256566730394646e-06,
+      "loss": 0.9185,
+      "step": 17306
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6251431257212314e-06,
+      "loss": 0.9272,
+      "step": 17307
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6246296437900854e-06,
+      "loss": 0.9793,
+      "step": 17308
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6241162272559735e-06,
+      "loss": 0.9198,
+      "step": 17309
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6236028761288452e-06,
+      "loss": 0.8259,
+      "step": 17310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6230895904186438e-06,
+      "loss": 0.9046,
+      "step": 17311
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.622576370135317e-06,
+      "loss": 0.8887,
+      "step": 17312
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.622063215288804e-06,
+      "loss": 1.0095,
+      "step": 17313
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.621550125889053e-06,
+      "loss": 0.9639,
+      "step": 17314
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6210371019460003e-06,
+      "loss": 0.9326,
+      "step": 17315
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.620524143469589e-06,
+      "loss": 1.0522,
+      "step": 17316
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.620011250469754e-06,
+      "loss": 0.9171,
+      "step": 17317
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6194984229564358e-06,
+      "loss": 0.9494,
+      "step": 17318
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6189856609395676e-06,
+      "loss": 0.9322,
+      "step": 17319
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.618472964429087e-06,
+      "loss": 0.8875,
+      "step": 17320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6179603334349237e-06,
+      "loss": 0.9617,
+      "step": 17321
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6174477679670132e-06,
+      "loss": 0.9964,
+      "step": 17322
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6169352680352835e-06,
+      "loss": 0.9601,
+      "step": 17323
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6164228336496673e-06,
+      "loss": 0.7934,
+      "step": 17324
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6159104648200879e-06,
+      "loss": 0.9298,
+      "step": 17325
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.615398161556479e-06,
+      "loss": 0.8984,
+      "step": 17326
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6148859238687609e-06,
+      "loss": 1.0719,
+      "step": 17327
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6143737517668623e-06,
+      "loss": 1.0136,
+      "step": 17328
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6138616452607026e-06,
+      "loss": 0.9996,
+      "step": 17329
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6133496043602065e-06,
+      "loss": 0.8835,
+      "step": 17330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6128376290752922e-06,
+      "loss": 0.8042,
+      "step": 17331
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6123257194158826e-06,
+      "loss": 1.097,
+      "step": 17332
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6118138753918917e-06,
+      "loss": 1.0165,
+      "step": 17333
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6113020970132398e-06,
+      "loss": 0.8787,
+      "step": 17334
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6107903842898398e-06,
+      "loss": 1.0136,
+      "step": 17335
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6102787372316075e-06,
+      "loss": 0.7991,
+      "step": 17336
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6097671558484568e-06,
+      "loss": 0.768,
+      "step": 17337
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6092556401502973e-06,
+      "loss": 0.8569,
+      "step": 17338
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.608744190147042e-06,
+      "loss": 0.8897,
+      "step": 17339
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6082328058485968e-06,
+      "loss": 1.1928,
+      "step": 17340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.607721487264874e-06,
+      "loss": 0.9653,
+      "step": 17341
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6072102344057755e-06,
+      "loss": 0.9246,
+      "step": 17342
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6066990472812116e-06,
+      "loss": 0.7717,
+      "step": 17343
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6061879259010816e-06,
+      "loss": 1.0568,
+      "step": 17344
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6056768702752924e-06,
+      "loss": 0.9013,
+      "step": 17345
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6051658804137426e-06,
+      "loss": 1.0117,
+      "step": 17346
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6046549563263346e-06,
+      "loss": 0.7325,
+      "step": 17347
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6041440980229644e-06,
+      "loss": 0.9087,
+      "step": 17348
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6036333055135345e-06,
+      "loss": 0.7299,
+      "step": 17349
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.603122578807938e-06,
+      "loss": 0.7619,
+      "step": 17350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6026119179160721e-06,
+      "loss": 0.787,
+      "step": 17351
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6021013228478277e-06,
+      "loss": 0.7595,
+      "step": 17352
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6015907936131015e-06,
+      "loss": 0.8262,
+      "step": 17353
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.601080330221781e-06,
+      "loss": 0.8433,
+      "step": 17354
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6005699326837599e-06,
+      "loss": 0.8771,
+      "step": 17355
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6000596010089237e-06,
+      "loss": 0.9807,
+      "step": 17356
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5995493352071629e-06,
+      "loss": 0.8907,
+      "step": 17357
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5990391352883606e-06,
+      "loss": 0.9815,
+      "step": 17358
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5985290012624056e-06,
+      "loss": 0.9418,
+      "step": 17359
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5980189331391776e-06,
+      "loss": 0.9292,
+      "step": 17360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5975089309285613e-06,
+      "loss": 0.9204,
+      "step": 17361
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5969989946404391e-06,
+      "loss": 0.9783,
+      "step": 17362
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.596489124284688e-06,
+      "loss": 0.7802,
+      "step": 17363
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5959793198711898e-06,
+      "loss": 0.8323,
+      "step": 17364
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5954695814098181e-06,
+      "loss": 0.8959,
+      "step": 17365
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5949599089104512e-06,
+      "loss": 0.9975,
+      "step": 17366
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5944503023829655e-06,
+      "loss": 0.9656,
+      "step": 17367
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5939407618372305e-06,
+      "loss": 0.9561,
+      "step": 17368
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5934312872831226e-06,
+      "loss": 0.9093,
+      "step": 17369
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5929218787305096e-06,
+      "loss": 0.9184,
+      "step": 17370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5924125361892618e-06,
+      "loss": 0.9647,
+      "step": 17371
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5919032596692496e-06,
+      "loss": 0.9637,
+      "step": 17372
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5913940491803376e-06,
+      "loss": 0.8519,
+      "step": 17373
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5908849047323938e-06,
+      "loss": 0.9387,
+      "step": 17374
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5903758263352802e-06,
+      "loss": 1.0002,
+      "step": 17375
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5898668139988632e-06,
+      "loss": 0.9183,
+      "step": 17376
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.589357867733002e-06,
+      "loss": 1.0439,
+      "step": 17377
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5888489875475594e-06,
+      "loss": 0.9396,
+      "step": 17378
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5883401734523923e-06,
+      "loss": 0.9166,
+      "step": 17379
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5878314254573622e-06,
+      "loss": 1.0757,
+      "step": 17380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5873227435723226e-06,
+      "loss": 0.8459,
+      "step": 17381
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5868141278071325e-06,
+      "loss": 0.7736,
+      "step": 17382
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5863055781716407e-06,
+      "loss": 0.8062,
+      "step": 17383
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5857970946757078e-06,
+      "loss": 0.9538,
+      "step": 17384
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5852886773291799e-06,
+      "loss": 0.944,
+      "step": 17385
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5847803261419109e-06,
+      "loss": 1.0314,
+      "step": 17386
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5842720411237466e-06,
+      "loss": 0.8898,
+      "step": 17387
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5837638222845385e-06,
+      "loss": 0.9962,
+      "step": 17388
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5832556696341295e-06,
+      "loss": 1.0533,
+      "step": 17389
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5827475831823692e-06,
+      "loss": 1.0004,
+      "step": 17390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5822395629390968e-06,
+      "loss": 0.9679,
+      "step": 17391
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5817316089141598e-06,
+      "loss": 0.8439,
+      "step": 17392
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5812237211173953e-06,
+      "loss": 0.8195,
+      "step": 17393
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.580715899558648e-06,
+      "loss": 1.0216,
+      "step": 17394
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5802081442477523e-06,
+      "loss": 1.0407,
+      "step": 17395
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5797004551945488e-06,
+      "loss": 0.7413,
+      "step": 17396
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.579192832408874e-06,
+      "loss": 0.9247,
+      "step": 17397
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.578685275900561e-06,
+      "loss": 1.0428,
+      "step": 17398
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5781777856794462e-06,
+      "loss": 1.0398,
+      "step": 17399
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5776703617553585e-06,
+      "loss": 0.8449,
+      "step": 17400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5771630041381337e-06,
+      "loss": 1.1003,
+      "step": 17401
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.576655712837597e-06,
+      "loss": 0.9477,
+      "step": 17402
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5761484878635809e-06,
+      "loss": 1.0149,
+      "step": 17403
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.57564132922591e-06,
+      "loss": 1.048,
+      "step": 17404
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.575134236934413e-06,
+      "loss": 1.0406,
+      "step": 17405
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5746272109989097e-06,
+      "loss": 0.7671,
+      "step": 17406
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.574120251429231e-06,
+      "loss": 0.9745,
+      "step": 17407
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5736133582351931e-06,
+      "loss": 0.978,
+      "step": 17408
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5731065314266214e-06,
+      "loss": 0.9746,
+      "step": 17409
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5725997710133311e-06,
+      "loss": 0.867,
+      "step": 17410
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5720930770051446e-06,
+      "loss": 0.8045,
+      "step": 17411
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.571586449411876e-06,
+      "loss": 0.711,
+      "step": 17412
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5710798882433432e-06,
+      "loss": 0.7925,
+      "step": 17413
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5705733935093586e-06,
+      "loss": 1.1293,
+      "step": 17414
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.570066965219738e-06,
+      "loss": 1.0538,
+      "step": 17415
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5695606033842903e-06,
+      "loss": 1.2048,
+      "step": 17416
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.569054308012829e-06,
+      "loss": 0.9729,
+      "step": 17417
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5685480791151598e-06,
+      "loss": 0.8039,
+      "step": 17418
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5680419167010958e-06,
+      "loss": 1.0701,
+      "step": 17419
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5675358207804398e-06,
+      "loss": 0.8445,
+      "step": 17420
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.567029791363e-06,
+      "loss": 0.8768,
+      "step": 17421
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5665238284585777e-06,
+      "loss": 0.8889,
+      "step": 17422
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5660179320769792e-06,
+      "loss": 0.9458,
+      "step": 17423
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5655121022280028e-06,
+      "loss": 0.8698,
+      "step": 17424
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5650063389214515e-06,
+      "loss": 0.9016,
+      "step": 17425
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5645006421671222e-06,
+      "loss": 0.8872,
+      "step": 17426
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5639950119748148e-06,
+      "loss": 1.0655,
+      "step": 17427
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5634894483543234e-06,
+      "loss": 0.9487,
+      "step": 17428
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5629839513154438e-06,
+      "loss": 0.8371,
+      "step": 17429
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5624785208679727e-06,
+      "loss": 0.9171,
+      "step": 17430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5619731570216984e-06,
+      "loss": 0.9505,
+      "step": 17431
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5614678597864163e-06,
+      "loss": 0.9405,
+      "step": 17432
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5609626291719126e-06,
+      "loss": 1.0003,
+      "step": 17433
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5604574651879794e-06,
+      "loss": 0.892,
+      "step": 17434
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5599523678444012e-06,
+      "loss": 0.9131,
+      "step": 17435
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5594473371509672e-06,
+      "loss": 0.939,
+      "step": 17436
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5589423731174592e-06,
+      "loss": 0.8841,
+      "step": 17437
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.558437475753663e-06,
+      "loss": 0.9584,
+      "step": 17438
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5579326450693594e-06,
+      "loss": 0.8423,
+      "step": 17439
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5574278810743309e-06,
+      "loss": 1.0505,
+      "step": 17440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5569231837783532e-06,
+      "loss": 0.8377,
+      "step": 17441
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5564185531912112e-06,
+      "loss": 1.0525,
+      "step": 17442
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5559139893226766e-06,
+      "loss": 0.8015,
+      "step": 17443
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5554094921825291e-06,
+      "loss": 1.0097,
+      "step": 17444
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.55490506178054e-06,
+      "loss": 0.9897,
+      "step": 17445
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5544006981264848e-06,
+      "loss": 0.9859,
+      "step": 17446
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5538964012301334e-06,
+      "loss": 0.9872,
+      "step": 17447
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5533921711012589e-06,
+      "loss": 0.8049,
+      "step": 17448
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5528880077496278e-06,
+      "loss": 0.9809,
+      "step": 17449
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5523839111850108e-06,
+      "loss": 0.9726,
+      "step": 17450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5518798814171725e-06,
+      "loss": 1.0852,
+      "step": 17451
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5513759184558807e-06,
+      "loss": 0.89,
+      "step": 17452
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5508720223108964e-06,
+      "loss": 0.7876,
+      "step": 17453
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5503681929919845e-06,
+      "loss": 0.8724,
+      "step": 17454
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5498644305089077e-06,
+      "loss": 0.9608,
+      "step": 17455
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5493607348714235e-06,
+      "loss": 0.8846,
+      "step": 17456
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5488571060892944e-06,
+      "loss": 0.9546,
+      "step": 17457
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5483535441722735e-06,
+      "loss": 1.0271,
+      "step": 17458
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5478500491301218e-06,
+      "loss": 1.0568,
+      "step": 17459
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5473466209725907e-06,
+      "loss": 0.8078,
+      "step": 17460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5468432597094374e-06,
+      "loss": 0.822,
+      "step": 17461
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5463399653504107e-06,
+      "loss": 1.0882,
+      "step": 17462
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5458367379052642e-06,
+      "loss": 0.9843,
+      "step": 17463
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.545333577383747e-06,
+      "loss": 0.9037,
+      "step": 17464
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5448304837956095e-06,
+      "loss": 1.057,
+      "step": 17465
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5443274571505963e-06,
+      "loss": 0.8516,
+      "step": 17466
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.543824497458456e-06,
+      "loss": 0.9081,
+      "step": 17467
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5433216047289313e-06,
+      "loss": 1.0506,
+      "step": 17468
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5428187789717675e-06,
+      "loss": 0.9833,
+      "step": 17469
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5423160201967036e-06,
+      "loss": 0.9633,
+      "step": 17470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5418133284134845e-06,
+      "loss": 0.897,
+      "step": 17471
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5413107036318465e-06,
+      "loss": 1.0421,
+      "step": 17472
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5408081458615298e-06,
+      "loss": 0.8165,
+      "step": 17473
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5403056551122697e-06,
+      "loss": 0.6215,
+      "step": 17474
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5398032313938039e-06,
+      "loss": 0.9955,
+      "step": 17475
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5393008747158627e-06,
+      "loss": 0.9793,
+      "step": 17476
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5387985850881853e-06,
+      "loss": 0.7946,
+      "step": 17477
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5382963625204982e-06,
+      "loss": 0.9301,
+      "step": 17478
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5377942070225355e-06,
+      "loss": 0.7831,
+      "step": 17479
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5372921186040235e-06,
+      "loss": 0.8615,
+      "step": 17480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5367900972746924e-06,
+      "loss": 1.121,
+      "step": 17481
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.536288143044266e-06,
+      "loss": 1.1359,
+      "step": 17482
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5357862559224729e-06,
+      "loss": 0.9926,
+      "step": 17483
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5352844359190338e-06,
+      "loss": 0.9096,
+      "step": 17484
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5347826830436742e-06,
+      "loss": 1.0768,
+      "step": 17485
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5342809973061123e-06,
+      "loss": 0.8689,
+      "step": 17486
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.533779378716072e-06,
+      "loss": 1.1492,
+      "step": 17487
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5332778272832683e-06,
+      "loss": 0.9341,
+      "step": 17488
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5327763430174197e-06,
+      "loss": 0.9942,
+      "step": 17489
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.532274925928245e-06,
+      "loss": 1.0247,
+      "step": 17490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5317735760254554e-06,
+      "loss": 0.9644,
+      "step": 17491
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.531272293318768e-06,
+      "loss": 0.9753,
+      "step": 17492
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5307710778178908e-06,
+      "loss": 0.8719,
+      "step": 17493
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.530269929532539e-06,
+      "loss": 0.8796,
+      "step": 17494
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5297688484724189e-06,
+      "loss": 0.9126,
+      "step": 17495
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.529267834647241e-06,
+      "loss": 1.0509,
+      "step": 17496
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5287668880667107e-06,
+      "loss": 1.0045,
+      "step": 17497
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5282660087405355e-06,
+      "loss": 0.9702,
+      "step": 17498
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5277651966784158e-06,
+      "loss": 0.8348,
+      "step": 17499
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5272644518900615e-06,
+      "loss": 0.9043,
+      "step": 17500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5267637743851688e-06,
+      "loss": 0.9557,
+      "step": 17501
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5262631641734415e-06,
+      "loss": 0.9869,
+      "step": 17502
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.525762621264576e-06,
+      "loss": 0.7902,
+      "step": 17503
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.525262145668273e-06,
+      "loss": 0.9312,
+      "step": 17504
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5247617373942263e-06,
+      "loss": 0.88,
+      "step": 17505
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.524261396452134e-06,
+      "loss": 0.8956,
+      "step": 17506
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5237611228516875e-06,
+      "loss": 0.9277,
+      "step": 17507
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.523260916602582e-06,
+      "loss": 0.9253,
+      "step": 17508
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5227607777145066e-06,
+      "loss": 0.8954,
+      "step": 17509
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5222607061971535e-06,
+      "loss": 0.9741,
+      "step": 17510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5217607020602066e-06,
+      "loss": 1.0375,
+      "step": 17511
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.521260765313361e-06,
+      "loss": 0.8064,
+      "step": 17512
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5207608959662973e-06,
+      "loss": 1.0673,
+      "step": 17513
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5202610940287032e-06,
+      "loss": 1.1424,
+      "step": 17514
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.51976135951026e-06,
+      "loss": 0.8226,
+      "step": 17515
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5192616924206522e-06,
+      "loss": 0.8389,
+      "step": 17516
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5187620927695585e-06,
+      "loss": 1.0432,
+      "step": 17517
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5182625605666607e-06,
+      "loss": 0.899,
+      "step": 17518
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5177630958216344e-06,
+      "loss": 0.9074,
+      "step": 17519
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5172636985441597e-06,
+      "loss": 0.7962,
+      "step": 17520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5167643687439098e-06,
+      "loss": 1.0067,
+      "step": 17521
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5162651064305594e-06,
+      "loss": 0.9809,
+      "step": 17522
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5157659116137847e-06,
+      "loss": 0.9635,
+      "step": 17523
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5152667843032525e-06,
+      "loss": 1.0283,
+      "step": 17524
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5147677245086383e-06,
+      "loss": 1.0356,
+      "step": 17525
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5142687322396071e-06,
+      "loss": 0.9915,
+      "step": 17526
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.51376980750583e-06,
+      "loss": 0.8658,
+      "step": 17527
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5132709503169706e-06,
+      "loss": 1.0066,
+      "step": 17528
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5127721606826973e-06,
+      "loss": 0.9762,
+      "step": 17529
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5122734386126703e-06,
+      "loss": 0.8754,
+      "step": 17530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5117747841165565e-06,
+      "loss": 0.9014,
+      "step": 17531
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.511276197204013e-06,
+      "loss": 0.8688,
+      "step": 17532
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5107776778847033e-06,
+      "loss": 0.8139,
+      "step": 17533
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5102792261682813e-06,
+      "loss": 1.0563,
+      "step": 17534
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5097808420644117e-06,
+      "loss": 0.9901,
+      "step": 17535
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5092825255827442e-06,
+      "loss": 1.1342,
+      "step": 17536
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5087842767329374e-06,
+      "loss": 0.7235,
+      "step": 17537
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5082860955246426e-06,
+      "loss": 0.9241,
+      "step": 17538
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5077879819675133e-06,
+      "loss": 0.8948,
+      "step": 17539
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5072899360711985e-06,
+      "loss": 0.9587,
+      "step": 17540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5067919578453504e-06,
+      "loss": 0.8753,
+      "step": 17541
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5062940472996146e-06,
+      "loss": 0.8862,
+      "step": 17542
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5057962044436403e-06,
+      "loss": 0.9813,
+      "step": 17543
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5052984292870704e-06,
+      "loss": 1.2005,
+      "step": 17544
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5048007218395521e-06,
+      "loss": 0.955,
+      "step": 17545
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5043030821107257e-06,
+      "loss": 1.016,
+      "step": 17546
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.503805510110234e-06,
+      "loss": 0.8221,
+      "step": 17547
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.503308005847719e-06,
+      "loss": 0.9316,
+      "step": 17548
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5028105693328166e-06,
+      "loss": 0.9002,
+      "step": 17549
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5023132005751678e-06,
+      "loss": 0.9837,
+      "step": 17550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5018158995844057e-06,
+      "loss": 0.8547,
+      "step": 17551
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5013186663701685e-06,
+      "loss": 0.7885,
+      "step": 17552
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5008215009420874e-06,
+      "loss": 0.9456,
+      "step": 17553
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.500324403309797e-06,
+      "loss": 0.9124,
+      "step": 17554
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499827373482926e-06,
+      "loss": 0.8251,
+      "step": 17555
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499330411471107e-06,
+      "loss": 0.9886,
+      "step": 17556
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4988335172839647e-06,
+      "loss": 0.8817,
+      "step": 17557
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4983366909311315e-06,
+      "loss": 0.8336,
+      "step": 17558
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4978399324222293e-06,
+      "loss": 1.0183,
+      "step": 17559
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4973432417668853e-06,
+      "loss": 1.0597,
+      "step": 17560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.49684661897472e-06,
+      "loss": 1.02,
+      "step": 17561
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4963500640553586e-06,
+      "loss": 1.1439,
+      "step": 17562
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4958535770184179e-06,
+      "loss": 0.9717,
+      "step": 17563
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.495357157873521e-06,
+      "loss": 0.8466,
+      "step": 17564
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4948608066302827e-06,
+      "loss": 0.6439,
+      "step": 17565
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4943645232983229e-06,
+      "loss": 0.9928,
+      "step": 17566
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4938683078872534e-06,
+      "loss": 0.7947,
+      "step": 17567
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4933721604066914e-06,
+      "loss": 0.9951,
+      "step": 17568
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4928760808662458e-06,
+      "loss": 0.9985,
+      "step": 17569
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4923800692755336e-06,
+      "loss": 0.7433,
+      "step": 17570
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4918841256441603e-06,
+      "loss": 0.8129,
+      "step": 17571
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4913882499817377e-06,
+      "loss": 0.9898,
+      "step": 17572
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4908924422978705e-06,
+      "loss": 0.931,
+      "step": 17573
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4903967026021676e-06,
+      "loss": 0.88,
+      "step": 17574
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.489901030904231e-06,
+      "loss": 0.8989,
+      "step": 17575
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.489405427213667e-06,
+      "loss": 0.84,
+      "step": 17576
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4889098915400757e-06,
+      "loss": 0.9067,
+      "step": 17577
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4884144238930598e-06,
+      "loss": 0.822,
+      "step": 17578
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.487919024282216e-06,
+      "loss": 0.9696,
+      "step": 17579
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4874236927171465e-06,
+      "loss": 1.085,
+      "step": 17580
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.486928429207445e-06,
+      "loss": 0.7435,
+      "step": 17581
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4864332337627075e-06,
+      "loss": 1.0754,
+      "step": 17582
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.485938106392531e-06,
+      "loss": 0.9519,
+      "step": 17583
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.485443047106505e-06,
+      "loss": 1.0799,
+      "step": 17584
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4849480559142244e-06,
+      "loss": 1.0652,
+      "step": 17585
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.484453132825276e-06,
+      "loss": 0.8095,
+      "step": 17586
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4839582778492522e-06,
+      "loss": 0.8843,
+      "step": 17587
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.483463490995738e-06,
+      "loss": 0.8706,
+      "step": 17588
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4829687722743225e-06,
+      "loss": 0.9991,
+      "step": 17589
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4824741216945876e-06,
+      "loss": 1.0099,
+      "step": 17590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4819795392661201e-06,
+      "loss": 0.992,
+      "step": 17591
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.481485024998498e-06,
+      "loss": 0.9747,
+      "step": 17592
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4809905789013085e-06,
+      "loss": 0.8957,
+      "step": 17593
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4804962009841267e-06,
+      "loss": 0.9335,
+      "step": 17594
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4800018912565339e-06,
+      "loss": 0.7311,
+      "step": 17595
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4795076497281052e-06,
+      "loss": 1.0786,
+      "step": 17596
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.479013476408418e-06,
+      "loss": 0.8317,
+      "step": 17597
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.478519371307045e-06,
+      "loss": 0.9987,
+      "step": 17598
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4780253344335615e-06,
+      "loss": 0.8763,
+      "step": 17599
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.477531365797537e-06,
+      "loss": 0.8406,
+      "step": 17600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4770374654085446e-06,
+      "loss": 0.9407,
+      "step": 17601
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4765436332761507e-06,
+      "loss": 1.0278,
+      "step": 17602
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4760498694099261e-06,
+      "loss": 1.1042,
+      "step": 17603
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4755561738194329e-06,
+      "loss": 1.1224,
+      "step": 17604
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4750625465142425e-06,
+      "loss": 0.9717,
+      "step": 17605
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4745689875039144e-06,
+      "loss": 0.8445,
+      "step": 17606
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4740754967980137e-06,
+      "loss": 0.8888,
+      "step": 17607
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.473582074406099e-06,
+      "loss": 0.9701,
+      "step": 17608
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.473088720337733e-06,
+      "loss": 1.0083,
+      "step": 17609
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4725954346024717e-06,
+      "loss": 0.9984,
+      "step": 17610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4721022172098754e-06,
+      "loss": 0.7588,
+      "step": 17611
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.471609068169496e-06,
+      "loss": 0.8577,
+      "step": 17612
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4711159874908931e-06,
+      "loss": 0.9559,
+      "step": 17613
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4706229751836153e-06,
+      "loss": 1.1029,
+      "step": 17614
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.470130031257217e-06,
+      "loss": 0.8969,
+      "step": 17615
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4696371557212501e-06,
+      "loss": 0.9171,
+      "step": 17616
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4691443485852608e-06,
+      "loss": 0.8839,
+      "step": 17617
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4686516098588005e-06,
+      "loss": 1.0558,
+      "step": 17618
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.468158939551413e-06,
+      "loss": 0.8419,
+      "step": 17619
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4676663376726457e-06,
+      "loss": 0.9873,
+      "step": 17620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.467173804232041e-06,
+      "loss": 1.0199,
+      "step": 17621
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4666813392391437e-06,
+      "loss": 1.0744,
+      "step": 17622
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4661889427034926e-06,
+      "loss": 0.8524,
+      "step": 17623
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4656966146346301e-06,
+      "loss": 0.8933,
+      "step": 17624
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.465204355042093e-06,
+      "loss": 0.8742,
+      "step": 17625
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4647121639354212e-06,
+      "loss": 0.8966,
+      "step": 17626
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4642200413241464e-06,
+      "loss": 0.867,
+      "step": 17627
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4637279872178089e-06,
+      "loss": 0.9079,
+      "step": 17628
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4632360016259377e-06,
+      "loss": 0.8618,
+      "step": 17629
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4627440845580687e-06,
+      "loss": 0.8624,
+      "step": 17630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.462252236023729e-06,
+      "loss": 0.9949,
+      "step": 17631
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4617604560324516e-06,
+      "loss": 1.0193,
+      "step": 17632
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4612687445937612e-06,
+      "loss": 0.8014,
+      "step": 17633
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4607771017171874e-06,
+      "loss": 1.0411,
+      "step": 17634
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.460285527412253e-06,
+      "loss": 0.9245,
+      "step": 17635
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.459794021688485e-06,
+      "loss": 1.0936,
+      "step": 17636
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.459302584555403e-06,
+      "loss": 0.9499,
+      "step": 17637
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.458811216022532e-06,
+      "loss": 0.8878,
+      "step": 17638
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4583199160993888e-06,
+      "loss": 1.0074,
+      "step": 17639
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4578286847954943e-06,
+      "loss": 0.9074,
+      "step": 17640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4573375221203667e-06,
+      "loss": 0.93,
+      "step": 17641
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4568464280835192e-06,
+      "loss": 0.9622,
+      "step": 17642
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4563554026944704e-06,
+      "loss": 0.9326,
+      "step": 17643
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4558644459627302e-06,
+      "loss": 0.8736,
+      "step": 17644
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.455373557897814e-06,
+      "loss": 1.0492,
+      "step": 17645
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4548827385092295e-06,
+      "loss": 0.9092,
+      "step": 17646
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4543919878064893e-06,
+      "loss": 0.8945,
+      "step": 17647
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4539013057990986e-06,
+      "loss": 1.0151,
+      "step": 17648
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4534106924965669e-06,
+      "loss": 0.8876,
+      "step": 17649
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.452920147908396e-06,
+      "loss": 0.8952,
+      "step": 17650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.452429672044095e-06,
+      "loss": 0.9782,
+      "step": 17651
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4519392649131637e-06,
+      "loss": 0.9508,
+      "step": 17652
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4514489265251052e-06,
+      "loss": 0.9669,
+      "step": 17653
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4509586568894173e-06,
+      "loss": 0.9224,
+      "step": 17654
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4504684560156018e-06,
+      "loss": 0.9045,
+      "step": 17655
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4499783239131537e-06,
+      "loss": 0.9223,
+      "step": 17656
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4494882605915717e-06,
+      "loss": 0.8126,
+      "step": 17657
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4489982660603474e-06,
+      "loss": 1.0527,
+      "step": 17658
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4485083403289784e-06,
+      "loss": 0.9657,
+      "step": 17659
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4480184834069527e-06,
+      "loss": 1.0056,
+      "step": 17660
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4475286953037648e-06,
+      "loss": 0.9963,
+      "step": 17661
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4470389760288995e-06,
+      "loss": 0.9664,
+      "step": 17662
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.446549325591851e-06,
+      "loss": 0.9724,
+      "step": 17663
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4460597440021018e-06,
+      "loss": 1.0086,
+      "step": 17664
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4455702312691406e-06,
+      "loss": 0.7992,
+      "step": 17665
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.445080787402448e-06,
+      "loss": 0.8117,
+      "step": 17666
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4445914124115107e-06,
+      "loss": 0.8159,
+      "step": 17667
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4441021063058063e-06,
+      "loss": 0.8419,
+      "step": 17668
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4436128690948192e-06,
+      "loss": 0.9009,
+      "step": 17669
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.443123700788024e-06,
+      "loss": 0.9124,
+      "step": 17670
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4426346013949016e-06,
+      "loss": 0.7893,
+      "step": 17671
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.442145570924926e-06,
+      "loss": 0.7764,
+      "step": 17672
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4416566093875721e-06,
+      "loss": 0.9058,
+      "step": 17673
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4411677167923166e-06,
+      "loss": 0.9794,
+      "step": 17674
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.440678893148627e-06,
+      "loss": 0.7242,
+      "step": 17675
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4401901384659782e-06,
+      "loss": 0.9711,
+      "step": 17676
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4397014527538366e-06,
+      "loss": 0.7839,
+      "step": 17677
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4392128360216733e-06,
+      "loss": 1.0375,
+      "step": 17678
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.438724288278952e-06,
+      "loss": 0.7109,
+      "step": 17679
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4382358095351417e-06,
+      "loss": 0.955,
+      "step": 17680
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4377473997997028e-06,
+      "loss": 1.1873,
+      "step": 17681
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4372590590821012e-06,
+      "loss": 0.9964,
+      "step": 17682
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4367707873917957e-06,
+      "loss": 0.9784,
+      "step": 17683
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.43628258473825e-06,
+      "loss": 0.9045,
+      "step": 17684
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.435794451130917e-06,
+      "loss": 1.062,
+      "step": 17685
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4353063865792616e-06,
+      "loss": 0.8579,
+      "step": 17686
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.434818391092735e-06,
+      "loss": 0.8099,
+      "step": 17687
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4343304646807942e-06,
+      "loss": 0.8886,
+      "step": 17688
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4338426073528905e-06,
+      "loss": 0.7433,
+      "step": 17689
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4333548191184794e-06,
+      "loss": 0.8116,
+      "step": 17690
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4328670999870075e-06,
+      "loss": 0.8935,
+      "step": 17691
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4323794499679284e-06,
+      "loss": 0.828,
+      "step": 17692
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4318918690706862e-06,
+      "loss": 0.8912,
+      "step": 17693
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4314043573047315e-06,
+      "loss": 0.9235,
+      "step": 17694
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.430916914679506e-06,
+      "loss": 0.9478,
+      "step": 17695
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4304295412044577e-06,
+      "loss": 0.9119,
+      "step": 17696
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4299422368890237e-06,
+      "loss": 1.0056,
+      "step": 17697
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4294550017426522e-06,
+      "loss": 0.945,
+      "step": 17698
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4289678357747788e-06,
+      "loss": 1.066,
+      "step": 17699
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4284807389948452e-06,
+      "loss": 0.8763,
+      "step": 17700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4279937114122855e-06,
+      "loss": 0.9141,
+      "step": 17701
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4275067530365388e-06,
+      "loss": 0.8773,
+      "step": 17702
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4270198638770372e-06,
+      "loss": 1.0435,
+      "step": 17703
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4265330439432162e-06,
+      "loss": 0.8345,
+      "step": 17704
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4260462932445057e-06,
+      "loss": 0.9321,
+      "step": 17705
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4255596117903392e-06,
+      "loss": 0.9556,
+      "step": 17706
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4250729995901424e-06,
+      "loss": 0.9256,
+      "step": 17707
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4245864566533458e-06,
+      "loss": 1.0101,
+      "step": 17708
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4240999829893764e-06,
+      "loss": 0.9063,
+      "step": 17709
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4236135786076578e-06,
+      "loss": 0.9906,
+      "step": 17710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4231272435176163e-06,
+      "loss": 0.9638,
+      "step": 17711
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4226409777286716e-06,
+      "loss": 0.817,
+      "step": 17712
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.422154781250248e-06,
+      "loss": 0.769,
+      "step": 17713
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4216686540917617e-06,
+      "loss": 0.8611,
+      "step": 17714
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4211825962626357e-06,
+      "loss": 0.7462,
+      "step": 17715
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4206966077722839e-06,
+      "loss": 0.8935,
+      "step": 17716
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4202106886301243e-06,
+      "loss": 0.9876,
+      "step": 17717
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4197248388455693e-06,
+      "loss": 1.0186,
+      "step": 17718
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4192390584280347e-06,
+      "loss": 0.9811,
+      "step": 17719
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4187533473869286e-06,
+      "loss": 1.0731,
+      "step": 17720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4182677057316663e-06,
+      "loss": 1.1025,
+      "step": 17721
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.417782133471654e-06,
+      "loss": 0.9642,
+      "step": 17722
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.417296630616301e-06,
+      "loss": 0.881,
+      "step": 17723
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.416811197175012e-06,
+      "loss": 1.0342,
+      "step": 17724
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.416325833157195e-06,
+      "loss": 0.9596,
+      "step": 17725
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4158405385722501e-06,
+      "loss": 0.829,
+      "step": 17726
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4153553134295838e-06,
+      "loss": 0.9507,
+      "step": 17727
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.414870157738593e-06,
+      "loss": 0.8937,
+      "step": 17728
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4143850715086815e-06,
+      "loss": 0.95,
+      "step": 17729
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4139000547492444e-06,
+      "loss": 1.0538,
+      "step": 17730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4134151074696816e-06,
+      "loss": 0.9364,
+      "step": 17731
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4129302296793862e-06,
+      "loss": 1.0487,
+      "step": 17732
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4124454213877543e-06,
+      "loss": 0.8841,
+      "step": 17733
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4119606826041794e-06,
+      "loss": 0.9327,
+      "step": 17734
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4114760133380511e-06,
+      "loss": 1.0147,
+      "step": 17735
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.410991413598763e-06,
+      "loss": 0.8342,
+      "step": 17736
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4105068833956997e-06,
+      "loss": 0.9938,
+      "step": 17737
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4100224227382537e-06,
+      "loss": 1.037,
+      "step": 17738
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4095380316358064e-06,
+      "loss": 0.9906,
+      "step": 17739
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4090537100977476e-06,
+      "loss": 0.6884,
+      "step": 17740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.408569458133457e-06,
+      "loss": 0.897,
+      "step": 17741
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.40808527575232e-06,
+      "loss": 0.9109,
+      "step": 17742
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4076011629637126e-06,
+      "loss": 0.9428,
+      "step": 17743
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4071171197770212e-06,
+      "loss": 1.1017,
+      "step": 17744
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4066331462016186e-06,
+      "loss": 0.9502,
+      "step": 17745
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4061492422468858e-06,
+      "loss": 1.008,
+      "step": 17746
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4056654079221943e-06,
+      "loss": 1.0019,
+      "step": 17747
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.405181643236922e-06,
+      "loss": 0.8792,
+      "step": 17748
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4046979482004381e-06,
+      "loss": 0.8389,
+      "step": 17749
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4042143228221178e-06,
+      "loss": 0.8081,
+      "step": 17750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4037307671113276e-06,
+      "loss": 1.1053,
+      "step": 17751
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4032472810774394e-06,
+      "loss": 0.9527,
+      "step": 17752
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4027638647298175e-06,
+      "loss": 0.8413,
+      "step": 17753
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4022805180778321e-06,
+      "loss": 0.9231,
+      "step": 17754
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4017972411308417e-06,
+      "loss": 1.0586,
+      "step": 17755
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4013140338982168e-06,
+      "loss": 0.9586,
+      "step": 17756
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4008308963893146e-06,
+      "loss": 1.1562,
+      "step": 17757
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4003478286134992e-06,
+      "loss": 1.0624,
+      "step": 17758
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3998648305801266e-06,
+      "loss": 1.1248,
+      "step": 17759
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3993819022985572e-06,
+      "loss": 0.9581,
+      "step": 17760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3988990437781458e-06,
+      "loss": 0.9237,
+      "step": 17761
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3984162550282498e-06,
+      "loss": 0.9433,
+      "step": 17762
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3979335360582203e-06,
+      "loss": 0.9882,
+      "step": 17763
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3974508868774134e-06,
+      "loss": 0.8362,
+      "step": 17764
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3969683074951761e-06,
+      "loss": 0.8511,
+      "step": 17765
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3964857979208613e-06,
+      "loss": 0.924,
+      "step": 17766
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3960033581638176e-06,
+      "loss": 1.0491,
+      "step": 17767
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3955209882333903e-06,
+      "loss": 0.9733,
+      "step": 17768
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.395038688138927e-06,
+      "loss": 0.988,
+      "step": 17769
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3945564578897707e-06,
+      "loss": 1.0165,
+      "step": 17770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3940742974952665e-06,
+      "loss": 0.9928,
+      "step": 17771
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3935922069647534e-06,
+      "loss": 0.9049,
+      "step": 17772
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3931101863075741e-06,
+      "loss": 0.7852,
+      "step": 17773
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.392628235533066e-06,
+      "loss": 1.0782,
+      "step": 17774
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3921463546505687e-06,
+      "loss": 0.9072,
+      "step": 17775
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.391664543669416e-06,
+      "loss": 0.9614,
+      "step": 17776
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3911828025989455e-06,
+      "loss": 0.924,
+      "step": 17777
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3907011314484869e-06,
+      "loss": 0.9392,
+      "step": 17778
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.390219530227378e-06,
+      "loss": 0.9928,
+      "step": 17779
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3897379989449455e-06,
+      "loss": 0.9897,
+      "step": 17780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3892565376105216e-06,
+      "loss": 0.9394,
+      "step": 17781
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3887751462334321e-06,
+      "loss": 0.8874,
+      "step": 17782
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.388293824823006e-06,
+      "loss": 0.9469,
+      "step": 17783
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.387812573388566e-06,
+      "loss": 0.9955,
+      "step": 17784
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3873313919394405e-06,
+      "loss": 0.9011,
+      "step": 17785
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3868502804849476e-06,
+      "loss": 1.0047,
+      "step": 17786
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3863692390344125e-06,
+      "loss": 0.8818,
+      "step": 17787
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3858882675971523e-06,
+      "loss": 0.8651,
+      "step": 17788
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.385407366182489e-06,
+      "loss": 1.0032,
+      "step": 17789
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3849265347997359e-06,
+      "loss": 0.9047,
+      "step": 17790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3844457734582111e-06,
+      "loss": 1.0083,
+      "step": 17791
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3839650821672301e-06,
+      "loss": 0.9392,
+      "step": 17792
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3834844609361064e-06,
+      "loss": 0.9039,
+      "step": 17793
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3830039097741494e-06,
+      "loss": 0.8916,
+      "step": 17794
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3825234286906725e-06,
+      "loss": 1.124,
+      "step": 17795
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3820430176949822e-06,
+      "loss": 0.8928,
+      "step": 17796
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.381562676796389e-06,
+      "loss": 0.9614,
+      "step": 17797
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3810824060041972e-06,
+      "loss": 1.0531,
+      "step": 17798
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.380602205327714e-06,
+      "loss": 0.8681,
+      "step": 17799
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3801220747762406e-06,
+      "loss": 0.758,
+      "step": 17800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.379642014359081e-06,
+      "loss": 1.078,
+      "step": 17801
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3791620240855374e-06,
+      "loss": 0.8891,
+      "step": 17802
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3786821039649073e-06,
+      "loss": 0.9769,
+      "step": 17803
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.378202254006491e-06,
+      "loss": 1.0686,
+      "step": 17804
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3777224742195827e-06,
+      "loss": 0.9584,
+      "step": 17805
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3772427646134816e-06,
+      "loss": 1.0192,
+      "step": 17806
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3767631251974788e-06,
+      "loss": 1.0851,
+      "step": 17807
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.37628355598087e-06,
+      "loss": 0.9429,
+      "step": 17808
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.375804056972944e-06,
+      "loss": 0.9285,
+      "step": 17809
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3753246281829936e-06,
+      "loss": 0.9818,
+      "step": 17810
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3748452696203046e-06,
+      "loss": 0.9921,
+      "step": 17811
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3743659812941678e-06,
+      "loss": 0.928,
+      "step": 17812
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3738867632138646e-06,
+      "loss": 0.7682,
+      "step": 17813
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.373407615388686e-06,
+      "loss": 1.0162,
+      "step": 17814
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.372928537827911e-06,
+      "loss": 0.9515,
+      "step": 17815
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3724495305408236e-06,
+      "loss": 0.9443,
+      "step": 17816
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3719705935367028e-06,
+      "loss": 0.9508,
+      "step": 17817
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.37149172682483e-06,
+      "loss": 0.89,
+      "step": 17818
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3710129304144803e-06,
+      "loss": 0.8933,
+      "step": 17819
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3705342043149339e-06,
+      "loss": 0.8936,
+      "step": 17820
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3700555485354622e-06,
+      "loss": 0.8208,
+      "step": 17821
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3695769630853423e-06,
+      "loss": 0.943,
+      "step": 17822
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3690984479738434e-06,
+      "loss": 0.8532,
+      "step": 17823
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3686200032102403e-06,
+      "loss": 0.9486,
+      "step": 17824
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.368141628803799e-06,
+      "loss": 1.0859,
+      "step": 17825
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3676633247637906e-06,
+      "loss": 0.8755,
+      "step": 17826
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3671850910994821e-06,
+      "loss": 1.0879,
+      "step": 17827
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3667069278201373e-06,
+      "loss": 0.9564,
+      "step": 17828
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3662288349350229e-06,
+      "loss": 0.9407,
+      "step": 17829
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3657508124533992e-06,
+      "loss": 0.9063,
+      "step": 17830
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3652728603845306e-06,
+      "loss": 0.9819,
+      "step": 17831
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.364794978737674e-06,
+      "loss": 0.7907,
+      "step": 17832
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3643171675220924e-06,
+      "loss": 0.7702,
+      "step": 17833
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3638394267470384e-06,
+      "loss": 1.0392,
+      "step": 17834
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3633617564217728e-06,
+      "loss": 1.1259,
+      "step": 17835
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.362884156555545e-06,
+      "loss": 0.9311,
+      "step": 17836
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3624066271576153e-06,
+      "loss": 0.9864,
+      "step": 17837
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.36192916823723e-06,
+      "loss": 0.8865,
+      "step": 17838
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3614517798036432e-06,
+      "loss": 0.9416,
+      "step": 17839
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3609744618661013e-06,
+      "loss": 1.1036,
+      "step": 17840
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3604972144338558e-06,
+      "loss": 1.0477,
+      "step": 17841
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3600200375161487e-06,
+      "loss": 0.9655,
+      "step": 17842
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3595429311222303e-06,
+      "loss": 1.0848,
+      "step": 17843
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3590658952613394e-06,
+      "loss": 0.8749,
+      "step": 17844
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3585889299427229e-06,
+      "loss": 0.7247,
+      "step": 17845
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3581120351756182e-06,
+      "loss": 0.774,
+      "step": 17846
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.357635210969268e-06,
+      "loss": 0.7985,
+      "step": 17847
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3571584573329067e-06,
+      "loss": 1.0094,
+      "step": 17848
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.356681774275777e-06,
+      "loss": 0.9859,
+      "step": 17849
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3562051618071098e-06,
+      "loss": 1.0135,
+      "step": 17850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3557286199361425e-06,
+      "loss": 1.0315,
+      "step": 17851
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3552521486721054e-06,
+      "loss": 1.021,
+      "step": 17852
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3547757480242324e-06,
+      "loss": 0.9354,
+      "step": 17853
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3542994180017505e-06,
+      "loss": 0.8364,
+      "step": 17854
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3538231586138922e-06,
+      "loss": 1.0257,
+      "step": 17855
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.353346969869882e-06,
+      "loss": 0.9565,
+      "step": 17856
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.352870851778948e-06,
+      "loss": 0.942,
+      "step": 17857
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3523948043503127e-06,
+      "loss": 0.8798,
+      "step": 17858
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3519188275932004e-06,
+      "loss": 0.8375,
+      "step": 17859
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3514429215168352e-06,
+      "loss": 1.0616,
+      "step": 17860
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.350967086130434e-06,
+      "loss": 0.8712,
+      "step": 17861
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3504913214432197e-06,
+      "loss": 1.2325,
+      "step": 17862
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.350015627464406e-06,
+      "loss": 0.9006,
+      "step": 17863
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3495400042032136e-06,
+      "loss": 0.841,
+      "step": 17864
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3490644516688539e-06,
+      "loss": 0.923,
+      "step": 17865
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3485889698705434e-06,
+      "loss": 0.9231,
+      "step": 17866
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3481135588174926e-06,
+      "loss": 0.8024,
+      "step": 17867
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3476382185189142e-06,
+      "loss": 0.9694,
+      "step": 17868
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3471629489840154e-06,
+      "loss": 1.0384,
+      "step": 17869
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3466877502220071e-06,
+      "loss": 0.8134,
+      "step": 17870
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3462126222420924e-06,
+      "loss": 1.0693,
+      "step": 17871
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3457375650534822e-06,
+      "loss": 1.0027,
+      "step": 17872
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3452625786653762e-06,
+      "loss": 0.8988,
+      "step": 17873
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.34478766308698e-06,
+      "loss": 0.9249,
+      "step": 17874
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3443128183274923e-06,
+      "loss": 0.9995,
+      "step": 17875
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3438380443961163e-06,
+      "loss": 0.8273,
+      "step": 17876
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.343363341302047e-06,
+      "loss": 0.9562,
+      "step": 17877
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.342888709054485e-06,
+      "loss": 0.9868,
+      "step": 17878
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3424141476626235e-06,
+      "loss": 0.8863,
+      "step": 17879
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.341939657135659e-06,
+      "loss": 0.8918,
+      "step": 17880
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.341465237482783e-06,
+      "loss": 1.038,
+      "step": 17881
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3409908887131896e-06,
+      "loss": 0.8485,
+      "step": 17882
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3405166108360662e-06,
+      "loss": 1.0038,
+      "step": 17883
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3400424038606031e-06,
+      "loss": 0.8116,
+      "step": 17884
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3395682677959897e-06,
+      "loss": 1.0362,
+      "step": 17885
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3390942026514098e-06,
+      "loss": 1.0063,
+      "step": 17886
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3386202084360505e-06,
+      "loss": 1.0165,
+      "step": 17887
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3381462851590927e-06,
+      "loss": 0.8354,
+      "step": 17888
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3376724328297197e-06,
+      "loss": 0.9202,
+      "step": 17889
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.337198651457114e-06,
+      "loss": 0.8823,
+      "step": 17890
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3367249410504525e-06,
+      "loss": 1.0158,
+      "step": 17891
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3362513016189154e-06,
+      "loss": 0.7626,
+      "step": 17892
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.335777733171677e-06,
+      "loss": 0.8647,
+      "step": 17893
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3353042357179135e-06,
+      "loss": 0.9095,
+      "step": 17894
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3348308092668012e-06,
+      "loss": 0.9903,
+      "step": 17895
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3343574538275089e-06,
+      "loss": 0.9643,
+      "step": 17896
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3338841694092108e-06,
+      "loss": 1.0062,
+      "step": 17897
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3334109560210735e-06,
+      "loss": 0.8618,
+      "step": 17898
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3329378136722693e-06,
+      "loss": 0.8101,
+      "step": 17899
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3324647423719611e-06,
+      "loss": 0.9343,
+      "step": 17900
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3319917421293182e-06,
+      "loss": 0.9991,
+      "step": 17901
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3315188129535024e-06,
+      "loss": 0.792,
+      "step": 17902
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.331045954853678e-06,
+      "loss": 0.756,
+      "step": 17903
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.330573167839005e-06,
+      "loss": 0.8431,
+      "step": 17904
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3301004519186462e-06,
+      "loss": 0.9901,
+      "step": 17905
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3296278071017548e-06,
+      "loss": 0.9749,
+      "step": 17906
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.329155233397496e-06,
+      "loss": 1.0191,
+      "step": 17907
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3286827308150197e-06,
+      "loss": 1.0911,
+      "step": 17908
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.328210299363485e-06,
+      "loss": 0.8283,
+      "step": 17909
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3277379390520405e-06,
+      "loss": 0.9927,
+      "step": 17910
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3272656498898422e-06,
+      "loss": 0.9474,
+      "step": 17911
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3267934318860377e-06,
+      "loss": 1.0189,
+      "step": 17912
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3263212850497782e-06,
+      "loss": 0.9086,
+      "step": 17913
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3258492093902092e-06,
+      "loss": 0.8689,
+      "step": 17914
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3253772049164799e-06,
+      "loss": 0.9696,
+      "step": 17915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.324905271637732e-06,
+      "loss": 0.9848,
+      "step": 17916
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3244334095631123e-06,
+      "loss": 0.9726,
+      "step": 17917
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.32396161870176e-06,
+      "loss": 1.1277,
+      "step": 17918
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.323489899062818e-06,
+      "loss": 0.8527,
+      "step": 17919
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3230182506554263e-06,
+      "loss": 0.9071,
+      "step": 17920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3225466734887205e-06,
+      "loss": 0.7834,
+      "step": 17921
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3220751675718402e-06,
+      "loss": 0.8371,
+      "step": 17922
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3216037329139176e-06,
+      "loss": 0.9096,
+      "step": 17923
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3211323695240901e-06,
+      "loss": 0.9582,
+      "step": 17924
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.320661077411487e-06,
+      "loss": 1.0715,
+      "step": 17925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3201898565852417e-06,
+      "loss": 0.8735,
+      "step": 17926
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3197187070544827e-06,
+      "loss": 0.8606,
+      "step": 17927
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3192476288283395e-06,
+      "loss": 0.97,
+      "step": 17928
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3187766219159359e-06,
+      "loss": 1.1759,
+      "step": 17929
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3183056863264031e-06,
+      "loss": 1.0882,
+      "step": 17930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3178348220688615e-06,
+      "loss": 0.8483,
+      "step": 17931
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3173640291524358e-06,
+      "loss": 0.9147,
+      "step": 17932
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3168933075862455e-06,
+      "loss": 0.8409,
+      "step": 17933
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3164226573794137e-06,
+      "loss": 0.9459,
+      "step": 17934
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3159520785410552e-06,
+      "loss": 0.8363,
+      "step": 17935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3154815710802914e-06,
+      "loss": 0.9741,
+      "step": 17936
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.315011135006235e-06,
+      "loss": 0.9036,
+      "step": 17937
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3145407703280033e-06,
+      "loss": 0.8321,
+      "step": 17938
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3140704770547065e-06,
+      "loss": 0.8663,
+      "step": 17939
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3136002551954596e-06,
+      "loss": 0.9134,
+      "step": 17940
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3131301047593685e-06,
+      "loss": 0.894,
+      "step": 17941
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3126600257555488e-06,
+      "loss": 1.0752,
+      "step": 17942
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.312190018193103e-06,
+      "loss": 1.0441,
+      "step": 17943
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.31172008208114e-06,
+      "loss": 0.8591,
+      "step": 17944
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3112502174287628e-06,
+      "loss": 0.9979,
+      "step": 17945
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3107804242450773e-06,
+      "loss": 1.0062,
+      "step": 17946
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3103107025391827e-06,
+      "loss": 0.9266,
+      "step": 17947
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3098410523201827e-06,
+      "loss": 0.8485,
+      "step": 17948
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3093714735971736e-06,
+      "loss": 1.0738,
+      "step": 17949
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.308901966379257e-06,
+      "loss": 0.9381,
+      "step": 17950
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3084325306755258e-06,
+      "loss": 1.0251,
+      "step": 17951
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3079631664950766e-06,
+      "loss": 0.757,
+      "step": 17952
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3074938738470055e-06,
+      "loss": 0.9821,
+      "step": 17953
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3070246527404012e-06,
+      "loss": 0.9102,
+      "step": 17954
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.306555503184358e-06,
+      "loss": 0.9607,
+      "step": 17955
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3060864251879628e-06,
+      "loss": 1.0187,
+      "step": 17956
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.305617418760307e-06,
+      "loss": 1.0919,
+      "step": 17957
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3051484839104738e-06,
+      "loss": 0.9707,
+      "step": 17958
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.304679620647552e-06,
+      "loss": 0.9677,
+      "step": 17959
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3042108289806232e-06,
+      "loss": 0.9718,
+      "step": 17960
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3037421089187724e-06,
+      "loss": 0.8885,
+      "step": 17961
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3032734604710784e-06,
+      "loss": 0.8731,
+      "step": 17962
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.302804883646624e-06,
+      "loss": 0.9648,
+      "step": 17963
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3023363784544834e-06,
+      "loss": 0.8888,
+      "step": 17964
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3018679449037396e-06,
+      "loss": 0.9957,
+      "step": 17965
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3013995830034642e-06,
+      "loss": 0.8975,
+      "step": 17966
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3009312927627342e-06,
+      "loss": 0.9384,
+      "step": 17967
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3004630741906199e-06,
+      "loss": 1.0756,
+      "step": 17968
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2999949272961953e-06,
+      "loss": 0.8623,
+      "step": 17969
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.299526852088529e-06,
+      "loss": 0.8038,
+      "step": 17970
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.299058848576692e-06,
+      "loss": 1.1063,
+      "step": 17971
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.298590916769748e-06,
+      "loss": 0.9127,
+      "step": 17972
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2981230566767678e-06,
+      "loss": 0.9331,
+      "step": 17973
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2976552683068117e-06,
+      "loss": 1.0292,
+      "step": 17974
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2971875516689465e-06,
+      "loss": 1.0297,
+      "step": 17975
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2967199067722308e-06,
+      "loss": 1.0488,
+      "step": 17976
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2962523336257271e-06,
+      "loss": 0.8071,
+      "step": 17977
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2957848322384959e-06,
+      "loss": 1.0041,
+      "step": 17978
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2953174026195914e-06,
+      "loss": 0.9249,
+      "step": 17979
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2948500447780738e-06,
+      "loss": 0.8229,
+      "step": 17980
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2943827587229945e-06,
+      "loss": 0.9361,
+      "step": 17981
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2939155444634099e-06,
+      "loss": 0.8721,
+      "step": 17982
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2934484020083697e-06,
+      "loss": 0.8493,
+      "step": 17983
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2929813313669276e-06,
+      "loss": 1.02,
+      "step": 17984
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2925143325481292e-06,
+      "loss": 0.9926,
+      "step": 17985
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2920474055610265e-06,
+      "loss": 0.8364,
+      "step": 17986
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2915805504146616e-06,
+      "loss": 0.8103,
+      "step": 17987
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2911137671180852e-06,
+      "loss": 0.8131,
+      "step": 17988
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2906470556803363e-06,
+      "loss": 1.033,
+      "step": 17989
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2901804161104608e-06,
+      "loss": 0.9227,
+      "step": 17990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.289713848417497e-06,
+      "loss": 0.9252,
+      "step": 17991
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.289247352610487e-06,
+      "loss": 1.0112,
+      "step": 17992
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2887809286984664e-06,
+      "loss": 0.9746,
+      "step": 17993
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2883145766904754e-06,
+      "loss": 0.8898,
+      "step": 17994
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2878482965955457e-06,
+      "loss": 0.9553,
+      "step": 17995
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2873820884227145e-06,
+      "loss": 0.9468,
+      "step": 17996
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2869159521810115e-06,
+      "loss": 1.0261,
+      "step": 17997
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2864498878794719e-06,
+      "loss": 1.0332,
+      "step": 17998
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2859838955271198e-06,
+      "loss": 0.9626,
+      "step": 17999
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2855179751329904e-06,
+      "loss": 0.8975,
+      "step": 18000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2850521267061061e-06,
+      "loss": 0.8991,
+      "step": 18001
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2845863502554956e-06,
+      "loss": 0.935,
+      "step": 18002
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.28412064579018e-06,
+      "loss": 1.0398,
+      "step": 18003
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.283655013319186e-06,
+      "loss": 0.9784,
+      "step": 18004
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2831894528515314e-06,
+      "loss": 0.9177,
+      "step": 18005
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2827239643962396e-06,
+      "loss": 0.654,
+      "step": 18006
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2822585479623263e-06,
+      "loss": 0.9293,
+      "step": 18007
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2817932035588115e-06,
+      "loss": 0.8387,
+      "step": 18008
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.281327931194709e-06,
+      "loss": 0.9868,
+      "step": 18009
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2808627308790355e-06,
+      "loss": 0.9583,
+      "step": 18010
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2803976026208015e-06,
+      "loss": 0.9264,
+      "step": 18011
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2799325464290202e-06,
+      "loss": 0.8957,
+      "step": 18012
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2794675623127034e-06,
+      "loss": 0.8621,
+      "step": 18013
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2790026502808567e-06,
+      "loss": 0.8114,
+      "step": 18014
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2785378103424917e-06,
+      "loss": 0.9349,
+      "step": 18015
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2780730425066108e-06,
+      "loss": 1.0276,
+      "step": 18016
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.277608346782222e-06,
+      "loss": 0.96,
+      "step": 18017
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2771437231783251e-06,
+      "loss": 1.0515,
+      "step": 18018
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2766791717039262e-06,
+      "loss": 0.8963,
+      "step": 18019
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.276214692368022e-06,
+      "loss": 0.8115,
+      "step": 18020
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.275750285179615e-06,
+      "loss": 0.9111,
+      "step": 18021
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2752859501476984e-06,
+      "loss": 0.9352,
+      "step": 18022
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2748216872812747e-06,
+      "loss": 1.0001,
+      "step": 18023
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.274357496589334e-06,
+      "loss": 0.8024,
+      "step": 18024
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2738933780808738e-06,
+      "loss": 1.0785,
+      "step": 18025
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2734293317648821e-06,
+      "loss": 0.932,
+      "step": 18026
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.272965357650353e-06,
+      "loss": 0.8869,
+      "step": 18027
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.272501455746274e-06,
+      "loss": 1.0825,
+      "step": 18028
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2720376260616345e-06,
+      "loss": 0.9346,
+      "step": 18029
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2715738686054197e-06,
+      "loss": 0.9895,
+      "step": 18030
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2711101833866163e-06,
+      "loss": 0.8165,
+      "step": 18031
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2706465704142058e-06,
+      "loss": 0.9286,
+      "step": 18032
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2701830296971734e-06,
+      "loss": 0.9944,
+      "step": 18033
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2697195612444963e-06,
+      "loss": 1.0276,
+      "step": 18034
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.269256165065159e-06,
+      "loss": 0.8516,
+      "step": 18035
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.268792841168136e-06,
+      "loss": 0.8498,
+      "step": 18036
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.268329589562407e-06,
+      "loss": 0.9031,
+      "step": 18037
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2678664102569437e-06,
+      "loss": 0.7996,
+      "step": 18038
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.267403303260724e-06,
+      "loss": 0.8411,
+      "step": 18039
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.266940268582717e-06,
+      "loss": 0.7555,
+      "step": 18040
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.266477306231897e-06,
+      "loss": 0.7993,
+      "step": 18041
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2660144162172305e-06,
+      "loss": 0.9132,
+      "step": 18042
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2655515985476896e-06,
+      "loss": 0.9379,
+      "step": 18043
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2650888532322375e-06,
+      "loss": 0.8887,
+      "step": 18044
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2646261802798421e-06,
+      "loss": 0.9274,
+      "step": 18045
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2641635796994684e-06,
+      "loss": 1.0071,
+      "step": 18046
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2637010515000764e-06,
+      "loss": 0.8002,
+      "step": 18047
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2632385956906306e-06,
+      "loss": 0.9904,
+      "step": 18048
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2627762122800874e-06,
+      "loss": 0.9762,
+      "step": 18049
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2623139012774093e-06,
+      "loss": 0.9425,
+      "step": 18050
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.26185166269155e-06,
+      "loss": 0.8388,
+      "step": 18051
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2613894965314682e-06,
+      "loss": 0.9851,
+      "step": 18052
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2609274028061153e-06,
+      "loss": 0.731,
+      "step": 18053
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2604653815244478e-06,
+      "loss": 1.047,
+      "step": 18054
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.260003432695413e-06,
+      "loss": 1.0046,
+      "step": 18055
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.259541556327965e-06,
+      "loss": 0.822,
+      "step": 18056
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2590797524310483e-06,
+      "loss": 0.7839,
+      "step": 18057
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2586180210136156e-06,
+      "loss": 0.9182,
+      "step": 18058
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2581563620846082e-06,
+      "loss": 0.8897,
+      "step": 18059
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2576947756529746e-06,
+      "loss": 0.7801,
+      "step": 18060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2572332617276539e-06,
+      "loss": 0.9548,
+      "step": 18061
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.256771820317591e-06,
+      "loss": 1.0795,
+      "step": 18062
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2563104514317243e-06,
+      "loss": 1.021,
+      "step": 18063
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.255849155078994e-06,
+      "loss": 0.8932,
+      "step": 18064
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2553879312683358e-06,
+      "loss": 1.0076,
+      "step": 18065
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2549267800086889e-06,
+      "loss": 0.9215,
+      "step": 18066
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2544657013089845e-06,
+      "loss": 1.0029,
+      "step": 18067
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2540046951781587e-06,
+      "loss": 0.9026,
+      "step": 18068
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2535437616251412e-06,
+      "loss": 0.9325,
+      "step": 18069
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.253082900658863e-06,
+      "loss": 0.9505,
+      "step": 18070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2526221122882553e-06,
+      "loss": 0.963,
+      "step": 18071
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2521613965222424e-06,
+      "loss": 0.8338,
+      "step": 18072
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.251700753369754e-06,
+      "loss": 0.9853,
+      "step": 18073
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2512401828397119e-06,
+      "loss": 0.6844,
+      "step": 18074
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.250779684941042e-06,
+      "loss": 1.0644,
+      "step": 18075
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.250319259682664e-06,
+      "loss": 0.9045,
+      "step": 18076
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2498589070735013e-06,
+      "loss": 1.0088,
+      "step": 18077
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.24939862712247e-06,
+      "loss": 1.0097,
+      "step": 18078
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2489384198384912e-06,
+      "loss": 0.9163,
+      "step": 18079
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2484782852304767e-06,
+      "loss": 0.8293,
+      "step": 18080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2480182233073473e-06,
+      "loss": 1.0234,
+      "step": 18081
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2475582340780118e-06,
+      "loss": 0.8833,
+      "step": 18082
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2470983175513863e-06,
+      "loss": 1.0068,
+      "step": 18083
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.246638473736378e-06,
+      "loss": 0.8856,
+      "step": 18084
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2461787026418998e-06,
+      "loss": 0.906,
+      "step": 18085
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2457190042768558e-06,
+      "loss": 0.983,
+      "step": 18086
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2452593786501554e-06,
+      "loss": 0.9097,
+      "step": 18087
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2447998257707022e-06,
+      "loss": 0.821,
+      "step": 18088
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2443403456474017e-06,
+      "loss": 0.9317,
+      "step": 18089
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2438809382891537e-06,
+      "loss": 0.9516,
+      "step": 18090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2434216037048618e-06,
+      "loss": 0.8537,
+      "step": 18091
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2429623419034215e-06,
+      "loss": 1.0115,
+      "step": 18092
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2425031528937364e-06,
+      "loss": 0.9907,
+      "step": 18093
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2420440366846987e-06,
+      "loss": 1.0302,
+      "step": 18094
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.241584993285207e-06,
+      "loss": 0.8972,
+      "step": 18095
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2411260227041523e-06,
+      "loss": 0.8858,
+      "step": 18096
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2406671249504292e-06,
+      "loss": 0.7534,
+      "step": 18097
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2402083000329268e-06,
+      "loss": 1.0001,
+      "step": 18098
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.239749547960537e-06,
+      "loss": 0.8744,
+      "step": 18099
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2392908687421451e-06,
+      "loss": 1.0521,
+      "step": 18100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2388322623866411e-06,
+      "loss": 1.0613,
+      "step": 18101
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2383737289029073e-06,
+      "loss": 0.9855,
+      "step": 18102
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2379152682998308e-06,
+      "loss": 0.9021,
+      "step": 18103
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2374568805862913e-06,
+      "loss": 0.951,
+      "step": 18104
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2369985657711714e-06,
+      "loss": 0.7126,
+      "step": 18105
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2365403238633517e-06,
+      "loss": 1.0181,
+      "step": 18106
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2360821548717083e-06,
+      "loss": 0.8517,
+      "step": 18107
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2356240588051204e-06,
+      "loss": 0.8401,
+      "step": 18108
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2351660356724616e-06,
+      "loss": 1.0236,
+      "step": 18109
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2347080854826076e-06,
+      "loss": 0.7221,
+      "step": 18110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2342502082444296e-06,
+      "loss": 1.1001,
+      "step": 18111
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2337924039668009e-06,
+      "loss": 0.9144,
+      "step": 18112
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2333346726585881e-06,
+      "loss": 1.0273,
+      "step": 18113
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.232877014328664e-06,
+      "loss": 0.9411,
+      "step": 18114
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2324194289858892e-06,
+      "loss": 1.0403,
+      "step": 18115
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2319619166391371e-06,
+      "loss": 1.0626,
+      "step": 18116
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2315044772972662e-06,
+      "loss": 0.8267,
+      "step": 18117
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.231047110969143e-06,
+      "loss": 0.9479,
+      "step": 18118
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2305898176636255e-06,
+      "loss": 0.8842,
+      "step": 18119
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2301325973895768e-06,
+      "loss": 0.8802,
+      "step": 18120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2296754501558523e-06,
+      "loss": 0.9968,
+      "step": 18121
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2292183759713116e-06,
+      "loss": 0.9448,
+      "step": 18122
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2287613748448091e-06,
+      "loss": 1.1486,
+      "step": 18123
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2283044467852002e-06,
+      "loss": 1.0679,
+      "step": 18124
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.227847591801336e-06,
+      "loss": 0.905,
+      "step": 18125
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2273908099020703e-06,
+      "loss": 0.8649,
+      "step": 18126
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2269341010962488e-06,
+      "loss": 0.9737,
+      "step": 18127
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2264774653927263e-06,
+      "loss": 0.8049,
+      "step": 18128
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2260209028003456e-06,
+      "loss": 0.8734,
+      "step": 18129
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2255644133279554e-06,
+      "loss": 0.7999,
+      "step": 18130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.225107996984397e-06,
+      "loss": 0.9827,
+      "step": 18131
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2246516537785158e-06,
+      "loss": 1.0101,
+      "step": 18132
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2241953837191517e-06,
+      "loss": 0.9376,
+      "step": 18133
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2237391868151465e-06,
+      "loss": 0.9477,
+      "step": 18134
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2232830630753367e-06,
+      "loss": 0.8935,
+      "step": 18135
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2228270125085623e-06,
+      "loss": 1.0276,
+      "step": 18136
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2223710351236562e-06,
+      "loss": 0.9894,
+      "step": 18137
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2219151309294547e-06,
+      "loss": 1.0037,
+      "step": 18138
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2214592999347912e-06,
+      "loss": 1.0003,
+      "step": 18139
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.221003542148496e-06,
+      "loss": 1.0311,
+      "step": 18140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.220547857579401e-06,
+      "loss": 0.96,
+      "step": 18141
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2200922462363324e-06,
+      "loss": 0.8554,
+      "step": 18142
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2196367081281208e-06,
+      "loss": 0.9249,
+      "step": 18143
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2191812432635892e-06,
+      "loss": 0.8695,
+      "step": 18144
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2187258516515644e-06,
+      "loss": 1.0107,
+      "step": 18145
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2182705333008672e-06,
+      "loss": 0.9477,
+      "step": 18146
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.217815288220322e-06,
+      "loss": 0.9713,
+      "step": 18147
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2173601164187466e-06,
+      "loss": 1.0523,
+      "step": 18148
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2169050179049618e-06,
+      "loss": 0.9501,
+      "step": 18149
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2164499926877814e-06,
+      "loss": 0.9075,
+      "step": 18150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.215995040776027e-06,
+      "loss": 0.8116,
+      "step": 18151
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.215540162178509e-06,
+      "loss": 0.9939,
+      "step": 18152
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2150853569040432e-06,
+      "loss": 0.8718,
+      "step": 18153
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2146306249614386e-06,
+      "loss": 1.2287,
+      "step": 18154
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2141759663595087e-06,
+      "loss": 0.881,
+      "step": 18155
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2137213811070587e-06,
+      "loss": 0.9124,
+      "step": 18156
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2132668692129001e-06,
+      "loss": 0.9657,
+      "step": 18157
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2128124306858352e-06,
+      "loss": 0.9138,
+      "step": 18158
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2123580655346718e-06,
+      "loss": 1.0234,
+      "step": 18159
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2119037737682104e-06,
+      "loss": 0.9235,
+      "step": 18160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2114495553952555e-06,
+      "loss": 0.8993,
+      "step": 18161
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2109954104246047e-06,
+      "loss": 1.0814,
+      "step": 18162
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2105413388650577e-06,
+      "loss": 0.9753,
+      "step": 18163
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.210087340725415e-06,
+      "loss": 1.1271,
+      "step": 18164
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2096334160144685e-06,
+      "loss": 0.8784,
+      "step": 18165
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2091795647410166e-06,
+      "loss": 0.8917,
+      "step": 18166
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2087257869138485e-06,
+      "loss": 0.8837,
+      "step": 18167
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2082720825417605e-06,
+      "loss": 0.8659,
+      "step": 18168
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.207818451633539e-06,
+      "loss": 0.9888,
+      "step": 18169
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.207364894197976e-06,
+      "loss": 1.0392,
+      "step": 18170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2069114102438567e-06,
+      "loss": 0.9736,
+      "step": 18171
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2064579997799698e-06,
+      "loss": 1.2081,
+      "step": 18172
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2060046628150961e-06,
+      "loss": 0.8722,
+      "step": 18173
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2055513993580236e-06,
+      "loss": 1.0115,
+      "step": 18174
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2050982094175312e-06,
+      "loss": 0.8977,
+      "step": 18175
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2046450930024012e-06,
+      "loss": 0.9903,
+      "step": 18176
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2041920501214105e-06,
+      "loss": 1.075,
+      "step": 18177
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2037390807833394e-06,
+      "loss": 1.0563,
+      "step": 18178
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2032861849969602e-06,
+      "loss": 0.9213,
+      "step": 18179
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.202833362771052e-06,
+      "loss": 0.849,
+      "step": 18180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2023806141143845e-06,
+      "loss": 0.835,
+      "step": 18181
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2019279390357324e-06,
+      "loss": 1.0046,
+      "step": 18182
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2014753375438636e-06,
+      "loss": 0.8379,
+      "step": 18183
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2010228096475502e-06,
+      "loss": 0.9193,
+      "step": 18184
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2005703553555553e-06,
+      "loss": 0.9463,
+      "step": 18185
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2001179746766505e-06,
+      "loss": 0.7655,
+      "step": 18186
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1996656676195968e-06,
+      "loss": 1.0781,
+      "step": 18187
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1992134341931594e-06,
+      "loss": 0.9093,
+      "step": 18188
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.198761274406099e-06,
+      "loss": 0.9362,
+      "step": 18189
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1983091882671776e-06,
+      "loss": 1.0321,
+      "step": 18190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.197857175785152e-06,
+      "loss": 0.9825,
+      "step": 18191
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1974052369687828e-06,
+      "loss": 1.0697,
+      "step": 18192
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1969533718268228e-06,
+      "loss": 0.898,
+      "step": 18193
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1965015803680302e-06,
+      "loss": 0.7605,
+      "step": 18194
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1960498626011553e-06,
+      "loss": 0.9982,
+      "step": 18195
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1955982185349508e-06,
+      "loss": 0.8985,
+      "step": 18196
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1951466481781698e-06,
+      "loss": 1.0172,
+      "step": 18197
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1946951515395584e-06,
+      "loss": 1.028,
+      "step": 18198
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1942437286278657e-06,
+      "loss": 0.8638,
+      "step": 18199
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.193792379451837e-06,
+      "loss": 0.9652,
+      "step": 18200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1933411040202181e-06,
+      "loss": 0.8852,
+      "step": 18201
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1928899023417512e-06,
+      "loss": 1.1603,
+      "step": 18202
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1924387744251797e-06,
+      "loss": 1.0231,
+      "step": 18203
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1919877202792419e-06,
+      "loss": 0.9612,
+      "step": 18204
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1915367399126798e-06,
+      "loss": 1.0207,
+      "step": 18205
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1910858333342279e-06,
+      "loss": 0.942,
+      "step": 18206
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1906350005526257e-06,
+      "loss": 0.9636,
+      "step": 18207
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.190184241576603e-06,
+      "loss": 0.9369,
+      "step": 18208
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1897335564148998e-06,
+      "loss": 0.9085,
+      "step": 18209
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1892829450762428e-06,
+      "loss": 0.9335,
+      "step": 18210
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1888324075693659e-06,
+      "loss": 0.9806,
+      "step": 18211
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1883819439029947e-06,
+      "loss": 1.0052,
+      "step": 18212
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.187931554085861e-06,
+      "loss": 0.9779,
+      "step": 18213
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1874812381266865e-06,
+      "loss": 0.7638,
+      "step": 18214
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1870309960342002e-06,
+      "loss": 0.8629,
+      "step": 18215
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1865808278171215e-06,
+      "loss": 0.8869,
+      "step": 18216
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.186130733484176e-06,
+      "loss": 0.8941,
+      "step": 18217
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1856807130440806e-06,
+      "loss": 1.025,
+      "step": 18218
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1852307665055579e-06,
+      "loss": 0.8433,
+      "step": 18219
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1847808938773208e-06,
+      "loss": 1.1201,
+      "step": 18220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1843310951680914e-06,
+      "loss": 0.9035,
+      "step": 18221
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1838813703865803e-06,
+      "loss": 1.0091,
+      "step": 18222
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.183431719541503e-06,
+      "loss": 0.8857,
+      "step": 18223
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1829821426415694e-06,
+      "loss": 0.8267,
+      "step": 18224
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1825326396954918e-06,
+      "loss": 0.9862,
+      "step": 18225
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1820832107119773e-06,
+      "loss": 0.9177,
+      "step": 18226
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.181633855699736e-06,
+      "loss": 0.926,
+      "step": 18227
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.181184574667471e-06,
+      "loss": 0.915,
+      "step": 18228
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1807353676238902e-06,
+      "loss": 1.1822,
+      "step": 18229
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1802862345776938e-06,
+      "loss": 0.8691,
+      "step": 18230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.179837175537585e-06,
+      "loss": 1.0387,
+      "step": 18231
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1793881905122656e-06,
+      "loss": 0.9709,
+      "step": 18232
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1789392795104326e-06,
+      "loss": 1.1012,
+      "step": 18233
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.178490442540785e-06,
+      "loss": 1.0388,
+      "step": 18234
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1780416796120176e-06,
+      "loss": 0.9439,
+      "step": 18235
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1775929907328266e-06,
+      "loss": 0.9858,
+      "step": 18236
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1771443759119028e-06,
+      "loss": 1.0021,
+      "step": 18237
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.176695835157941e-06,
+      "loss": 1.019,
+      "step": 18238
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1762473684796287e-06,
+      "loss": 1.0471,
+      "step": 18239
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.175798975885658e-06,
+      "loss": 0.8041,
+      "step": 18240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1753506573847135e-06,
+      "loss": 0.743,
+      "step": 18241
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.174902412985483e-06,
+      "loss": 0.7899,
+      "step": 18242
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1744542426966481e-06,
+      "loss": 1.126,
+      "step": 18243
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1740061465268977e-06,
+      "loss": 1.0798,
+      "step": 18244
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1735581244849088e-06,
+      "loss": 1.0612,
+      "step": 18245
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1731101765793645e-06,
+      "loss": 0.9921,
+      "step": 18246
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.172662302818941e-06,
+      "loss": 0.9116,
+      "step": 18247
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1722145032123183e-06,
+      "loss": 0.7145,
+      "step": 18248
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1717667777681702e-06,
+      "loss": 0.7961,
+      "step": 18249
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.171319126495174e-06,
+      "loss": 0.9312,
+      "step": 18250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1708715494019996e-06,
+      "loss": 0.8909,
+      "step": 18251
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1704240464973215e-06,
+      "loss": 0.9806,
+      "step": 18252
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.169976617789808e-06,
+      "loss": 0.7278,
+      "step": 18253
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.16952926328813e-06,
+      "loss": 0.9709,
+      "step": 18254
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.169081983000952e-06,
+      "loss": 0.9984,
+      "step": 18255
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1686347769369415e-06,
+      "loss": 0.9292,
+      "step": 18256
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1681876451047647e-06,
+      "loss": 1.1057,
+      "step": 18257
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1677405875130815e-06,
+      "loss": 1.0222,
+      "step": 18258
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1672936041705568e-06,
+      "loss": 0.9599,
+      "step": 18259
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1668466950858475e-06,
+      "loss": 0.9081,
+      "step": 18260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1663998602676153e-06,
+      "loss": 0.7946,
+      "step": 18261
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.165953099724515e-06,
+      "loss": 0.8794,
+      "step": 18262
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.165506413465205e-06,
+      "loss": 0.7922,
+      "step": 18263
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1650598014983373e-06,
+      "loss": 0.9198,
+      "step": 18264
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1646132638325669e-06,
+      "loss": 0.9765,
+      "step": 18265
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1641668004765416e-06,
+      "loss": 1.0082,
+      "step": 18266
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1637204114389179e-06,
+      "loss": 0.804,
+      "step": 18267
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1632740967283395e-06,
+      "loss": 0.8188,
+      "step": 18268
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1628278563534562e-06,
+      "loss": 0.7844,
+      "step": 18269
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1623816903229114e-06,
+      "loss": 0.9896,
+      "step": 18270
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1619355986453523e-06,
+      "loss": 0.7501,
+      "step": 18271
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1614895813294192e-06,
+      "loss": 0.9988,
+      "step": 18272
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.161043638383756e-06,
+      "loss": 0.9233,
+      "step": 18273
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1605977698170001e-06,
+      "loss": 0.9697,
+      "step": 18274
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1601519756377926e-06,
+      "loss": 0.9389,
+      "step": 18275
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1597062558547683e-06,
+      "loss": 0.9471,
+      "step": 18276
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1592606104765658e-06,
+      "loss": 0.7812,
+      "step": 18277
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1588150395118146e-06,
+      "loss": 1.1471,
+      "step": 18278
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1583695429691544e-06,
+      "loss": 0.9598,
+      "step": 18279
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1579241208572107e-06,
+      "loss": 0.9997,
+      "step": 18280
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.157478773184617e-06,
+      "loss": 0.8423,
+      "step": 18281
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1570334999599992e-06,
+      "loss": 0.8526,
+      "step": 18282
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.156588301191987e-06,
+      "loss": 0.9562,
+      "step": 18283
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1561431768892029e-06,
+      "loss": 0.906,
+      "step": 18284
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1556981270602745e-06,
+      "loss": 0.9882,
+      "step": 18285
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1552531517138215e-06,
+      "loss": 0.8693,
+      "step": 18286
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.154808250858468e-06,
+      "loss": 0.7837,
+      "step": 18287
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1543634245028307e-06,
+      "loss": 0.9348,
+      "step": 18288
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.15391867265553e-06,
+      "loss": 1.0544,
+      "step": 18289
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1534739953251838e-06,
+      "loss": 0.8497,
+      "step": 18290
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.153029392520405e-06,
+      "loss": 0.9573,
+      "step": 18291
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1525848642498111e-06,
+      "loss": 1.0342,
+      "step": 18292
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1521404105220107e-06,
+      "loss": 0.9413,
+      "step": 18293
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1516960313456183e-06,
+      "loss": 0.9261,
+      "step": 18294
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.151251726729241e-06,
+      "loss": 0.8714,
+      "step": 18295
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1508074966814902e-06,
+      "loss": 0.7889,
+      "step": 18296
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1503633412109694e-06,
+      "loss": 0.9765,
+      "step": 18297
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.149919260326287e-06,
+      "loss": 0.9822,
+      "step": 18298
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1494752540360444e-06,
+      "loss": 0.8677,
+      "step": 18299
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1490313223488463e-06,
+      "loss": 0.941,
+      "step": 18300
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1485874652732904e-06,
+      "loss": 0.7679,
+      "step": 18301
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1481436828179815e-06,
+      "loss": 0.8245,
+      "step": 18302
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1476999749915136e-06,
+      "loss": 1.0399,
+      "step": 18303
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1472563418024857e-06,
+      "loss": 1.0447,
+      "step": 18304
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1468127832594912e-06,
+      "loss": 0.708,
+      "step": 18305
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1463692993711267e-06,
+      "loss": 0.8122,
+      "step": 18306
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1459258901459808e-06,
+      "loss": 1.0099,
+      "step": 18307
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1454825555926486e-06,
+      "loss": 1.0214,
+      "step": 18308
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1450392957197154e-06,
+      "loss": 0.9111,
+      "step": 18309
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1445961105357734e-06,
+      "loss": 0.8908,
+      "step": 18310
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1441530000494055e-06,
+      "loss": 0.9607,
+      "step": 18311
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1437099642692e-06,
+      "loss": 0.8591,
+      "step": 18312
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1432670032037379e-06,
+      "loss": 0.9401,
+      "step": 18313
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1428241168616027e-06,
+      "loss": 0.8348,
+      "step": 18314
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.142381305251377e-06,
+      "loss": 0.9486,
+      "step": 18315
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.141938568381637e-06,
+      "loss": 1.093,
+      "step": 18316
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1414959062609615e-06,
+      "loss": 0.9173,
+      "step": 18317
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1410533188979295e-06,
+      "loss": 0.7114,
+      "step": 18318
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1406108063011123e-06,
+      "loss": 1.0409,
+      "step": 18319
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.140168368479087e-06,
+      "loss": 0.9599,
+      "step": 18320
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1397260054404225e-06,
+      "loss": 1.0181,
+      "step": 18321
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.139283717193692e-06,
+      "loss": 1.1098,
+      "step": 18322
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.138841503747463e-06,
+      "loss": 0.9715,
+      "step": 18323
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1383993651103037e-06,
+      "loss": 0.8949,
+      "step": 18324
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1379573012907823e-06,
+      "loss": 0.8241,
+      "step": 18325
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1375153122974602e-06,
+      "loss": 0.8258,
+      "step": 18326
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1370733981389049e-06,
+      "loss": 0.8373,
+      "step": 18327
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1366315588236743e-06,
+      "loss": 0.7165,
+      "step": 18328
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1361897943603327e-06,
+      "loss": 0.9355,
+      "step": 18329
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.135748104757436e-06,
+      "loss": 0.9452,
+      "step": 18330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.135306490023544e-06,
+      "loss": 0.9849,
+      "step": 18331
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.134864950167211e-06,
+      "loss": 0.8801,
+      "step": 18332
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1344234851969943e-06,
+      "loss": 0.9358,
+      "step": 18333
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1339820951214442e-06,
+      "loss": 1.0525,
+      "step": 18334
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1335407799491161e-06,
+      "loss": 0.8953,
+      "step": 18335
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1330995396885547e-06,
+      "loss": 0.8606,
+      "step": 18336
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1326583743483156e-06,
+      "loss": 0.8199,
+      "step": 18337
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1322172839369422e-06,
+      "loss": 0.9655,
+      "step": 18338
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1317762684629823e-06,
+      "loss": 1.0196,
+      "step": 18339
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1313353279349787e-06,
+      "loss": 0.9951,
+      "step": 18340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.130894462361477e-06,
+      "loss": 0.9695,
+      "step": 18341
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1304536717510162e-06,
+      "loss": 0.9049,
+      "step": 18342
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1300129561121387e-06,
+      "loss": 0.8294,
+      "step": 18343
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1295723154533812e-06,
+      "loss": 1.2596,
+      "step": 18344
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1291317497832831e-06,
+      "loss": 0.9369,
+      "step": 18345
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1286912591103788e-06,
+      "loss": 0.9498,
+      "step": 18346
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1282508434432038e-06,
+      "loss": 1.0819,
+      "step": 18347
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1278105027902898e-06,
+      "loss": 0.7568,
+      "step": 18348
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.127370237160168e-06,
+      "loss": 1.0659,
+      "step": 18349
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1269300465613714e-06,
+      "loss": 0.7876,
+      "step": 18350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1264899310024252e-06,
+      "loss": 0.9817,
+      "step": 18351
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1260498904918587e-06,
+      "loss": 1.06,
+      "step": 18352
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1256099250381963e-06,
+      "loss": 0.8601,
+      "step": 18353
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1251700346499633e-06,
+      "loss": 0.8963,
+      "step": 18354
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1247302193356813e-06,
+      "loss": 1.0073,
+      "step": 18355
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1242904791038734e-06,
+      "loss": 0.9338,
+      "step": 18356
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1238508139630567e-06,
+      "loss": 1.0277,
+      "step": 18357
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1234112239217525e-06,
+      "loss": 0.8369,
+      "step": 18358
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.122971708988474e-06,
+      "loss": 0.9423,
+      "step": 18359
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1225322691717422e-06,
+      "loss": 0.8341,
+      "step": 18360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.122092904480066e-06,
+      "loss": 1.0213,
+      "step": 18361
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.121653614921962e-06,
+      "loss": 1.0881,
+      "step": 18362
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1212144005059384e-06,
+      "loss": 0.862,
+      "step": 18363
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1207752612405066e-06,
+      "loss": 1.1043,
+      "step": 18364
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.120336197134173e-06,
+      "loss": 0.8736,
+      "step": 18365
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1198972081954472e-06,
+      "loss": 0.7948,
+      "step": 18366
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1194582944328315e-06,
+      "loss": 0.8508,
+      "step": 18367
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1190194558548317e-06,
+      "loss": 1.1893,
+      "step": 18368
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1185806924699487e-06,
+      "loss": 1.1823,
+      "step": 18369
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1181420042866858e-06,
+      "loss": 1.0202,
+      "step": 18370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1177033913135382e-06,
+      "loss": 0.9714,
+      "step": 18371
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1172648535590097e-06,
+      "loss": 1.0823,
+      "step": 18372
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1168263910315924e-06,
+      "loss": 1.0061,
+      "step": 18373
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.116388003739784e-06,
+      "loss": 1.1284,
+      "step": 18374
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1159496916920758e-06,
+      "loss": 0.9496,
+      "step": 18375
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1155114548969632e-06,
+      "loss": 0.6487,
+      "step": 18376
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1150732933629332e-06,
+      "loss": 0.7395,
+      "step": 18377
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1146352070984778e-06,
+      "loss": 0.9512,
+      "step": 18378
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1141971961120835e-06,
+      "loss": 0.6139,
+      "step": 18379
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1137592604122377e-06,
+      "loss": 0.9196,
+      "step": 18380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.113321400007424e-06,
+      "loss": 1.0413,
+      "step": 18381
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1128836149061256e-06,
+      "loss": 1.1395,
+      "step": 18382
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1124459051168273e-06,
+      "loss": 0.9832,
+      "step": 18383
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1120082706480067e-06,
+      "loss": 0.8181,
+      "step": 18384
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1115707115081448e-06,
+      "loss": 0.9693,
+      "step": 18385
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.111133227705717e-06,
+      "loss": 0.9285,
+      "step": 18386
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1106958192492024e-06,
+      "loss": 0.9901,
+      "step": 18387
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1102584861470722e-06,
+      "loss": 0.9497,
+      "step": 18388
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1098212284078037e-06,
+      "loss": 0.9512,
+      "step": 18389
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1093840460398642e-06,
+      "loss": 0.9501,
+      "step": 18390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1089469390517283e-06,
+      "loss": 0.8824,
+      "step": 18391
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1085099074518613e-06,
+      "loss": 0.8543,
+      "step": 18392
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1080729512487332e-06,
+      "loss": 0.9257,
+      "step": 18393
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.107636070450806e-06,
+      "loss": 0.8286,
+      "step": 18394
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1071992650665503e-06,
+      "loss": 1.0994,
+      "step": 18395
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1067625351044242e-06,
+      "loss": 0.7831,
+      "step": 18396
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1063258805728922e-06,
+      "loss": 1.0292,
+      "step": 18397
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1058893014804112e-06,
+      "loss": 1.0505,
+      "step": 18398
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1054527978354434e-06,
+      "loss": 0.8391,
+      "step": 18399
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1050163696464428e-06,
+      "loss": 0.866,
+      "step": 18400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1045800169218678e-06,
+      "loss": 0.8731,
+      "step": 18401
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1041437396701705e-06,
+      "loss": 0.9688,
+      "step": 18402
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1037075378998053e-06,
+      "loss": 0.9868,
+      "step": 18403
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1032714116192218e-06,
+      "loss": 1.0522,
+      "step": 18404
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.102835360836872e-06,
+      "loss": 0.8104,
+      "step": 18405
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1023993855612015e-06,
+      "loss": 1.2309,
+      "step": 18406
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.101963485800659e-06,
+      "loss": 1.0067,
+      "step": 18407
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1015276615636905e-06,
+      "loss": 0.737,
+      "step": 18408
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1010919128587383e-06,
+      "loss": 0.9731,
+      "step": 18409
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1006562396942478e-06,
+      "loss": 1.0681,
+      "step": 18410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1002206420786559e-06,
+      "loss": 1.0267,
+      "step": 18411
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0997851200204064e-06,
+      "loss": 0.9028,
+      "step": 18412
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0993496735279335e-06,
+      "loss": 0.8437,
+      "step": 18413
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0989143026096765e-06,
+      "loss": 0.8419,
+      "step": 18414
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0984790072740708e-06,
+      "loss": 0.9181,
+      "step": 18415
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0980437875295485e-06,
+      "loss": 0.9382,
+      "step": 18416
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0976086433845424e-06,
+      "loss": 0.981,
+      "step": 18417
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.097173574847485e-06,
+      "loss": 1.0814,
+      "step": 18418
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0967385819268028e-06,
+      "loss": 0.9965,
+      "step": 18419
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0963036646309266e-06,
+      "loss": 0.9463,
+      "step": 18420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0958688229682801e-06,
+      "loss": 1.0426,
+      "step": 18421
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.095434056947291e-06,
+      "loss": 1.0092,
+      "step": 18422
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0949993665763798e-06,
+      "loss": 0.8927,
+      "step": 18423
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.094564751863972e-06,
+      "loss": 0.8433,
+      "step": 18424
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0941302128184839e-06,
+      "loss": 0.8598,
+      "step": 18425
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0936957494483391e-06,
+      "loss": 0.8627,
+      "step": 18426
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0932613617619513e-06,
+      "loss": 0.9833,
+      "step": 18427
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0928270497677396e-06,
+      "loss": 0.8032,
+      "step": 18428
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0923928134741152e-06,
+      "loss": 0.9521,
+      "step": 18429
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0919586528894966e-06,
+      "loss": 0.9769,
+      "step": 18430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0915245680222908e-06,
+      "loss": 0.9042,
+      "step": 18431
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0910905588809112e-06,
+      "loss": 0.8547,
+      "step": 18432
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0906566254737638e-06,
+      "loss": 0.8788,
+      "step": 18433
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0902227678092592e-06,
+      "loss": 1.0353,
+      "step": 18434
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0897889858958005e-06,
+      "loss": 0.8994,
+      "step": 18435
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.089355279741794e-06,
+      "loss": 0.8871,
+      "step": 18436
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.088921649355641e-06,
+      "loss": 0.7952,
+      "step": 18437
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.088488094745745e-06,
+      "loss": 0.7271,
+      "step": 18438
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0880546159205035e-06,
+      "loss": 0.8886,
+      "step": 18439
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0876212128883178e-06,
+      "loss": 0.8553,
+      "step": 18440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0871878856575823e-06,
+      "loss": 0.9319,
+      "step": 18441
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0867546342366935e-06,
+      "loss": 1.0533,
+      "step": 18442
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0863214586340482e-06,
+      "loss": 0.9122,
+      "step": 18443
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0858883588580348e-06,
+      "loss": 0.9938,
+      "step": 18444
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0854553349170483e-06,
+      "loss": 1.0405,
+      "step": 18445
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0850223868194755e-06,
+      "loss": 0.9075,
+      "step": 18446
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0845895145737067e-06,
+      "loss": 0.9208,
+      "step": 18447
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.084156718188127e-06,
+      "loss": 0.8737,
+      "step": 18448
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.083723997671124e-06,
+      "loss": 0.8747,
+      "step": 18449
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0832913530310784e-06,
+      "loss": 0.8554,
+      "step": 18450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.082858784276376e-06,
+      "loss": 0.8124,
+      "step": 18451
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0824262914153938e-06,
+      "loss": 1.1591,
+      "step": 18452
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0819938744565156e-06,
+      "loss": 0.8272,
+      "step": 18453
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0815615334081164e-06,
+      "loss": 0.7432,
+      "step": 18454
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0811292682785745e-06,
+      "loss": 0.8732,
+      "step": 18455
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0806970790762632e-06,
+      "loss": 0.8641,
+      "step": 18456
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0802649658095582e-06,
+      "loss": 0.7996,
+      "step": 18457
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0798329284868286e-06,
+      "loss": 0.8171,
+      "step": 18458
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0794009671164484e-06,
+      "loss": 0.8382,
+      "step": 18459
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0789690817067843e-06,
+      "loss": 1.0114,
+      "step": 18460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0785372722662058e-06,
+      "loss": 0.7553,
+      "step": 18461
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0781055388030765e-06,
+      "loss": 1.0479,
+      "step": 18462
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0776738813257643e-06,
+      "loss": 0.851,
+      "step": 18463
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.077242299842628e-06,
+      "loss": 0.9279,
+      "step": 18464
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0768107943620354e-06,
+      "loss": 0.9465,
+      "step": 18465
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.076379364892342e-06,
+      "loss": 0.9204,
+      "step": 18466
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0759480114419097e-06,
+      "loss": 1.0999,
+      "step": 18467
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0755167340190937e-06,
+      "loss": 0.8742,
+      "step": 18468
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0750855326322518e-06,
+      "loss": 1.0019,
+      "step": 18469
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0746544072897358e-06,
+      "loss": 0.8828,
+      "step": 18470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0742233579999018e-06,
+      "loss": 0.8799,
+      "step": 18471
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0737923847710985e-06,
+      "loss": 0.9349,
+      "step": 18472
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0733614876116782e-06,
+      "loss": 0.7515,
+      "step": 18473
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.072930666529987e-06,
+      "loss": 0.8611,
+      "step": 18474
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.072499921534374e-06,
+      "loss": 0.8987,
+      "step": 18475
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.072069252633185e-06,
+      "loss": 1.1125,
+      "step": 18476
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0716386598347618e-06,
+      "loss": 0.8455,
+      "step": 18477
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0712081431474496e-06,
+      "loss": 0.9236,
+      "step": 18478
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0707777025795874e-06,
+      "loss": 0.9986,
+      "step": 18479
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0703473381395175e-06,
+      "loss": 0.8013,
+      "step": 18480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.069917049835575e-06,
+      "loss": 0.8068,
+      "step": 18481
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0694868376760993e-06,
+      "loss": 0.8364,
+      "step": 18482
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.069056701669423e-06,
+      "loss": 1.0364,
+      "step": 18483
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0686266418238833e-06,
+      "loss": 0.7666,
+      "step": 18484
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0681966581478087e-06,
+      "loss": 0.9034,
+      "step": 18485
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0677667506495337e-06,
+      "loss": 1.1109,
+      "step": 18486
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0673369193373833e-06,
+      "loss": 1.0511,
+      "step": 18487
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0669071642196903e-06,
+      "loss": 0.9257,
+      "step": 18488
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0664774853047782e-06,
+      "loss": 0.8559,
+      "step": 18489
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0660478826009736e-06,
+      "loss": 1.0419,
+      "step": 18490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0656183561165972e-06,
+      "loss": 0.9516,
+      "step": 18491
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.065188905859974e-06,
+      "loss": 0.825,
+      "step": 18492
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0647595318394222e-06,
+      "loss": 0.913,
+      "step": 18493
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0643302340632633e-06,
+      "loss": 1.0481,
+      "step": 18494
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.063901012539812e-06,
+      "loss": 0.8485,
+      "step": 18495
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0634718672773863e-06,
+      "loss": 0.922,
+      "step": 18496
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0630427982842995e-06,
+      "loss": 1.1849,
+      "step": 18497
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0626138055688667e-06,
+      "loss": 1.0422,
+      "step": 18498
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0621848891393972e-06,
+      "loss": 0.9146,
+      "step": 18499
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0617560490042018e-06,
+      "loss": 0.8908,
+      "step": 18500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0613272851715912e-06,
+      "loss": 0.9925,
+      "step": 18501
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0608985976498693e-06,
+      "loss": 1.1068,
+      "step": 18502
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.060469986447345e-06,
+      "loss": 0.8476,
+      "step": 18503
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0600414515723202e-06,
+      "loss": 0.8859,
+      "step": 18504
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0596129930330995e-06,
+      "loss": 0.8991,
+      "step": 18505
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0591846108379816e-06,
+      "loss": 0.9817,
+      "step": 18506
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0587563049952699e-06,
+      "loss": 0.9765,
+      "step": 18507
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0583280755132592e-06,
+      "loss": 1.2003,
+      "step": 18508
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.057899922400249e-06,
+      "loss": 0.9441,
+      "step": 18509
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.057471845664531e-06,
+      "loss": 0.9295,
+      "step": 18510
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0570438453144044e-06,
+      "loss": 1.0203,
+      "step": 18511
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0566159213581577e-06,
+      "loss": 0.8851,
+      "step": 18512
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0561880738040841e-06,
+      "loss": 0.8141,
+      "step": 18513
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0557603026604713e-06,
+      "loss": 0.9148,
+      "step": 18514
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0553326079356086e-06,
+      "loss": 1.0122,
+      "step": 18515
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0549049896377805e-06,
+      "loss": 0.8312,
+      "step": 18516
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0544774477752751e-06,
+      "loss": 1.0274,
+      "step": 18517
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0540499823563722e-06,
+      "loss": 0.7806,
+      "step": 18518
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0536225933893574e-06,
+      "loss": 0.8164,
+      "step": 18519
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0531952808825086e-06,
+      "loss": 1.0054,
+      "step": 18520
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0527680448441069e-06,
+      "loss": 1.0403,
+      "step": 18521
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0523408852824262e-06,
+      "loss": 0.8961,
+      "step": 18522
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0519138022057485e-06,
+      "loss": 0.8332,
+      "step": 18523
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.051486795622344e-06,
+      "loss": 1.0365,
+      "step": 18524
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.051059865540488e-06,
+      "loss": 0.8813,
+      "step": 18525
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0506330119684504e-06,
+      "loss": 0.8939,
+      "step": 18526
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0502062349145037e-06,
+      "loss": 0.8601,
+      "step": 18527
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0497795343869132e-06,
+      "loss": 0.9388,
+      "step": 18528
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0493529103939503e-06,
+      "loss": 1.0564,
+      "step": 18529
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0489263629438767e-06,
+      "loss": 0.8823,
+      "step": 18530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0484998920449608e-06,
+      "loss": 1.0358,
+      "step": 18531
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0480734977054608e-06,
+      "loss": 0.7878,
+      "step": 18532
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0476471799336424e-06,
+      "loss": 0.8317,
+      "step": 18533
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0472209387377613e-06,
+      "loss": 0.8236,
+      "step": 18534
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.046794774126078e-06,
+      "loss": 0.8406,
+      "step": 18535
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0463686861068513e-06,
+      "loss": 0.8541,
+      "step": 18536
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0459426746883322e-06,
+      "loss": 0.9802,
+      "step": 18537
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0455167398787791e-06,
+      "loss": 1.0869,
+      "step": 18538
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0450908816864402e-06,
+      "loss": 1.0526,
+      "step": 18539
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0446651001195697e-06,
+      "loss": 1.0256,
+      "step": 18540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0442393951864143e-06,
+      "loss": 0.878,
+      "step": 18541
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0438137668952247e-06,
+      "loss": 0.9889,
+      "step": 18542
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0433882152542452e-06,
+      "loss": 0.9875,
+      "step": 18543
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0429627402717218e-06,
+      "loss": 0.944,
+      "step": 18544
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.042537341955896e-06,
+      "loss": 0.9904,
+      "step": 18545
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0421120203150136e-06,
+      "loss": 0.8912,
+      "step": 18546
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0416867753573119e-06,
+      "loss": 0.9855,
+      "step": 18547
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.041261607091032e-06,
+      "loss": 0.8333,
+      "step": 18548
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0408365155244094e-06,
+      "loss": 0.9735,
+      "step": 18549
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0404115006656828e-06,
+      "loss": 0.8485,
+      "step": 18550
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0399865625230837e-06,
+      "loss": 0.8339,
+      "step": 18551
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0395617011048475e-06,
+      "loss": 0.9671,
+      "step": 18552
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0391369164192044e-06,
+      "loss": 0.8543,
+      "step": 18553
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0387122084743862e-06,
+      "loss": 0.8448,
+      "step": 18554
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0382875772786188e-06,
+      "loss": 1.0434,
+      "step": 18555
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0378630228401321e-06,
+      "loss": 1.0442,
+      "step": 18556
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0374385451671482e-06,
+      "loss": 1.0618,
+      "step": 18557
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0370141442678967e-06,
+      "loss": 0.8053,
+      "step": 18558
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0365898201505953e-06,
+      "loss": 0.9204,
+      "step": 18559
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0361655728234683e-06,
+      "loss": 0.948,
+      "step": 18560
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.035741402294733e-06,
+      "loss": 0.95,
+      "step": 18561
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0353173085726103e-06,
+      "loss": 0.9357,
+      "step": 18562
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0348932916653143e-06,
+      "loss": 1.0516,
+      "step": 18563
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0344693515810622e-06,
+      "loss": 1.0189,
+      "step": 18564
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.034045488328066e-06,
+      "loss": 1.0135,
+      "step": 18565
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0336217019145396e-06,
+      "loss": 0.9698,
+      "step": 18566
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0331979923486917e-06,
+      "loss": 0.9601,
+      "step": 18567
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0327743596387335e-06,
+      "loss": 0.8458,
+      "step": 18568
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.032350803792873e-06,
+      "loss": 0.8137,
+      "step": 18569
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0319273248193145e-06,
+      "loss": 0.8976,
+      "step": 18570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0315039227262651e-06,
+      "loss": 1.001,
+      "step": 18571
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0310805975219256e-06,
+      "loss": 0.86,
+      "step": 18572
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.030657349214501e-06,
+      "loss": 0.9192,
+      "step": 18573
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0302341778121882e-06,
+      "loss": 0.7934,
+      "step": 18574
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0298110833231894e-06,
+      "loss": 1.0232,
+      "step": 18575
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0293880657556987e-06,
+      "loss": 0.8075,
+      "step": 18576
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0289651251179144e-06,
+      "loss": 0.8901,
+      "step": 18577
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0285422614180297e-06,
+      "loss": 1.0367,
+      "step": 18578
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.028119474664238e-06,
+      "loss": 0.9524,
+      "step": 18579
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.027696764864729e-06,
+      "loss": 1.0328,
+      "step": 18580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0272741320276957e-06,
+      "loss": 1.1746,
+      "step": 18581
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0268515761613246e-06,
+      "loss": 1.097,
+      "step": 18582
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0264290972738034e-06,
+      "loss": 0.9034,
+      "step": 18583
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.026006695373316e-06,
+      "loss": 0.9192,
+      "step": 18584
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.025584370468049e-06,
+      "loss": 0.792,
+      "step": 18585
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0251621225661818e-06,
+      "loss": 0.9219,
+      "step": 18586
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0247399516758983e-06,
+      "loss": 0.8174,
+      "step": 18587
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0243178578053752e-06,
+      "loss": 0.8293,
+      "step": 18588
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0238958409627931e-06,
+      "loss": 1.0803,
+      "step": 18589
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0234739011563256e-06,
+      "loss": 0.8866,
+      "step": 18590
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0230520383941506e-06,
+      "loss": 0.8955,
+      "step": 18591
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0226302526844388e-06,
+      "loss": 0.9239,
+      "step": 18592
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.022208544035364e-06,
+      "loss": 0.893,
+      "step": 18593
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.021786912455097e-06,
+      "loss": 0.8431,
+      "step": 18594
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0213653579518057e-06,
+      "loss": 0.7618,
+      "step": 18595
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0209438805336591e-06,
+      "loss": 0.9798,
+      "step": 18596
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0205224802088203e-06,
+      "loss": 0.8853,
+      "step": 18597
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.020101156985458e-06,
+      "loss": 0.9941,
+      "step": 18598
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0196799108717309e-06,
+      "loss": 0.9269,
+      "step": 18599
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0192587418758044e-06,
+      "loss": 0.8799,
+      "step": 18600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0188376500058355e-06,
+      "loss": 0.8269,
+      "step": 18601
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0184166352699854e-06,
+      "loss": 0.8385,
+      "step": 18602
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0179956976764066e-06,
+      "loss": 1.0159,
+      "step": 18603
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0175748372332611e-06,
+      "loss": 0.9837,
+      "step": 18604
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0171540539486986e-06,
+      "loss": 0.8366,
+      "step": 18605
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0167333478308738e-06,
+      "loss": 0.7788,
+      "step": 18606
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0163127188879352e-06,
+      "loss": 0.8935,
+      "step": 18607
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0158921671280358e-06,
+      "loss": 0.9784,
+      "step": 18608
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0154716925593205e-06,
+      "loss": 0.9646,
+      "step": 18609
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0150512951899387e-06,
+      "loss": 0.9103,
+      "step": 18610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0146309750280326e-06,
+      "loss": 1.0426,
+      "step": 18611
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.014210732081749e-06,
+      "loss": 0.7777,
+      "step": 18612
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0137905663592267e-06,
+      "loss": 0.7954,
+      "step": 18613
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0133704778686093e-06,
+      "loss": 0.9054,
+      "step": 18614
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0129504666180323e-06,
+      "loss": 0.9273,
+      "step": 18615
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0125305326156376e-06,
+      "loss": 1.0667,
+      "step": 18616
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0121106758695586e-06,
+      "loss": 1.0183,
+      "step": 18617
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0116908963879313e-06,
+      "loss": 0.921,
+      "step": 18618
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0112711941788877e-06,
+      "loss": 0.9902,
+      "step": 18619
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0108515692505605e-06,
+      "loss": 0.9049,
+      "step": 18620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0104320216110785e-06,
+      "loss": 0.8779,
+      "step": 18621
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.010012551268572e-06,
+      "loss": 1.0607,
+      "step": 18622
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0095931582311663e-06,
+      "loss": 0.9764,
+      "step": 18623
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0091738425069892e-06,
+      "loss": 1.0951,
+      "step": 18624
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0087546041041618e-06,
+      "loss": 0.9069,
+      "step": 18625
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.008335443030809e-06,
+      "loss": 0.7348,
+      "step": 18626
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.007916359295053e-06,
+      "loss": 0.9699,
+      "step": 18627
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0074973529050103e-06,
+      "loss": 0.8259,
+      "step": 18628
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0070784238688026e-06,
+      "loss": 0.8234,
+      "step": 18629
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0066595721945426e-06,
+      "loss": 0.8967,
+      "step": 18630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0062407978903494e-06,
+      "loss": 1.1048,
+      "step": 18631
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0058221009643331e-06,
+      "loss": 0.837,
+      "step": 18632
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0054034814246093e-06,
+      "loss": 0.9282,
+      "step": 18633
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0049849392792854e-06,
+      "loss": 0.7375,
+      "step": 18634
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0045664745364736e-06,
+      "loss": 1.0216,
+      "step": 18635
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0041480872042786e-06,
+      "loss": 0.9602,
+      "step": 18636
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0037297772908094e-06,
+      "loss": 1.0209,
+      "step": 18637
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0033115448041665e-06,
+      "loss": 0.8386,
+      "step": 18638
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0028933897524583e-06,
+      "loss": 1.0335,
+      "step": 18639
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.002475312143783e-06,
+      "loss": 0.9184,
+      "step": 18640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0020573119862425e-06,
+      "loss": 0.9921,
+      "step": 18641
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.001639389287934e-06,
+      "loss": 1.002,
+      "step": 18642
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0012215440569567e-06,
+      "loss": 0.8586,
+      "step": 18643
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0008037763014033e-06,
+      "loss": 1.0806,
+      "step": 18644
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0003860860293708e-06,
+      "loss": 0.7518,
+      "step": 18645
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.999684732489496e-07,
+      "loss": 0.7207,
+      "step": 18646
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.995509379682327e-07,
+      "loss": 0.8056,
+      "step": 18647
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.991334801953078e-07,
+      "loss": 0.8571,
+      "step": 18648
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.987160999382655e-07,
+      "loss": 1.0901,
+      "step": 18649
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.982987972051889e-07,
+      "loss": 0.9542,
+      "step": 18650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.978815720041673e-07,
+      "loss": 0.8983,
+      "step": 18651
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.974644243432818e-07,
+      "loss": 0.8664,
+      "step": 18652
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.97047354230616e-07,
+      "loss": 0.9443,
+      "step": 18653
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.966303616742484e-07,
+      "loss": 0.8852,
+      "step": 18654
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.962134466822609e-07,
+      "loss": 0.8429,
+      "step": 18655
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.957966092627286e-07,
+      "loss": 0.9597,
+      "step": 18656
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.953798494237294e-07,
+      "loss": 0.9936,
+      "step": 18657
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.94963167173336e-07,
+      "loss": 0.8132,
+      "step": 18658
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.94546562519625e-07,
+      "loss": 0.8554,
+      "step": 18659
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.941300354706635e-07,
+      "loss": 1.017,
+      "step": 18660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.937135860345236e-07,
+      "loss": 0.8434,
+      "step": 18661
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.932972142192764e-07,
+      "loss": 0.8832,
+      "step": 18662
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.928809200329852e-07,
+      "loss": 0.9415,
+      "step": 18663
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.92464703483718e-07,
+      "loss": 0.9558,
+      "step": 18664
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.920485645795374e-07,
+      "loss": 0.9065,
+      "step": 18665
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.91632503328508e-07,
+      "loss": 0.9332,
+      "step": 18666
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.912165197386881e-07,
+      "loss": 0.8999,
+      "step": 18667
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.908006138181403e-07,
+      "loss": 1.0201,
+      "step": 18668
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.903847855749194e-07,
+      "loss": 0.9061,
+      "step": 18669
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.899690350170854e-07,
+      "loss": 0.911,
+      "step": 18670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.895533621526904e-07,
+      "loss": 0.8544,
+      "step": 18671
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.891377669897906e-07,
+      "loss": 0.9731,
+      "step": 18672
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.88722249536434e-07,
+      "loss": 0.8123,
+      "step": 18673
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.883068098006765e-07,
+      "loss": 0.8994,
+      "step": 18674
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.878914477905633e-07,
+      "loss": 0.9282,
+      "step": 18675
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.874761635141444e-07,
+      "loss": 1.0828,
+      "step": 18676
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.870609569794637e-07,
+      "loss": 0.9156,
+      "step": 18677
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.86645828194568e-07,
+      "loss": 1.1293,
+      "step": 18678
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.862307771674973e-07,
+      "loss": 0.8982,
+      "step": 18679
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.85815803906296e-07,
+      "loss": 0.941,
+      "step": 18680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.85400908419002e-07,
+      "loss": 0.8128,
+      "step": 18681
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.84986090713656e-07,
+      "loss": 0.8871,
+      "step": 18682
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.845713507982925e-07,
+      "loss": 0.8995,
+      "step": 18683
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.841566886809489e-07,
+      "loss": 1.0692,
+      "step": 18684
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.837421043696577e-07,
+      "loss": 0.8752,
+      "step": 18685
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.833275978724522e-07,
+      "loss": 0.8931,
+      "step": 18686
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.829131691973648e-07,
+      "loss": 1.0126,
+      "step": 18687
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.82498818352422e-07,
+      "loss": 0.897,
+      "step": 18688
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.820845453456545e-07,
+      "loss": 0.8998,
+      "step": 18689
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.816703501850865e-07,
+      "loss": 1.0916,
+      "step": 18690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.81256232878745e-07,
+      "loss": 0.9132,
+      "step": 18691
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.808421934346512e-07,
+      "loss": 0.9962,
+      "step": 18692
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.804282318608298e-07,
+      "loss": 0.9751,
+      "step": 18693
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.80014348165298e-07,
+      "loss": 0.7765,
+      "step": 18694
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.796005423560778e-07,
+      "loss": 0.881,
+      "step": 18695
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.79186814441182e-07,
+      "loss": 0.9538,
+      "step": 18696
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.787731644286325e-07,
+      "loss": 0.9238,
+      "step": 18697
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.7835959232644e-07,
+      "loss": 0.9295,
+      "step": 18698
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.779460981426187e-07,
+      "loss": 0.8619,
+      "step": 18699
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.775326818851783e-07,
+      "loss": 0.8335,
+      "step": 18700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.771193435621312e-07,
+      "loss": 0.8804,
+      "step": 18701
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.767060831814828e-07,
+      "loss": 0.8947,
+      "step": 18702
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.762929007512428e-07,
+      "loss": 0.9884,
+      "step": 18703
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.758797962794138e-07,
+      "loss": 0.9514,
+      "step": 18704
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.754667697740023e-07,
+      "loss": 0.9887,
+      "step": 18705
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.750538212430077e-07,
+      "loss": 0.9705,
+      "step": 18706
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.746409506944337e-07,
+      "loss": 0.9284,
+      "step": 18707
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.742281581362755e-07,
+      "loss": 1.0057,
+      "step": 18708
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.738154435765363e-07,
+      "loss": 0.9995,
+      "step": 18709
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.734028070232083e-07,
+      "loss": 1.0089,
+      "step": 18710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.72990248484289e-07,
+      "loss": 0.9176,
+      "step": 18711
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.725777679677683e-07,
+      "loss": 1.1094,
+      "step": 18712
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.721653654816416e-07,
+      "loss": 0.8975,
+      "step": 18713
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.717530410338961e-07,
+      "loss": 0.8475,
+      "step": 18714
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.713407946325222e-07,
+      "loss": 0.9957,
+      "step": 18715
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.709286262855055e-07,
+      "loss": 1.0651,
+      "step": 18716
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.705165360008345e-07,
+      "loss": 0.9898,
+      "step": 18717
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.701045237864898e-07,
+      "loss": 0.9193,
+      "step": 18718
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.69692589650456e-07,
+      "loss": 0.902,
+      "step": 18719
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.692807336007154e-07,
+      "loss": 0.9749,
+      "step": 18720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.688689556452452e-07,
+      "loss": 0.9643,
+      "step": 18721
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.684572557920258e-07,
+      "loss": 1.0463,
+      "step": 18722
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.680456340490307e-07,
+      "loss": 0.827,
+      "step": 18723
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.676340904242386e-07,
+      "loss": 0.843,
+      "step": 18724
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.6722262492562e-07,
+      "loss": 1.0046,
+      "step": 18725
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.668112375611494e-07,
+      "loss": 0.9206,
+      "step": 18726
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.663999283387949e-07,
+      "loss": 0.8906,
+      "step": 18727
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.659886972665284e-07,
+      "loss": 1.0091,
+      "step": 18728
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.65577544352314e-07,
+      "loss": 0.9415,
+      "step": 18729
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.651664696041214e-07,
+      "loss": 1.0645,
+      "step": 18730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.647554730299102e-07,
+      "loss": 0.8907,
+      "step": 18731
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.64344554637649e-07,
+      "loss": 0.9287,
+      "step": 18732
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.639337144352956e-07,
+      "loss": 1.0717,
+      "step": 18733
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.63522952430812e-07,
+      "loss": 1.0598,
+      "step": 18734
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.631122686321542e-07,
+      "loss": 0.9874,
+      "step": 18735
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.62701663047282e-07,
+      "loss": 0.8357,
+      "step": 18736
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.622911356841475e-07,
+      "loss": 0.9699,
+      "step": 18737
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.61880686550708e-07,
+      "loss": 1.037,
+      "step": 18738
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.614703156549132e-07,
+      "loss": 0.9741,
+      "step": 18739
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.610600230047158e-07,
+      "loss": 0.8825,
+      "step": 18740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.60649808608063e-07,
+      "loss": 0.8912,
+      "step": 18741
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.602396724729051e-07,
+      "loss": 1.0153,
+      "step": 18742
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.59829614607185e-07,
+      "loss": 0.9222,
+      "step": 18743
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.594196350188523e-07,
+      "loss": 1.1158,
+      "step": 18744
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.59009733715846e-07,
+      "loss": 0.7729,
+      "step": 18745
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.585999107061105e-07,
+      "loss": 1.0007,
+      "step": 18746
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.58190165997584e-07,
+      "loss": 0.9356,
+      "step": 18747
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.577804995982075e-07,
+      "loss": 0.8469,
+      "step": 18748
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.573709115159157e-07,
+      "loss": 0.8449,
+      "step": 18749
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.569614017586465e-07,
+      "loss": 0.9157,
+      "step": 18750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.565519703343319e-07,
+      "loss": 0.7127,
+      "step": 18751
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.56142617250907e-07,
+      "loss": 0.794,
+      "step": 18752
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.557333425163e-07,
+      "loss": 0.9224,
+      "step": 18753
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.553241461384428e-07,
+      "loss": 1.0221,
+      "step": 18754
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.549150281252633e-07,
+      "loss": 0.8299,
+      "step": 18755
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.545059884846863e-07,
+      "loss": 0.7741,
+      "step": 18756
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.540970272246397e-07,
+      "loss": 0.9152,
+      "step": 18757
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.536881443530444e-07,
+      "loss": 0.8756,
+      "step": 18758
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.532793398778245e-07,
+      "loss": 1.1036,
+      "step": 18759
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.528706138068977e-07,
+      "loss": 0.8261,
+      "step": 18760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.524619661481865e-07,
+      "loss": 1.0229,
+      "step": 18761
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.520533969096052e-07,
+      "loss": 1.0723,
+      "step": 18762
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.516449060990723e-07,
+      "loss": 0.8992,
+      "step": 18763
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.512364937245e-07,
+      "loss": 0.8156,
+      "step": 18764
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.508281597938029e-07,
+      "loss": 0.8711,
+      "step": 18765
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.504199043148898e-07,
+      "loss": 0.8717,
+      "step": 18766
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.500117272956749e-07,
+      "loss": 0.9052,
+      "step": 18767
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.496036287440624e-07,
+      "loss": 1.1222,
+      "step": 18768
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.491956086679621e-07,
+      "loss": 0.9843,
+      "step": 18769
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.487876670752772e-07,
+      "loss": 0.7498,
+      "step": 18770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.483798039739134e-07,
+      "loss": 1.0674,
+      "step": 18771
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.479720193717706e-07,
+      "loss": 0.926,
+      "step": 18772
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.47564313276752e-07,
+      "loss": 0.7992,
+      "step": 18773
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.47156685696754e-07,
+      "loss": 0.9517,
+      "step": 18774
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.467491366396775e-07,
+      "loss": 0.895,
+      "step": 18775
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.463416661134156e-07,
+      "loss": 0.7338,
+      "step": 18776
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.45934274125866e-07,
+      "loss": 0.8792,
+      "step": 18777
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.455269606849187e-07,
+      "loss": 0.9299,
+      "step": 18778
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.451197257984668e-07,
+      "loss": 0.8551,
+      "step": 18779
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.447125694744013e-07,
+      "loss": 0.9253,
+      "step": 18780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.443054917206091e-07,
+      "loss": 0.9172,
+      "step": 18781
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.438984925449795e-07,
+      "loss": 0.9924,
+      "step": 18782
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.43491571955395e-07,
+      "loss": 0.9117,
+      "step": 18783
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.430847299597423e-07,
+      "loss": 0.9704,
+      "step": 18784
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.426779665659013e-07,
+      "loss": 0.8977,
+      "step": 18785
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.422712817817559e-07,
+      "loss": 0.8872,
+      "step": 18786
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.418646756151823e-07,
+      "loss": 0.867,
+      "step": 18787
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.41458148074062e-07,
+      "loss": 0.9482,
+      "step": 18788
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.410516991662666e-07,
+      "loss": 0.8793,
+      "step": 18789
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.406453288996759e-07,
+      "loss": 0.9584,
+      "step": 18790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.402390372821602e-07,
+      "loss": 1.0368,
+      "step": 18791
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.398328243215937e-07,
+      "loss": 0.9189,
+      "step": 18792
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.394266900258436e-07,
+      "loss": 0.961,
+      "step": 18793
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.390206344027819e-07,
+      "loss": 1.0051,
+      "step": 18794
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.386146574602729e-07,
+      "loss": 0.9641,
+      "step": 18795
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.382087592061851e-07,
+      "loss": 0.9083,
+      "step": 18796
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.378029396483795e-07,
+      "loss": 1.0427,
+      "step": 18797
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.373971987947217e-07,
+      "loss": 0.8549,
+      "step": 18798
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.369915366530708e-07,
+      "loss": 0.9296,
+      "step": 18799
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.365859532312882e-07,
+      "loss": 0.9653,
+      "step": 18800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.361804485372289e-07,
+      "loss": 0.9986,
+      "step": 18801
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.357750225787537e-07,
+      "loss": 0.9301,
+      "step": 18802
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.353696753637142e-07,
+      "loss": 0.9257,
+      "step": 18803
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.349644068999669e-07,
+      "loss": 0.8633,
+      "step": 18804
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.345592171953605e-07,
+      "loss": 0.8571,
+      "step": 18805
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.34154106257748e-07,
+      "loss": 0.8852,
+      "step": 18806
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.337490740949767e-07,
+      "loss": 0.8884,
+      "step": 18807
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.333441207148958e-07,
+      "loss": 0.9147,
+      "step": 18808
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.32939246125349e-07,
+      "loss": 1.148,
+      "step": 18809
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.325344503341832e-07,
+      "loss": 0.8575,
+      "step": 18810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.321297333492379e-07,
+      "loss": 0.8772,
+      "step": 18811
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.317250951783563e-07,
+      "loss": 0.8133,
+      "step": 18812
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.313205358293798e-07,
+      "loss": 1.1911,
+      "step": 18813
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.309160553101437e-07,
+      "loss": 1.0245,
+      "step": 18814
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.30511653628487e-07,
+      "loss": 0.9369,
+      "step": 18815
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.301073307922432e-07,
+      "loss": 0.8547,
+      "step": 18816
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.297030868092472e-07,
+      "loss": 1.1619,
+      "step": 18817
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.292989216873294e-07,
+      "loss": 0.9283,
+      "step": 18818
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.288948354343236e-07,
+      "loss": 0.9523,
+      "step": 18819
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.284908280580552e-07,
+      "loss": 0.8672,
+      "step": 18820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.280868995663545e-07,
+      "loss": 1.0405,
+      "step": 18821
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.276830499670453e-07,
+      "loss": 0.8325,
+      "step": 18822
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.272792792679541e-07,
+      "loss": 1.0283,
+      "step": 18823
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.268755874769009e-07,
+      "loss": 0.9397,
+      "step": 18824
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.26471974601712e-07,
+      "loss": 1.0646,
+      "step": 18825
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.260684406502029e-07,
+      "loss": 0.8134,
+      "step": 18826
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.256649856301953e-07,
+      "loss": 1.001,
+      "step": 18827
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.252616095495031e-07,
+      "loss": 0.9969,
+      "step": 18828
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.248583124159438e-07,
+      "loss": 0.9329,
+      "step": 18829
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.244550942373293e-07,
+      "loss": 0.6876,
+      "step": 18830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.240519550214738e-07,
+      "loss": 1.0053,
+      "step": 18831
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.236488947761862e-07,
+      "loss": 1.1152,
+      "step": 18832
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.23245913509278e-07,
+      "loss": 0.934,
+      "step": 18833
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.228430112285536e-07,
+      "loss": 1.0549,
+      "step": 18834
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.224401879418226e-07,
+      "loss": 1.0047,
+      "step": 18835
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.220374436568869e-07,
+      "loss": 0.8065,
+      "step": 18836
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.216347783815505e-07,
+      "loss": 0.7978,
+      "step": 18837
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.212321921236167e-07,
+      "loss": 0.9222,
+      "step": 18838
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.208296848908821e-07,
+      "loss": 0.8933,
+      "step": 18839
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.204272566911487e-07,
+      "loss": 1.0877,
+      "step": 18840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.200249075322104e-07,
+      "loss": 1.0117,
+      "step": 18841
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.196226374218642e-07,
+      "loss": 0.894,
+      "step": 18842
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.192204463679049e-07,
+      "loss": 0.8555,
+      "step": 18843
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.188183343781226e-07,
+      "loss": 0.8969,
+      "step": 18844
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.184163014603109e-07,
+      "loss": 0.9533,
+      "step": 18845
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.180143476222558e-07,
+      "loss": 0.8976,
+      "step": 18846
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.176124728717467e-07,
+      "loss": 0.8671,
+      "step": 18847
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.172106772165718e-07,
+      "loss": 1.1752,
+      "step": 18848
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.16808960664512e-07,
+      "loss": 0.7514,
+      "step": 18849
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.164073232233539e-07,
+      "loss": 0.9168,
+      "step": 18850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.160057649008758e-07,
+      "loss": 0.8981,
+      "step": 18851
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.156042857048614e-07,
+      "loss": 0.9305,
+      "step": 18852
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.152028856430861e-07,
+      "loss": 0.8608,
+      "step": 18853
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.148015647233299e-07,
+      "loss": 0.9801,
+      "step": 18854
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.144003229533649e-07,
+      "loss": 0.8777,
+      "step": 18855
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.139991603409681e-07,
+      "loss": 1.0323,
+      "step": 18856
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.135980768939101e-07,
+      "loss": 0.9432,
+      "step": 18857
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.13197072619963e-07,
+      "loss": 0.9839,
+      "step": 18858
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.127961475268937e-07,
+      "loss": 0.9304,
+      "step": 18859
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.123953016224741e-07,
+      "loss": 0.822,
+      "step": 18860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.119945349144665e-07,
+      "loss": 0.8799,
+      "step": 18861
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.115938474106395e-07,
+      "loss": 0.9592,
+      "step": 18862
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.111932391187527e-07,
+      "loss": 0.9722,
+      "step": 18863
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.107927100465708e-07,
+      "loss": 0.8724,
+      "step": 18864
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.103922602018511e-07,
+      "loss": 1.1108,
+      "step": 18865
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.099918895923554e-07,
+      "loss": 0.9298,
+      "step": 18866
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.095915982258375e-07,
+      "loss": 0.8152,
+      "step": 18867
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.091913861100554e-07,
+      "loss": 0.969,
+      "step": 18868
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.087912532527615e-07,
+      "loss": 1.0424,
+      "step": 18869
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.083911996617101e-07,
+      "loss": 0.7768,
+      "step": 18870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.079912253446499e-07,
+      "loss": 1.0563,
+      "step": 18871
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.075913303093315e-07,
+      "loss": 0.9082,
+      "step": 18872
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.07191514563504e-07,
+      "loss": 1.0149,
+      "step": 18873
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.06791778114911e-07,
+      "loss": 0.8095,
+      "step": 18874
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.063921209713e-07,
+      "loss": 1.1177,
+      "step": 18875
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.059925431404121e-07,
+      "loss": 0.9154,
+      "step": 18876
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.055930446299916e-07,
+      "loss": 0.8431,
+      "step": 18877
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.051936254477756e-07,
+      "loss": 0.9476,
+      "step": 18878
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.047942856015052e-07,
+      "loss": 0.976,
+      "step": 18879
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.043950250989153e-07,
+      "loss": 0.9294,
+      "step": 18880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.039958439477442e-07,
+      "loss": 0.9225,
+      "step": 18881
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.035967421557218e-07,
+      "loss": 0.8759,
+      "step": 18882
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.031977197305863e-07,
+      "loss": 0.9129,
+      "step": 18883
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.027987766800639e-07,
+      "loss": 0.7552,
+      "step": 18884
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.023999130118877e-07,
+      "loss": 0.8929,
+      "step": 18885
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.020011287337815e-07,
+      "loss": 0.9297,
+      "step": 18886
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.01602423853476e-07,
+      "loss": 0.9802,
+      "step": 18887
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.01203798378692e-07,
+      "loss": 1.015,
+      "step": 18888
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.008052523171563e-07,
+      "loss": 0.8767,
+      "step": 18889
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.004067856765875e-07,
+      "loss": 1.0664,
+      "step": 18890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.000083984647085e-07,
+      "loss": 0.9429,
+      "step": 18891
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.996100906892358e-07,
+      "loss": 0.8927,
+      "step": 18892
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.992118623578883e-07,
+      "loss": 1.0728,
+      "step": 18893
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.988137134783781e-07,
+      "loss": 1.0447,
+      "step": 18894
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.984156440584252e-07,
+      "loss": 0.947,
+      "step": 18895
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.980176541057367e-07,
+      "loss": 1.1276,
+      "step": 18896
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.976197436280276e-07,
+      "loss": 0.936,
+      "step": 18897
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.97221912633004e-07,
+      "loss": 0.9693,
+      "step": 18898
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.968241611283762e-07,
+      "loss": 1.1309,
+      "step": 18899
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.964264891218483e-07,
+      "loss": 0.9448,
+      "step": 18900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.960288966211272e-07,
+      "loss": 1.0473,
+      "step": 18901
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.956313836339148e-07,
+      "loss": 0.8489,
+      "step": 18902
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.952339501679142e-07,
+      "loss": 0.8797,
+      "step": 18903
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.948365962308236e-07,
+      "loss": 0.8235,
+      "step": 18904
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.944393218303427e-07,
+      "loss": 0.8314,
+      "step": 18905
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.940421269741695e-07,
+      "loss": 0.895,
+      "step": 18906
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.93645011669998e-07,
+      "loss": 0.9893,
+      "step": 18907
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.932479759255236e-07,
+      "loss": 1.0294,
+      "step": 18908
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.928510197484375e-07,
+      "loss": 0.9981,
+      "step": 18909
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.924541431464323e-07,
+      "loss": 0.9949,
+      "step": 18910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.920573461271948e-07,
+      "loss": 1.0273,
+      "step": 18911
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.916606286984158e-07,
+      "loss": 0.9489,
+      "step": 18912
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.91263990867779e-07,
+      "loss": 1.0364,
+      "step": 18913
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.908674326429711e-07,
+      "loss": 0.9537,
+      "step": 18914
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.904709540316741e-07,
+      "loss": 1.153,
+      "step": 18915
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.900745550415713e-07,
+      "loss": 1.0373,
+      "step": 18916
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.896782356803385e-07,
+      "loss": 0.9609,
+      "step": 18917
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.892819959556609e-07,
+      "loss": 0.8672,
+      "step": 18918
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.888858358752106e-07,
+      "loss": 0.9775,
+      "step": 18919
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.884897554466654e-07,
+      "loss": 1.0617,
+      "step": 18920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.88093754677698e-07,
+      "loss": 1.0806,
+      "step": 18921
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.876978335759823e-07,
+      "loss": 1.0886,
+      "step": 18922
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.873019921491872e-07,
+      "loss": 1.0098,
+      "step": 18923
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.869062304049841e-07,
+      "loss": 0.8111,
+      "step": 18924
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.865105483510383e-07,
+      "loss": 1.0361,
+      "step": 18925
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.861149459950191e-07,
+      "loss": 0.9142,
+      "step": 18926
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.857194233445887e-07,
+      "loss": 1.0111,
+      "step": 18927
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.853239804074121e-07,
+      "loss": 0.9359,
+      "step": 18928
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.849286171911487e-07,
+      "loss": 0.7505,
+      "step": 18929
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.845333337034601e-07,
+      "loss": 0.8699,
+      "step": 18930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.841381299520058e-07,
+      "loss": 0.8966,
+      "step": 18931
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.837430059444402e-07,
+      "loss": 0.8679,
+      "step": 18932
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.833479616884217e-07,
+      "loss": 0.9624,
+      "step": 18933
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.829529971916012e-07,
+      "loss": 1.0273,
+      "step": 18934
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.825581124616334e-07,
+      "loss": 0.6647,
+      "step": 18935
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.821633075061675e-07,
+      "loss": 1.0736,
+      "step": 18936
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.817685823328542e-07,
+      "loss": 0.8968,
+      "step": 18937
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.813739369493395e-07,
+      "loss": 0.9153,
+      "step": 18938
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.809793713632702e-07,
+      "loss": 1.0048,
+      "step": 18939
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.805848855822918e-07,
+      "loss": 0.9263,
+      "step": 18940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.801904796140476e-07,
+      "loss": 0.9835,
+      "step": 18941
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.79796153466177e-07,
+      "loss": 0.9209,
+      "step": 18942
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.794019071463228e-07,
+      "loss": 1.0854,
+      "step": 18943
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.790077406621206e-07,
+      "loss": 0.921,
+      "step": 18944
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.786136540212098e-07,
+      "loss": 1.0743,
+      "step": 18945
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.782196472312238e-07,
+      "loss": 1.1335,
+      "step": 18946
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.778257202997975e-07,
+      "loss": 0.9723,
+      "step": 18947
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.774318732345621e-07,
+      "loss": 0.9456,
+      "step": 18948
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.770381060431504e-07,
+      "loss": 1.001,
+      "step": 18949
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.766444187331885e-07,
+      "loss": 0.8758,
+      "step": 18950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.762508113123075e-07,
+      "loss": 1.0112,
+      "step": 18951
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.758572837881285e-07,
+      "loss": 0.9277,
+      "step": 18952
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.754638361682821e-07,
+      "loss": 0.9528,
+      "step": 18953
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.750704684603867e-07,
+      "loss": 0.9106,
+      "step": 18954
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.746771806720666e-07,
+      "loss": 1.0782,
+      "step": 18955
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.742839728109392e-07,
+      "loss": 0.9277,
+      "step": 18956
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.738908448846251e-07,
+      "loss": 0.9304,
+      "step": 18957
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.734977969007386e-07,
+      "loss": 0.9334,
+      "step": 18958
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.731048288668976e-07,
+      "loss": 0.8175,
+      "step": 18959
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.727119407907131e-07,
+      "loss": 0.8737,
+      "step": 18960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.723191326797997e-07,
+      "loss": 1.0084,
+      "step": 18961
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.719264045417658e-07,
+      "loss": 1.0235,
+      "step": 18962
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.715337563842224e-07,
+      "loss": 0.986,
+      "step": 18963
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.711411882147747e-07,
+      "loss": 0.8596,
+      "step": 18964
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.707487000410303e-07,
+      "loss": 0.9437,
+      "step": 18965
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.703562918705943e-07,
+      "loss": 0.9002,
+      "step": 18966
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.699639637110669e-07,
+      "loss": 0.9192,
+      "step": 18967
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.695717155700522e-07,
+      "loss": 1.0538,
+      "step": 18968
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.691795474551473e-07,
+      "loss": 0.8774,
+      "step": 18969
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.687874593739526e-07,
+      "loss": 0.7929,
+      "step": 18970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.683954513340632e-07,
+      "loss": 1.0041,
+      "step": 18971
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.680035233430756e-07,
+      "loss": 0.695,
+      "step": 18972
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.676116754085811e-07,
+      "loss": 0.9459,
+      "step": 18973
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.672199075381743e-07,
+      "loss": 0.8468,
+      "step": 18974
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.668282197394417e-07,
+      "loss": 0.9606,
+      "step": 18975
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.664366120199774e-07,
+      "loss": 0.8573,
+      "step": 18976
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.660450843873647e-07,
+      "loss": 0.8821,
+      "step": 18977
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.656536368491924e-07,
+      "loss": 0.8993,
+      "step": 18978
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.652622694130414e-07,
+      "loss": 0.8661,
+      "step": 18979
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.648709820864976e-07,
+      "loss": 0.9765,
+      "step": 18980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.644797748771394e-07,
+      "loss": 0.9391,
+      "step": 18981
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.640886477925481e-07,
+      "loss": 0.8527,
+      "step": 18982
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.636976008403003e-07,
+      "loss": 0.9802,
+      "step": 18983
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.633066340279744e-07,
+      "loss": 0.8899,
+      "step": 18984
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.629157473631428e-07,
+      "loss": 0.9049,
+      "step": 18985
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.625249408533814e-07,
+      "loss": 0.6916,
+      "step": 18986
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.621342145062583e-07,
+      "loss": 1.0457,
+      "step": 18987
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.617435683293484e-07,
+      "loss": 1.0312,
+      "step": 18988
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.613530023302169e-07,
+      "loss": 0.9749,
+      "step": 18989
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.609625165164337e-07,
+      "loss": 0.9715,
+      "step": 18990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.605721108955617e-07,
+      "loss": 1.02,
+      "step": 18991
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.601817854751671e-07,
+      "loss": 0.9348,
+      "step": 18992
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.597915402628099e-07,
+      "loss": 0.9516,
+      "step": 18993
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.594013752660541e-07,
+      "loss": 0.8207,
+      "step": 18994
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.590112904924558e-07,
+      "loss": 1.0438,
+      "step": 18995
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.586212859495758e-07,
+      "loss": 1.0149,
+      "step": 18996
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.582313616449678e-07,
+      "loss": 0.8882,
+      "step": 18997
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.578415175861876e-07,
+      "loss": 0.668,
+      "step": 18998
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.574517537807897e-07,
+      "loss": 0.9872,
+      "step": 18999
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.57062070236323e-07,
+      "loss": 0.9069,
+      "step": 19000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.566724669603405e-07,
+      "loss": 0.9442,
+      "step": 19001
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.562829439603876e-07,
+      "loss": 1.0205,
+      "step": 19002
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.55893501244014e-07,
+      "loss": 0.7958,
+      "step": 19003
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.555041388187624e-07,
+      "loss": 0.8781,
+      "step": 19004
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.551148566921791e-07,
+      "loss": 0.8685,
+      "step": 19005
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.547256548718042e-07,
+      "loss": 0.9366,
+      "step": 19006
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.54336533365181e-07,
+      "loss": 0.9372,
+      "step": 19007
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.539474921798452e-07,
+      "loss": 1.011,
+      "step": 19008
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.53558531323338e-07,
+      "loss": 0.9186,
+      "step": 19009
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.531696508031911e-07,
+      "loss": 0.9664,
+      "step": 19010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.527808506269436e-07,
+      "loss": 0.8753,
+      "step": 19011
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.523921308021255e-07,
+      "loss": 0.9257,
+      "step": 19012
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.520034913362707e-07,
+      "loss": 1.17,
+      "step": 19013
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.516149322369055e-07,
+      "loss": 1.0024,
+      "step": 19014
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.512264535115611e-07,
+      "loss": 1.0452,
+      "step": 19015
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.508380551677619e-07,
+      "loss": 0.9971,
+      "step": 19016
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.50449737213036e-07,
+      "loss": 0.9736,
+      "step": 19017
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.500614996549033e-07,
+      "loss": 0.8464,
+      "step": 19018
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.496733425008891e-07,
+      "loss": 0.9745,
+      "step": 19019
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.492852657585115e-07,
+      "loss": 0.942,
+      "step": 19020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.488972694352915e-07,
+      "loss": 0.9147,
+      "step": 19021
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.485093535387439e-07,
+      "loss": 0.9564,
+      "step": 19022
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.481215180763857e-07,
+      "loss": 0.8796,
+      "step": 19023
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.477337630557325e-07,
+      "loss": 0.9791,
+      "step": 19024
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.473460884842949e-07,
+      "loss": 0.8221,
+      "step": 19025
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.469584943695858e-07,
+      "loss": 0.9285,
+      "step": 19026
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.465709807191125e-07,
+      "loss": 0.9516,
+      "step": 19027
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.461835475403857e-07,
+      "loss": 1.1025,
+      "step": 19028
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.457961948409094e-07,
+      "loss": 0.8719,
+      "step": 19029
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.454089226281903e-07,
+      "loss": 1.1514,
+      "step": 19030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.450217309097297e-07,
+      "loss": 0.9895,
+      "step": 19031
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.446346196930321e-07,
+      "loss": 1.1268,
+      "step": 19032
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.442475889855933e-07,
+      "loss": 0.9113,
+      "step": 19033
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.438606387949178e-07,
+      "loss": 0.8644,
+      "step": 19034
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.434737691284977e-07,
+      "loss": 0.8518,
+      "step": 19035
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.430869799938323e-07,
+      "loss": 0.8916,
+      "step": 19036
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.427002713984122e-07,
+      "loss": 0.8357,
+      "step": 19037
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.423136433497325e-07,
+      "loss": 1.0282,
+      "step": 19038
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.419270958552816e-07,
+      "loss": 0.9518,
+      "step": 19039
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.415406289225514e-07,
+      "loss": 0.8358,
+      "step": 19040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.411542425590275e-07,
+      "loss": 0.8106,
+      "step": 19041
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.407679367721972e-07,
+      "loss": 0.8009,
+      "step": 19042
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.403817115695434e-07,
+      "loss": 1.0295,
+      "step": 19043
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.399955669585518e-07,
+      "loss": 0.8184,
+      "step": 19044
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.396095029467005e-07,
+      "loss": 0.8333,
+      "step": 19045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.392235195414732e-07,
+      "loss": 0.9663,
+      "step": 19046
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.388376167503454e-07,
+      "loss": 0.9692,
+      "step": 19047
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.384517945807957e-07,
+      "loss": 1.1303,
+      "step": 19048
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.380660530402979e-07,
+      "loss": 0.8809,
+      "step": 19049
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.37680392136327e-07,
+      "loss": 0.9599,
+      "step": 19050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.372948118763536e-07,
+      "loss": 1.0307,
+      "step": 19051
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.369093122678496e-07,
+      "loss": 0.9328,
+      "step": 19052
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.365238933182823e-07,
+      "loss": 0.942,
+      "step": 19053
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.361385550351214e-07,
+      "loss": 0.964,
+      "step": 19054
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.357532974258304e-07,
+      "loss": 0.9215,
+      "step": 19055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.353681204978753e-07,
+      "loss": 0.9629,
+      "step": 19056
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.349830242587176e-07,
+      "loss": 0.9658,
+      "step": 19057
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.34598008715819e-07,
+      "loss": 0.9939,
+      "step": 19058
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.342130738766396e-07,
+      "loss": 0.8471,
+      "step": 19059
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.338282197486364e-07,
+      "loss": 0.8798,
+      "step": 19060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.334434463392677e-07,
+      "loss": 0.9383,
+      "step": 19061
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.330587536559853e-07,
+      "loss": 0.924,
+      "step": 19062
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.326741417062462e-07,
+      "loss": 0.8647,
+      "step": 19063
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.322896104974987e-07,
+      "loss": 0.8388,
+      "step": 19064
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.319051600371958e-07,
+      "loss": 0.8511,
+      "step": 19065
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.315207903327838e-07,
+      "loss": 0.677,
+      "step": 19066
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.311365013917122e-07,
+      "loss": 0.7855,
+      "step": 19067
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.307522932214224e-07,
+      "loss": 0.8786,
+      "step": 19068
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.303681658293639e-07,
+      "loss": 1.0375,
+      "step": 19069
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.299841192229752e-07,
+      "loss": 1.0022,
+      "step": 19070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.296001534096998e-07,
+      "loss": 0.9055,
+      "step": 19071
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.292162683969734e-07,
+      "loss": 0.943,
+      "step": 19072
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.288324641922374e-07,
+      "loss": 0.9445,
+      "step": 19073
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.284487408029252e-07,
+      "loss": 1.13,
+      "step": 19074
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.280650982364735e-07,
+      "loss": 1.0511,
+      "step": 19075
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.276815365003127e-07,
+      "loss": 1.0198,
+      "step": 19076
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.272980556018773e-07,
+      "loss": 0.9274,
+      "step": 19077
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.269146555485941e-07,
+      "loss": 0.796,
+      "step": 19078
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.265313363478938e-07,
+      "loss": 0.8247,
+      "step": 19079
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.261480980072001e-07,
+      "loss": 0.8477,
+      "step": 19080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.257649405339424e-07,
+      "loss": 0.88,
+      "step": 19081
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.253818639355409e-07,
+      "loss": 0.9834,
+      "step": 19082
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.249988682194205e-07,
+      "loss": 0.9466,
+      "step": 19083
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.246159533929982e-07,
+      "loss": 0.8141,
+      "step": 19084
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.24233119463696e-07,
+      "loss": 0.9807,
+      "step": 19085
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.238503664389286e-07,
+      "loss": 0.8459,
+      "step": 19086
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.234676943261144e-07,
+      "loss": 1.029,
+      "step": 19087
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.230851031326653e-07,
+      "loss": 1.0068,
+      "step": 19088
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.227025928659954e-07,
+      "loss": 0.9819,
+      "step": 19089
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.223201635335148e-07,
+      "loss": 0.8327,
+      "step": 19090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.219378151426327e-07,
+      "loss": 0.8488,
+      "step": 19091
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.215555477007591e-07,
+      "loss": 0.8245,
+      "step": 19092
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.211733612152978e-07,
+      "loss": 0.8691,
+      "step": 19093
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.207912556936559e-07,
+      "loss": 1.0428,
+      "step": 19094
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.204092311432344e-07,
+      "loss": 0.9712,
+      "step": 19095
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.200272875714366e-07,
+      "loss": 0.8963,
+      "step": 19096
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.196454249856611e-07,
+      "loss": 1.0242,
+      "step": 19097
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.192636433933088e-07,
+      "loss": 0.9412,
+      "step": 19098
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.188819428017731e-07,
+      "loss": 1.0018,
+      "step": 19099
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.18500323218453e-07,
+      "loss": 0.9224,
+      "step": 19100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.181187846507393e-07,
+      "loss": 0.9221,
+      "step": 19101
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.177373271060268e-07,
+      "loss": 0.9741,
+      "step": 19102
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.173559505917023e-07,
+      "loss": 0.8806,
+      "step": 19103
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.169746551151603e-07,
+      "loss": 0.9137,
+      "step": 19104
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.165934406837838e-07,
+      "loss": 1.0025,
+      "step": 19105
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.162123073049616e-07,
+      "loss": 1.0653,
+      "step": 19106
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.158312549860758e-07,
+      "loss": 0.8531,
+      "step": 19107
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.154502837345119e-07,
+      "loss": 0.9969,
+      "step": 19108
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.150693935576481e-07,
+      "loss": 0.8915,
+      "step": 19109
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.146885844628666e-07,
+      "loss": 0.7644,
+      "step": 19110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.143078564575435e-07,
+      "loss": 1.0983,
+      "step": 19111
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.139272095490569e-07,
+      "loss": 1.0249,
+      "step": 19112
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.135466437447803e-07,
+      "loss": 0.9816,
+      "step": 19113
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.13166159052089e-07,
+      "loss": 0.8358,
+      "step": 19114
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.127857554783519e-07,
+      "loss": 0.9011,
+      "step": 19115
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.124054330309416e-07,
+      "loss": 1.0847,
+      "step": 19116
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.120251917172273e-07,
+      "loss": 0.9699,
+      "step": 19117
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.116450315445734e-07,
+      "loss": 0.8674,
+      "step": 19118
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.112649525203481e-07,
+      "loss": 0.7908,
+      "step": 19119
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.108849546519132e-07,
+      "loss": 1.1377,
+      "step": 19120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.105050379466334e-07,
+      "loss": 0.9087,
+      "step": 19121
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.101252024118666e-07,
+      "loss": 0.8918,
+      "step": 19122
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.097454480549749e-07,
+      "loss": 0.8365,
+      "step": 19123
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.093657748833134e-07,
+      "loss": 0.8403,
+      "step": 19124
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.089861829042406e-07,
+      "loss": 0.9062,
+      "step": 19125
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.086066721251073e-07,
+      "loss": 0.9143,
+      "step": 19126
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.082272425532716e-07,
+      "loss": 0.8457,
+      "step": 19127
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.07847894196081e-07,
+      "loss": 0.9787,
+      "step": 19128
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.074686270608878e-07,
+      "loss": 0.9804,
+      "step": 19129
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.070894411550373e-07,
+      "loss": 0.9888,
+      "step": 19130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.067103364858798e-07,
+      "loss": 0.824,
+      "step": 19131
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.063313130607564e-07,
+      "loss": 0.8992,
+      "step": 19132
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.059523708870138e-07,
+      "loss": 1.0201,
+      "step": 19133
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.055735099719919e-07,
+      "loss": 0.9292,
+      "step": 19134
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.051947303230323e-07,
+      "loss": 1.0049,
+      "step": 19135
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.048160319474723e-07,
+      "loss": 0.9145,
+      "step": 19136
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.044374148526518e-07,
+      "loss": 0.8181,
+      "step": 19137
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.04058879045902e-07,
+      "loss": 0.9154,
+      "step": 19138
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.036804245345614e-07,
+      "loss": 0.9889,
+      "step": 19139
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.033020513259604e-07,
+      "loss": 0.9663,
+      "step": 19140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.029237594274308e-07,
+      "loss": 0.9081,
+      "step": 19141
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.025455488463002e-07,
+      "loss": 0.8591,
+      "step": 19142
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.021674195898987e-07,
+      "loss": 0.886,
+      "step": 19143
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.017893716655501e-07,
+      "loss": 1.014,
+      "step": 19144
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.01411405080581e-07,
+      "loss": 0.9031,
+      "step": 19145
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.010335198423125e-07,
+      "loss": 0.8968,
+      "step": 19146
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.006557159580686e-07,
+      "loss": 1.0167,
+      "step": 19147
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.002779934351662e-07,
+      "loss": 0.886,
+      "step": 19148
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.999003522809251e-07,
+      "loss": 0.9182,
+      "step": 19149
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.995227925026627e-07,
+      "loss": 0.8967,
+      "step": 19150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.991453141076927e-07,
+      "loss": 0.843,
+      "step": 19151
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.987679171033303e-07,
+      "loss": 0.9661,
+      "step": 19152
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.98390601496885e-07,
+      "loss": 0.8615,
+      "step": 19153
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.980133672956702e-07,
+      "loss": 0.9544,
+      "step": 19154
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.976362145069922e-07,
+      "loss": 0.865,
+      "step": 19155
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.972591431381605e-07,
+      "loss": 0.8699,
+      "step": 19156
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.968821531964782e-07,
+      "loss": 0.9516,
+      "step": 19157
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.965052446892519e-07,
+      "loss": 0.9279,
+      "step": 19158
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.961284176237821e-07,
+      "loss": 0.801,
+      "step": 19159
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.957516720073716e-07,
+      "loss": 1.0633,
+      "step": 19160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.953750078473161e-07,
+      "loss": 1.1233,
+      "step": 19161
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.949984251509185e-07,
+      "loss": 0.8924,
+      "step": 19162
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.946219239254716e-07,
+      "loss": 0.8728,
+      "step": 19163
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.942455041782721e-07,
+      "loss": 0.8483,
+      "step": 19164
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.938691659166109e-07,
+      "loss": 1.0631,
+      "step": 19165
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.934929091477816e-07,
+      "loss": 0.9536,
+      "step": 19166
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.931167338790719e-07,
+      "loss": 1.2777,
+      "step": 19167
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.927406401177723e-07,
+      "loss": 0.9507,
+      "step": 19168
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.923646278711672e-07,
+      "loss": 0.9563,
+      "step": 19169
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.91988697146544e-07,
+      "loss": 1.0273,
+      "step": 19170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.916128479511842e-07,
+      "loss": 0.9781,
+      "step": 19171
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.912370802923719e-07,
+      "loss": 1.0024,
+      "step": 19172
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.908613941773846e-07,
+      "loss": 0.9001,
+      "step": 19173
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.904857896135049e-07,
+      "loss": 0.8247,
+      "step": 19174
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.901102666080069e-07,
+      "loss": 0.8992,
+      "step": 19175
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.897348251681691e-07,
+      "loss": 0.8422,
+      "step": 19176
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.893594653012627e-07,
+      "loss": 0.9322,
+      "step": 19177
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.889841870145626e-07,
+      "loss": 1.007,
+      "step": 19178
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.886089903153371e-07,
+      "loss": 0.8024,
+      "step": 19179
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.882338752108587e-07,
+      "loss": 0.9895,
+      "step": 19180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.878588417083927e-07,
+      "loss": 0.8346,
+      "step": 19181
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.87483889815207e-07,
+      "loss": 0.9787,
+      "step": 19182
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.871090195385645e-07,
+      "loss": 0.8267,
+      "step": 19183
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.867342308857284e-07,
+      "loss": 0.9528,
+      "step": 19184
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.863595238639626e-07,
+      "loss": 0.9164,
+      "step": 19185
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.85984898480524e-07,
+      "loss": 0.9199,
+      "step": 19186
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.85610354742673e-07,
+      "loss": 0.8855,
+      "step": 19187
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.852358926576647e-07,
+      "loss": 1.0288,
+      "step": 19188
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.848615122327558e-07,
+      "loss": 0.9912,
+      "step": 19189
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.844872134751974e-07,
+      "loss": 1.0021,
+      "step": 19190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.841129963922445e-07,
+      "loss": 0.8535,
+      "step": 19191
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.837388609911451e-07,
+      "loss": 0.9324,
+      "step": 19192
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.8336480727915e-07,
+      "loss": 1.0411,
+      "step": 19193
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.829908352635035e-07,
+      "loss": 0.9455,
+      "step": 19194
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.826169449514548e-07,
+      "loss": 0.9476,
+      "step": 19195
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.82243136350243e-07,
+      "loss": 0.7818,
+      "step": 19196
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.818694094671169e-07,
+      "loss": 1.0327,
+      "step": 19197
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.814957643093124e-07,
+      "loss": 0.8891,
+      "step": 19198
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.811222008840719e-07,
+      "loss": 1.0227,
+      "step": 19199
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.807487191986307e-07,
+      "loss": 0.8817,
+      "step": 19200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.803753192602271e-07,
+      "loss": 0.9362,
+      "step": 19201
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.800020010760933e-07,
+      "loss": 0.9489,
+      "step": 19202
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.79628764653465e-07,
+      "loss": 1.0502,
+      "step": 19203
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.792556099995701e-07,
+      "loss": 0.9897,
+      "step": 19204
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.788825371216424e-07,
+      "loss": 0.9614,
+      "step": 19205
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.785095460269059e-07,
+      "loss": 0.868,
+      "step": 19206
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.781366367225912e-07,
+      "loss": 0.9601,
+      "step": 19207
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.777638092159201e-07,
+      "loss": 0.8402,
+      "step": 19208
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.773910635141174e-07,
+      "loss": 0.9379,
+      "step": 19209
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.770183996244063e-07,
+      "loss": 1.0082,
+      "step": 19210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.766458175540048e-07,
+      "loss": 0.8804,
+      "step": 19211
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.762733173101333e-07,
+      "loss": 0.9208,
+      "step": 19212
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.759008989000072e-07,
+      "loss": 0.8087,
+      "step": 19213
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.755285623308445e-07,
+      "loss": 0.7818,
+      "step": 19214
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.751563076098556e-07,
+      "loss": 0.8366,
+      "step": 19215
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.74784134744257e-07,
+      "loss": 0.9745,
+      "step": 19216
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.744120437412556e-07,
+      "loss": 0.9722,
+      "step": 19217
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.740400346080634e-07,
+      "loss": 0.8482,
+      "step": 19218
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.736681073518854e-07,
+      "loss": 0.8562,
+      "step": 19219
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.732962619799306e-07,
+      "loss": 0.9953,
+      "step": 19220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.729244984994011e-07,
+      "loss": 0.7856,
+      "step": 19221
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.725528169175017e-07,
+      "loss": 1.0355,
+      "step": 19222
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.721812172414316e-07,
+      "loss": 0.8675,
+      "step": 19223
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.718096994783919e-07,
+      "loss": 0.8551,
+      "step": 19224
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.7143826363558e-07,
+      "loss": 1.0576,
+      "step": 19225
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.710669097201928e-07,
+      "loss": 0.907,
+      "step": 19226
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.706956377394242e-07,
+      "loss": 0.8658,
+      "step": 19227
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.703244477004696e-07,
+      "loss": 0.9241,
+      "step": 19228
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.699533396105174e-07,
+      "loss": 1.2256,
+      "step": 19229
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.695823134767616e-07,
+      "loss": 0.9097,
+      "step": 19230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.692113693063862e-07,
+      "loss": 0.8199,
+      "step": 19231
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.688405071065836e-07,
+      "loss": 1.0441,
+      "step": 19232
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.684697268845348e-07,
+      "loss": 0.8428,
+      "step": 19233
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.680990286474265e-07,
+      "loss": 0.7689,
+      "step": 19234
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.677284124024387e-07,
+      "loss": 0.9619,
+      "step": 19235
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.673578781567537e-07,
+      "loss": 0.7206,
+      "step": 19236
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.669874259175481e-07,
+      "loss": 1.0235,
+      "step": 19237
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.666170556920028e-07,
+      "loss": 0.9179,
+      "step": 19238
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.662467674872903e-07,
+      "loss": 0.8866,
+      "step": 19239
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.658765613105873e-07,
+      "loss": 0.9551,
+      "step": 19240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.655064371690646e-07,
+      "loss": 0.7548,
+      "step": 19241
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.651363950698936e-07,
+      "loss": 0.9909,
+      "step": 19242
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.647664350202461e-07,
+      "loss": 0.9829,
+      "step": 19243
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.643965570272866e-07,
+      "loss": 0.7789,
+      "step": 19244
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.640267610981845e-07,
+      "loss": 0.7899,
+      "step": 19245
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.636570472401017e-07,
+      "loss": 0.9768,
+      "step": 19246
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.632874154602038e-07,
+      "loss": 1.084,
+      "step": 19247
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.629178657656505e-07,
+      "loss": 0.9615,
+      "step": 19248
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.625483981636028e-07,
+      "loss": 0.9454,
+      "step": 19249
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.621790126612182e-07,
+      "loss": 1.0073,
+      "step": 19250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.618097092656545e-07,
+      "loss": 0.9295,
+      "step": 19251
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.614404879840653e-07,
+      "loss": 0.8961,
+      "step": 19252
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.610713488236066e-07,
+      "loss": 0.919,
+      "step": 19253
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.607022917914264e-07,
+      "loss": 0.9836,
+      "step": 19254
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.603333168946802e-07,
+      "loss": 1.0715,
+      "step": 19255
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.59964424140513e-07,
+      "loss": 0.9109,
+      "step": 19256
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.595956135360744e-07,
+      "loss": 0.8629,
+      "step": 19257
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.592268850885076e-07,
+      "loss": 1.0432,
+      "step": 19258
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.588582388049592e-07,
+      "loss": 0.8478,
+      "step": 19259
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.584896746925697e-07,
+      "loss": 1.007,
+      "step": 19260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.581211927584814e-07,
+      "loss": 1.0591,
+      "step": 19261
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.57752793009831e-07,
+      "loss": 0.8825,
+      "step": 19262
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.573844754537602e-07,
+      "loss": 0.9843,
+      "step": 19263
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.570162400974007e-07,
+      "loss": 1.0101,
+      "step": 19264
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.566480869478909e-07,
+      "loss": 1.1667,
+      "step": 19265
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.562800160123601e-07,
+      "loss": 0.9311,
+      "step": 19266
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.55912027297942e-07,
+      "loss": 0.885,
+      "step": 19267
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.555441208117653e-07,
+      "loss": 0.8262,
+      "step": 19268
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.551762965609594e-07,
+      "loss": 0.9605,
+      "step": 19269
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.54808554552649e-07,
+      "loss": 0.7949,
+      "step": 19270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.544408947939613e-07,
+      "loss": 0.9294,
+      "step": 19271
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.540733172920167e-07,
+      "loss": 0.8888,
+      "step": 19272
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.537058220539395e-07,
+      "loss": 1.0176,
+      "step": 19273
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.533384090868478e-07,
+      "loss": 1.0783,
+      "step": 19274
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.529710783978622e-07,
+      "loss": 0.8595,
+      "step": 19275
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.526038299940974e-07,
+      "loss": 0.9211,
+      "step": 19276
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.5223666388267e-07,
+      "loss": 0.9824,
+      "step": 19277
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.518695800706949e-07,
+      "loss": 0.8985,
+      "step": 19278
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.515025785652819e-07,
+      "loss": 0.9989,
+      "step": 19279
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.511356593735436e-07,
+      "loss": 0.8662,
+      "step": 19280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.507688225025872e-07,
+      "loss": 0.8658,
+      "step": 19281
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.504020679595226e-07,
+      "loss": 0.9869,
+      "step": 19282
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.500353957514517e-07,
+      "loss": 1.0446,
+      "step": 19283
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.496688058854829e-07,
+      "loss": 0.8411,
+      "step": 19284
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.493022983687154e-07,
+      "loss": 0.9311,
+      "step": 19285
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.48935873208253e-07,
+      "loss": 1.0312,
+      "step": 19286
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.485695304111923e-07,
+      "loss": 0.7098,
+      "step": 19287
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.482032699846342e-07,
+      "loss": 0.8405,
+      "step": 19288
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.478370919356703e-07,
+      "loss": 0.9113,
+      "step": 19289
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.474709962714011e-07,
+      "loss": 1.1052,
+      "step": 19290
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.471049829989153e-07,
+      "loss": 0.9922,
+      "step": 19291
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.467390521253071e-07,
+      "loss": 0.9095,
+      "step": 19292
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.463732036576638e-07,
+      "loss": 0.9352,
+      "step": 19293
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.460074376030757e-07,
+      "loss": 1.0062,
+      "step": 19294
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.456417539686272e-07,
+      "loss": 0.9991,
+      "step": 19295
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.45276152761406e-07,
+      "loss": 0.9478,
+      "step": 19296
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.449106339884932e-07,
+      "loss": 0.9366,
+      "step": 19297
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.445451976569728e-07,
+      "loss": 1.0497,
+      "step": 19298
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.441798437739228e-07,
+      "loss": 0.9984,
+      "step": 19299
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.438145723464241e-07,
+      "loss": 0.9409,
+      "step": 19300
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.434493833815509e-07,
+      "loss": 0.9663,
+      "step": 19301
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.430842768863811e-07,
+      "loss": 0.925,
+      "step": 19302
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.427192528679883e-07,
+      "loss": 0.9607,
+      "step": 19303
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.423543113334436e-07,
+      "loss": 1.0643,
+      "step": 19304
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.419894522898191e-07,
+      "loss": 0.6851,
+      "step": 19305
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.416246757441825e-07,
+      "loss": 0.897,
+      "step": 19306
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.412599817036026e-07,
+      "loss": 0.7951,
+      "step": 19307
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.408953701751431e-07,
+      "loss": 0.9312,
+      "step": 19308
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.405308411658713e-07,
+      "loss": 0.9538,
+      "step": 19309
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.401663946828469e-07,
+      "loss": 0.8538,
+      "step": 19310
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.398020307331339e-07,
+      "loss": 0.9706,
+      "step": 19311
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.39437749323787e-07,
+      "loss": 1.0441,
+      "step": 19312
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.390735504618702e-07,
+      "loss": 0.9516,
+      "step": 19313
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.387094341544354e-07,
+      "loss": 0.978,
+      "step": 19314
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.3834540040854e-07,
+      "loss": 0.935,
+      "step": 19315
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.379814492312343e-07,
+      "loss": 0.8464,
+      "step": 19316
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.376175806295727e-07,
+      "loss": 0.8522,
+      "step": 19317
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.372537946106023e-07,
+      "loss": 0.921,
+      "step": 19318
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.368900911813743e-07,
+      "loss": 0.9036,
+      "step": 19319
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.365264703489322e-07,
+      "loss": 1.0244,
+      "step": 19320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.361629321203234e-07,
+      "loss": 1.0,
+      "step": 19321
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.3579947650259e-07,
+      "loss": 0.8869,
+      "step": 19322
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.354361035027751e-07,
+      "loss": 0.8889,
+      "step": 19323
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.350728131279167e-07,
+      "loss": 0.9042,
+      "step": 19324
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.347096053850572e-07,
+      "loss": 0.89,
+      "step": 19325
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.3434648028123e-07,
+      "loss": 0.6536,
+      "step": 19326
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.339834378234734e-07,
+      "loss": 0.9394,
+      "step": 19327
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.336204780188194e-07,
+      "loss": 1.0707,
+      "step": 19328
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.332576008743014e-07,
+      "loss": 0.8865,
+      "step": 19329
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.328948063969488e-07,
+      "loss": 1.0114,
+      "step": 19330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.325320945937919e-07,
+      "loss": 0.8597,
+      "step": 19331
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.32169465471857e-07,
+      "loss": 0.9719,
+      "step": 19332
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.318069190381716e-07,
+      "loss": 1.0209,
+      "step": 19333
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.31444455299758e-07,
+      "loss": 0.8671,
+      "step": 19334
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.310820742636398e-07,
+      "loss": 1.1442,
+      "step": 19335
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.30719775936839e-07,
+      "loss": 1.1729,
+      "step": 19336
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.303575603263729e-07,
+      "loss": 0.9349,
+      "step": 19337
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.299954274392618e-07,
+      "loss": 0.9617,
+      "step": 19338
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.296333772825198e-07,
+      "loss": 0.9085,
+      "step": 19339
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.292714098631631e-07,
+      "loss": 0.858,
+      "step": 19340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.289095251882034e-07,
+      "loss": 0.9456,
+      "step": 19341
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.285477232646537e-07,
+      "loss": 0.951,
+      "step": 19342
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.28186004099522e-07,
+      "loss": 1.0605,
+      "step": 19343
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.278243676998181e-07,
+      "loss": 0.9711,
+      "step": 19344
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.274628140725475e-07,
+      "loss": 0.9335,
+      "step": 19345
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.271013432247165e-07,
+      "loss": 0.8892,
+      "step": 19346
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.267399551633253e-07,
+      "loss": 0.8514,
+      "step": 19347
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.263786498953806e-07,
+      "loss": 1.0197,
+      "step": 19348
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.260174274278791e-07,
+      "loss": 0.8932,
+      "step": 19349
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.256562877678214e-07,
+      "loss": 0.8466,
+      "step": 19350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.252952309222022e-07,
+      "loss": 0.8191,
+      "step": 19351
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.2493425689802e-07,
+      "loss": 1.1256,
+      "step": 19352
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.245733657022652e-07,
+      "loss": 0.9444,
+      "step": 19353
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.242125573419328e-07,
+      "loss": 0.9188,
+      "step": 19354
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.238518318240112e-07,
+      "loss": 0.9611,
+      "step": 19355
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.234911891554914e-07,
+      "loss": 0.9723,
+      "step": 19356
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.231306293433587e-07,
+      "loss": 1.02,
+      "step": 19357
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.227701523946007e-07,
+      "loss": 0.8496,
+      "step": 19358
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.224097583162005e-07,
+      "loss": 1.1704,
+      "step": 19359
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.220494471151401e-07,
+      "loss": 0.8245,
+      "step": 19360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.216892187984026e-07,
+      "loss": 0.9488,
+      "step": 19361
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.213290733729645e-07,
+      "loss": 0.7809,
+      "step": 19362
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.209690108458067e-07,
+      "loss": 0.8296,
+      "step": 19363
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.206090312239022e-07,
+      "loss": 0.8639,
+      "step": 19364
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.202491345142288e-07,
+      "loss": 1.0464,
+      "step": 19365
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.198893207237557e-07,
+      "loss": 1.0665,
+      "step": 19366
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.195295898594562e-07,
+      "loss": 0.9324,
+      "step": 19367
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.191699419283016e-07,
+      "loss": 0.9267,
+      "step": 19368
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.188103769372567e-07,
+      "loss": 0.8874,
+      "step": 19369
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.184508948932895e-07,
+      "loss": 0.8252,
+      "step": 19370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.180914958033663e-07,
+      "loss": 1.0919,
+      "step": 19371
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.177321796744479e-07,
+      "loss": 0.9737,
+      "step": 19372
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.173729465134987e-07,
+      "loss": 1.0482,
+      "step": 19373
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.170137963274759e-07,
+      "loss": 0.7548,
+      "step": 19374
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.1665472912334e-07,
+      "loss": 0.9424,
+      "step": 19375
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.162957449080466e-07,
+      "loss": 0.9008,
+      "step": 19376
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.159368436885516e-07,
+      "loss": 0.9988,
+      "step": 19377
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.155780254718081e-07,
+      "loss": 1.0642,
+      "step": 19378
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.152192902647692e-07,
+      "loss": 0.7998,
+      "step": 19379
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.148606380743839e-07,
+      "loss": 0.8045,
+      "step": 19380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.145020689076027e-07,
+      "loss": 0.7341,
+      "step": 19381
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.141435827713694e-07,
+      "loss": 0.8825,
+      "step": 19382
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.137851796726341e-07,
+      "loss": 1.0176,
+      "step": 19383
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.13426859618338e-07,
+      "loss": 1.0155,
+      "step": 19384
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.130686226154249e-07,
+      "loss": 0.8682,
+      "step": 19385
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.127104686708341e-07,
+      "loss": 1.0291,
+      "step": 19386
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.123523977915059e-07,
+      "loss": 0.8424,
+      "step": 19387
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.119944099843768e-07,
+      "loss": 0.97,
+      "step": 19388
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.116365052563846e-07,
+      "loss": 0.8177,
+      "step": 19389
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.112786836144608e-07,
+      "loss": 0.8315,
+      "step": 19390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.109209450655408e-07,
+      "loss": 0.9202,
+      "step": 19391
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.105632896165532e-07,
+      "loss": 0.9207,
+      "step": 19392
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.102057172744303e-07,
+      "loss": 0.9311,
+      "step": 19393
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.098482280460978e-07,
+      "loss": 0.9782,
+      "step": 19394
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.094908219384822e-07,
+      "loss": 0.8988,
+      "step": 19395
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.091334989585097e-07,
+      "loss": 0.7523,
+      "step": 19396
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.087762591131014e-07,
+      "loss": 1.0483,
+      "step": 19397
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.084191024091813e-07,
+      "loss": 0.7519,
+      "step": 19398
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.080620288536655e-07,
+      "loss": 0.8805,
+      "step": 19399
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.077050384534761e-07,
+      "loss": 0.9384,
+      "step": 19400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.073481312155267e-07,
+      "loss": 0.9239,
+      "step": 19401
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.069913071467344e-07,
+      "loss": 1.0964,
+      "step": 19402
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.066345662540108e-07,
+      "loss": 0.9798,
+      "step": 19403
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.0627790854427e-07,
+      "loss": 0.8764,
+      "step": 19404
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.059213340244181e-07,
+      "loss": 1.0381,
+      "step": 19405
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.055648427013684e-07,
+      "loss": 1.0509,
+      "step": 19406
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.052084345820249e-07,
+      "loss": 0.9728,
+      "step": 19407
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.048521096732947e-07,
+      "loss": 0.7552,
+      "step": 19408
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.044958679820796e-07,
+      "loss": 0.8057,
+      "step": 19409
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.041397095152841e-07,
+      "loss": 0.8267,
+      "step": 19410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.037836342798054e-07,
+      "loss": 0.8569,
+      "step": 19411
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.034276422825459e-07,
+      "loss": 0.9973,
+      "step": 19412
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.030717335303999e-07,
+      "loss": 0.9499,
+      "step": 19413
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.027159080302653e-07,
+      "loss": 1.0267,
+      "step": 19414
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.023601657890344e-07,
+      "loss": 0.8166,
+      "step": 19415
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.020045068136017e-07,
+      "loss": 1.1507,
+      "step": 19416
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.016489311108543e-07,
+      "loss": 0.9758,
+      "step": 19417
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.012934386876857e-07,
+      "loss": 0.9492,
+      "step": 19418
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.00938029550981e-07,
+      "loss": 0.8678,
+      "step": 19419
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.005827037076274e-07,
+      "loss": 0.9894,
+      "step": 19420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.002274611645083e-07,
+      "loss": 0.9585,
+      "step": 19421
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.99872301928507e-07,
+      "loss": 0.8793,
+      "step": 19422
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.995172260065036e-07,
+      "loss": 0.8515,
+      "step": 19423
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.991622334053799e-07,
+      "loss": 0.9522,
+      "step": 19424
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.988073241320109e-07,
+      "loss": 0.9311,
+      "step": 19425
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.984524981932756e-07,
+      "loss": 0.9435,
+      "step": 19426
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.980977555960461e-07,
+      "loss": 0.8716,
+      "step": 19427
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.977430963471971e-07,
+      "loss": 0.7484,
+      "step": 19428
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.973885204536007e-07,
+      "loss": 0.8812,
+      "step": 19429
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.970340279221244e-07,
+      "loss": 0.9412,
+      "step": 19430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.966796187596386e-07,
+      "loss": 1.1569,
+      "step": 19431
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.963252929730085e-07,
+      "loss": 0.9845,
+      "step": 19432
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.959710505691003e-07,
+      "loss": 0.9171,
+      "step": 19433
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.956168915547756e-07,
+      "loss": 0.9421,
+      "step": 19434
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.952628159368985e-07,
+      "loss": 0.9534,
+      "step": 19435
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.949088237223268e-07,
+      "loss": 0.8575,
+      "step": 19436
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.945549149179215e-07,
+      "loss": 0.8046,
+      "step": 19437
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.942010895305362e-07,
+      "loss": 1.1834,
+      "step": 19438
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.938473475670298e-07,
+      "loss": 0.9944,
+      "step": 19439
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.934936890342515e-07,
+      "loss": 0.9478,
+      "step": 19440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.931401139390586e-07,
+      "loss": 1.036,
+      "step": 19441
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.927866222882984e-07,
+      "loss": 1.0295,
+      "step": 19442
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.924332140888206e-07,
+      "loss": 1.0282,
+      "step": 19443
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.920798893474718e-07,
+      "loss": 0.9027,
+      "step": 19444
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.917266480710982e-07,
+      "loss": 1.0163,
+      "step": 19445
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.913734902665431e-07,
+      "loss": 0.8292,
+      "step": 19446
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.910204159406503e-07,
+      "loss": 0.9012,
+      "step": 19447
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.906674251002582e-07,
+      "loss": 1.0473,
+      "step": 19448
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.903145177522086e-07,
+      "loss": 0.9858,
+      "step": 19449
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.89961693903336e-07,
+      "loss": 0.8735,
+      "step": 19450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.896089535604794e-07,
+      "loss": 0.9586,
+      "step": 19451
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.892562967304706e-07,
+      "loss": 0.9009,
+      "step": 19452
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.889037234201429e-07,
+      "loss": 0.9212,
+      "step": 19453
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.885512336363287e-07,
+      "loss": 0.9475,
+      "step": 19454
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.881988273858553e-07,
+      "loss": 0.9268,
+      "step": 19455
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.878465046755529e-07,
+      "loss": 1.1455,
+      "step": 19456
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.874942655122452e-07,
+      "loss": 0.8468,
+      "step": 19457
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.871421099027586e-07,
+      "loss": 0.9606,
+      "step": 19458
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.867900378539145e-07,
+      "loss": 0.9913,
+      "step": 19459
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.864380493725359e-07,
+      "loss": 0.8984,
+      "step": 19460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.860861444654404e-07,
+      "loss": 0.9096,
+      "step": 19461
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.857343231394487e-07,
+      "loss": 1.0573,
+      "step": 19462
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.853825854013735e-07,
+      "loss": 0.9197,
+      "step": 19463
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.850309312580345e-07,
+      "loss": 0.9772,
+      "step": 19464
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.846793607162411e-07,
+      "loss": 0.8209,
+      "step": 19465
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.843278737828074e-07,
+      "loss": 1.0833,
+      "step": 19466
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.839764704645407e-07,
+      "loss": 0.9927,
+      "step": 19467
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.836251507682523e-07,
+      "loss": 0.8746,
+      "step": 19468
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.832739147007461e-07,
+      "loss": 0.8347,
+      "step": 19469
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.8292276226883e-07,
+      "loss": 1.0342,
+      "step": 19470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.825716934793048e-07,
+      "loss": 0.9667,
+      "step": 19471
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.822207083389748e-07,
+      "loss": 0.8842,
+      "step": 19472
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.818698068546381e-07,
+      "loss": 1.0405,
+      "step": 19473
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.815189890330959e-07,
+      "loss": 0.9309,
+      "step": 19474
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.81168254881141e-07,
+      "loss": 0.9416,
+      "step": 19475
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.808176044055737e-07,
+      "loss": 1.0494,
+      "step": 19476
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.804670376131845e-07,
+      "loss": 0.9476,
+      "step": 19477
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.801165545107679e-07,
+      "loss": 1.0172,
+      "step": 19478
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.797661551051116e-07,
+      "loss": 0.9936,
+      "step": 19479
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.794158394030076e-07,
+      "loss": 0.7913,
+      "step": 19480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.790656074112406e-07,
+      "loss": 0.8053,
+      "step": 19481
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.78715459136598e-07,
+      "loss": 1.0158,
+      "step": 19482
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.783653945858626e-07,
+      "loss": 0.9554,
+      "step": 19483
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.78015413765818e-07,
+      "loss": 0.9967,
+      "step": 19484
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.77665516683243e-07,
+      "loss": 1.0438,
+      "step": 19485
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.773157033449196e-07,
+      "loss": 0.9416,
+      "step": 19486
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.769659737576229e-07,
+      "loss": 0.978,
+      "step": 19487
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.766163279281302e-07,
+      "loss": 1.0493,
+      "step": 19488
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.762667658632161e-07,
+      "loss": 1.097,
+      "step": 19489
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.759172875696512e-07,
+      "loss": 0.9958,
+      "step": 19490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.755678930542092e-07,
+      "loss": 1.0297,
+      "step": 19491
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.752185823236579e-07,
+      "loss": 0.8005,
+      "step": 19492
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.748693553847658e-07,
+      "loss": 0.7499,
+      "step": 19493
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.745202122442984e-07,
+      "loss": 0.9426,
+      "step": 19494
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.741711529090212e-07,
+      "loss": 0.7288,
+      "step": 19495
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.738221773856962e-07,
+      "loss": 0.9641,
+      "step": 19496
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.734732856810855e-07,
+      "loss": 0.9279,
+      "step": 19497
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.731244778019469e-07,
+      "loss": 0.7534,
+      "step": 19498
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.727757537550423e-07,
+      "loss": 0.9921,
+      "step": 19499
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.724271135471244e-07,
+      "loss": 0.9269,
+      "step": 19500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.720785571849508e-07,
+      "loss": 1.1001,
+      "step": 19501
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.717300846752723e-07,
+      "loss": 0.8501,
+      "step": 19502
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.713816960248426e-07,
+      "loss": 0.9051,
+      "step": 19503
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.710333912404094e-07,
+      "loss": 0.9478,
+      "step": 19504
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.706851703287237e-07,
+      "loss": 0.9016,
+      "step": 19505
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.703370332965298e-07,
+      "loss": 0.9449,
+      "step": 19506
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.699889801505744e-07,
+      "loss": 0.9575,
+      "step": 19507
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.696410108975992e-07,
+      "loss": 0.8995,
+      "step": 19508
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.692931255443485e-07,
+      "loss": 0.9743,
+      "step": 19509
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.689453240975591e-07,
+      "loss": 0.8357,
+      "step": 19510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.685976065639732e-07,
+      "loss": 0.8787,
+      "step": 19511
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.682499729503256e-07,
+      "loss": 1.0772,
+      "step": 19512
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.67902423263353e-07,
+      "loss": 0.951,
+      "step": 19513
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.675549575097873e-07,
+      "loss": 0.9292,
+      "step": 19514
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.672075756963625e-07,
+      "loss": 0.884,
+      "step": 19515
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.668602778298072e-07,
+      "loss": 1.0818,
+      "step": 19516
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.665130639168521e-07,
+      "loss": 0.9946,
+      "step": 19517
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.661659339642223e-07,
+      "loss": 1.1049,
+      "step": 19518
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.658188879786454e-07,
+      "loss": 0.8657,
+      "step": 19519
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.654719259668435e-07,
+      "loss": 0.8612,
+      "step": 19520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.651250479355397e-07,
+      "loss": 1.0376,
+      "step": 19521
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.647782538914554e-07,
+      "loss": 0.8798,
+      "step": 19522
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.644315438413085e-07,
+      "loss": 1.0388,
+      "step": 19523
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.64084917791818e-07,
+      "loss": 0.9763,
+      "step": 19524
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.637383757496979e-07,
+      "loss": 0.7577,
+      "step": 19525
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.633919177216636e-07,
+      "loss": 0.9211,
+      "step": 19526
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.630455437144262e-07,
+      "loss": 1.0906,
+      "step": 19527
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.626992537346988e-07,
+      "loss": 0.9671,
+      "step": 19528
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.623530477891888e-07,
+      "loss": 0.8362,
+      "step": 19529
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.620069258846051e-07,
+      "loss": 0.8973,
+      "step": 19530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.616608880276526e-07,
+      "loss": 0.8589,
+      "step": 19531
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.61314934225037e-07,
+      "loss": 0.867,
+      "step": 19532
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.609690644834583e-07,
+      "loss": 0.9858,
+      "step": 19533
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.606232788096218e-07,
+      "loss": 0.8335,
+      "step": 19534
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.602775772102238e-07,
+      "loss": 0.9254,
+      "step": 19535
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.599319596919645e-07,
+      "loss": 0.9313,
+      "step": 19536
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.595864262615381e-07,
+      "loss": 0.839,
+      "step": 19537
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.592409769256408e-07,
+      "loss": 0.8458,
+      "step": 19538
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.588956116909645e-07,
+      "loss": 0.9082,
+      "step": 19539
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.585503305642016e-07,
+      "loss": 1.0079,
+      "step": 19540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.582051335520401e-07,
+      "loss": 1.0276,
+      "step": 19541
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.578600206611707e-07,
+      "loss": 0.8291,
+      "step": 19542
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.575149918982771e-07,
+      "loss": 1.0005,
+      "step": 19543
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.571700472700465e-07,
+      "loss": 0.9734,
+      "step": 19544
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.568251867831599e-07,
+      "loss": 0.784,
+      "step": 19545
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.564804104443006e-07,
+      "loss": 1.0205,
+      "step": 19546
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.561357182601485e-07,
+      "loss": 1.0378,
+      "step": 19547
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.55791110237381e-07,
+      "loss": 1.0541,
+      "step": 19548
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.554465863826765e-07,
+      "loss": 0.8323,
+      "step": 19549
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.551021467027075e-07,
+      "loss": 0.6542,
+      "step": 19550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.547577912041497e-07,
+      "loss": 0.9604,
+      "step": 19551
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.54413519893673e-07,
+      "loss": 0.8976,
+      "step": 19552
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.540693327779501e-07,
+      "loss": 0.7719,
+      "step": 19553
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.537252298636465e-07,
+      "loss": 0.9157,
+      "step": 19554
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.533812111574323e-07,
+      "loss": 0.9338,
+      "step": 19555
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.530372766659682e-07,
+      "loss": 0.9231,
+      "step": 19556
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.526934263959234e-07,
+      "loss": 1.0153,
+      "step": 19557
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.523496603539565e-07,
+      "loss": 0.751,
+      "step": 19558
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.520059785467297e-07,
+      "loss": 0.9468,
+      "step": 19559
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.516623809808998e-07,
+      "loss": 0.8127,
+      "step": 19560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.513188676631254e-07,
+      "loss": 0.7999,
+      "step": 19561
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.509754386000611e-07,
+      "loss": 0.9046,
+      "step": 19562
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.506320937983618e-07,
+      "loss": 0.7734,
+      "step": 19563
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.502888332646779e-07,
+      "loss": 1.0545,
+      "step": 19564
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.499456570056622e-07,
+      "loss": 0.8069,
+      "step": 19565
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.496025650279619e-07,
+      "loss": 0.7591,
+      "step": 19566
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.492595573382265e-07,
+      "loss": 0.8602,
+      "step": 19567
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.489166339430974e-07,
+      "loss": 1.0567,
+      "step": 19568
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.485737948492237e-07,
+      "loss": 0.9054,
+      "step": 19569
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.482310400632452e-07,
+      "loss": 0.9988,
+      "step": 19570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.478883695918036e-07,
+      "loss": 0.9735,
+      "step": 19571
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.475457834415372e-07,
+      "loss": 0.8815,
+      "step": 19572
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.472032816190843e-07,
+      "loss": 0.9497,
+      "step": 19573
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.468608641310797e-07,
+      "loss": 0.8851,
+      "step": 19574
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.465185309841598e-07,
+      "loss": 1.0142,
+      "step": 19575
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.461762821849543e-07,
+      "loss": 0.8787,
+      "step": 19576
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.458341177400967e-07,
+      "loss": 0.7974,
+      "step": 19577
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.454920376562146e-07,
+      "loss": 0.9862,
+      "step": 19578
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.451500419399376e-07,
+      "loss": 0.9553,
+      "step": 19579
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.448081305978892e-07,
+      "loss": 1.0891,
+      "step": 19580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.44466303636696e-07,
+      "loss": 1.0314,
+      "step": 19581
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.44124561062981e-07,
+      "loss": 0.9251,
+      "step": 19582
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.43782902883363e-07,
+      "loss": 0.8228,
+      "step": 19583
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.434413291044644e-07,
+      "loss": 0.8746,
+      "step": 19584
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.430998397329014e-07,
+      "loss": 0.9236,
+      "step": 19585
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.427584347752913e-07,
+      "loss": 1.0708,
+      "step": 19586
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.424171142382468e-07,
+      "loss": 0.9574,
+      "step": 19587
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.420758781283836e-07,
+      "loss": 0.8194,
+      "step": 19588
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.417347264523105e-07,
+      "loss": 1.0121,
+      "step": 19589
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.413936592166398e-07,
+      "loss": 1.0637,
+      "step": 19590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.410526764279756e-07,
+      "loss": 0.9064,
+      "step": 19591
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.407117780929295e-07,
+      "loss": 0.7946,
+      "step": 19592
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.403709642181028e-07,
+      "loss": 1.0987,
+      "step": 19593
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.400302348101011e-07,
+      "loss": 0.9885,
+      "step": 19594
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.396895898755229e-07,
+      "loss": 0.7881,
+      "step": 19595
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.393490294209714e-07,
+      "loss": 0.7989,
+      "step": 19596
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.390085534530422e-07,
+      "loss": 0.9184,
+      "step": 19597
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.386681619783336e-07,
+      "loss": 0.9222,
+      "step": 19598
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.383278550034394e-07,
+      "loss": 1.0148,
+      "step": 19599
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.379876325349543e-07,
+      "loss": 0.9593,
+      "step": 19600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.376474945794686e-07,
+      "loss": 0.928,
+      "step": 19601
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.373074411435742e-07,
+      "loss": 0.893,
+      "step": 19602
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.369674722338564e-07,
+      "loss": 0.9666,
+      "step": 19603
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.366275878569062e-07,
+      "loss": 0.9309,
+      "step": 19604
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.362877880193058e-07,
+      "loss": 0.9872,
+      "step": 19605
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.359480727276407e-07,
+      "loss": 0.8529,
+      "step": 19606
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.356084419884905e-07,
+      "loss": 0.8731,
+      "step": 19607
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.35268895808438e-07,
+      "loss": 0.7984,
+      "step": 19608
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.349294341940593e-07,
+      "loss": 0.9241,
+      "step": 19609
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.345900571519336e-07,
+      "loss": 1.0512,
+      "step": 19610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.342507646886342e-07,
+      "loss": 0.8189,
+      "step": 19611
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.339115568107373e-07,
+      "loss": 0.899,
+      "step": 19612
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.335724335248122e-07,
+      "loss": 1.0485,
+      "step": 19613
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.332333948374308e-07,
+      "loss": 0.9887,
+      "step": 19614
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.328944407551635e-07,
+      "loss": 0.7756,
+      "step": 19615
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.325555712845743e-07,
+      "loss": 1.0361,
+      "step": 19616
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.322167864322309e-07,
+      "loss": 0.8366,
+      "step": 19617
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.318780862046958e-07,
+      "loss": 0.8063,
+      "step": 19618
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.315394706085332e-07,
+      "loss": 0.9259,
+      "step": 19619
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.312009396503011e-07,
+      "loss": 1.0081,
+      "step": 19620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.308624933365604e-07,
+      "loss": 0.8806,
+      "step": 19621
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.305241316738675e-07,
+      "loss": 0.942,
+      "step": 19622
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.301858546687789e-07,
+      "loss": 0.9513,
+      "step": 19623
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.298476623278472e-07,
+      "loss": 1.0547,
+      "step": 19624
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.295095546576269e-07,
+      "loss": 0.9857,
+      "step": 19625
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.291715316646657e-07,
+      "loss": 0.8924,
+      "step": 19626
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.288335933555161e-07,
+      "loss": 0.9836,
+      "step": 19627
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.284957397367236e-07,
+      "loss": 1.0817,
+      "step": 19628
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.281579708148355e-07,
+      "loss": 0.9766,
+      "step": 19629
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.278202865963934e-07,
+      "loss": 1.0317,
+      "step": 19630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.27482687087943e-07,
+      "loss": 1.0177,
+      "step": 19631
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.271451722960225e-07,
+      "loss": 1.071,
+      "step": 19632
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.26807742227174e-07,
+      "loss": 0.9873,
+      "step": 19633
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.264703968879315e-07,
+      "loss": 1.0562,
+      "step": 19634
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.261331362848344e-07,
+      "loss": 0.8125,
+      "step": 19635
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.257959604244146e-07,
+      "loss": 0.9167,
+      "step": 19636
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.254588693132074e-07,
+      "loss": 0.9831,
+      "step": 19637
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.251218629577404e-07,
+      "loss": 1.0109,
+      "step": 19638
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.247849413645451e-07,
+      "loss": 0.9911,
+      "step": 19639
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.244481045401507e-07,
+      "loss": 0.9737,
+      "step": 19640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.241113524910802e-07,
+      "loss": 0.85,
+      "step": 19641
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.237746852238608e-07,
+      "loss": 0.8618,
+      "step": 19642
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.234381027450132e-07,
+      "loss": 1.0917,
+      "step": 19643
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.231016050610605e-07,
+      "loss": 0.951,
+      "step": 19644
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.227651921785205e-07,
+      "loss": 0.8557,
+      "step": 19645
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.224288641039128e-07,
+      "loss": 0.9985,
+      "step": 19646
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.220926208437517e-07,
+      "loss": 0.8813,
+      "step": 19647
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.217564624045541e-07,
+      "loss": 0.9785,
+      "step": 19648
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.2142038879283e-07,
+      "loss": 0.9562,
+      "step": 19649
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.210844000150951e-07,
+      "loss": 1.0704,
+      "step": 19650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.207484960778548e-07,
+      "loss": 0.7368,
+      "step": 19651
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.204126769876207e-07,
+      "loss": 0.8139,
+      "step": 19652
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.200769427508963e-07,
+      "loss": 0.9113,
+      "step": 19653
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.197412933741881e-07,
+      "loss": 1.0297,
+      "step": 19654
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.194057288639981e-07,
+      "loss": 0.8873,
+      "step": 19655
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.190702492268297e-07,
+      "loss": 0.9696,
+      "step": 19656
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.187348544691801e-07,
+      "loss": 0.9167,
+      "step": 19657
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.183995445975499e-07,
+      "loss": 0.8844,
+      "step": 19658
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.180643196184338e-07,
+      "loss": 1.0376,
+      "step": 19659
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.177291795383283e-07,
+      "loss": 0.9625,
+      "step": 19660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.173941243637244e-07,
+      "loss": 1.0957,
+      "step": 19661
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.170591541011167e-07,
+      "loss": 0.9856,
+      "step": 19662
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.167242687569924e-07,
+      "loss": 1.0989,
+      "step": 19663
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.16389468337843e-07,
+      "loss": 1.0521,
+      "step": 19664
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.160547528501515e-07,
+      "loss": 0.8774,
+      "step": 19665
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.157201223004061e-07,
+      "loss": 0.8635,
+      "step": 19666
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.153855766950873e-07,
+      "loss": 0.805,
+      "step": 19667
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.150511160406797e-07,
+      "loss": 1.0597,
+      "step": 19668
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.147167403436616e-07,
+      "loss": 0.9059,
+      "step": 19669
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.143824496105122e-07,
+      "loss": 0.8191,
+      "step": 19670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.140482438477069e-07,
+      "loss": 0.9967,
+      "step": 19671
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.137141230617227e-07,
+      "loss": 0.977,
+      "step": 19672
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.133800872590328e-07,
+      "loss": 0.9476,
+      "step": 19673
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.130461364461076e-07,
+      "loss": 1.1784,
+      "step": 19674
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.127122706294197e-07,
+      "loss": 0.9416,
+      "step": 19675
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.123784898154356e-07,
+      "loss": 1.0048,
+      "step": 19676
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.120447940106234e-07,
+      "loss": 1.1169,
+      "step": 19677
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.117111832214473e-07,
+      "loss": 1.0337,
+      "step": 19678
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.113776574543722e-07,
+      "loss": 0.9567,
+      "step": 19679
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.110442167158592e-07,
+      "loss": 0.9395,
+      "step": 19680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.107108610123697e-07,
+      "loss": 0.9834,
+      "step": 19681
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.103775903503601e-07,
+      "loss": 0.8618,
+      "step": 19682
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.100444047362902e-07,
+      "loss": 0.925,
+      "step": 19683
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.097113041766123e-07,
+      "loss": 1.0282,
+      "step": 19684
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.093782886777839e-07,
+      "loss": 0.8853,
+      "step": 19685
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.090453582462541e-07,
+      "loss": 0.8482,
+      "step": 19686
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.087125128884758e-07,
+      "loss": 1.1307,
+      "step": 19687
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.083797526108948e-07,
+      "loss": 0.8777,
+      "step": 19688
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.080470774199615e-07,
+      "loss": 0.9403,
+      "step": 19689
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.077144873221185e-07,
+      "loss": 1.0432,
+      "step": 19690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.073819823238125e-07,
+      "loss": 0.864,
+      "step": 19691
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.070495624314826e-07,
+      "loss": 0.8533,
+      "step": 19692
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.06717227651572e-07,
+      "loss": 0.6702,
+      "step": 19693
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.063849779905178e-07,
+      "loss": 1.0581,
+      "step": 19694
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.060528134547589e-07,
+      "loss": 0.9386,
+      "step": 19695
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.057207340507282e-07,
+      "loss": 0.8236,
+      "step": 19696
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.053887397848629e-07,
+      "loss": 0.9637,
+      "step": 19697
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.050568306635935e-07,
+      "loss": 0.9176,
+      "step": 19698
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.047250066933519e-07,
+      "loss": 0.9257,
+      "step": 19699
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.043932678805653e-07,
+      "loss": 1.0079,
+      "step": 19700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.04061614231663e-07,
+      "loss": 0.9759,
+      "step": 19701
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.037300457530687e-07,
+      "loss": 0.8311,
+      "step": 19702
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.033985624512084e-07,
+      "loss": 0.932,
+      "step": 19703
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.030671643325031e-07,
+      "loss": 1.0011,
+      "step": 19704
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.027358514033749e-07,
+      "loss": 0.9476,
+      "step": 19705
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.024046236702407e-07,
+      "loss": 1.2926,
+      "step": 19706
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.020734811395195e-07,
+      "loss": 0.8894,
+      "step": 19707
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.017424238176283e-07,
+      "loss": 0.9659,
+      "step": 19708
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.014114517109792e-07,
+      "loss": 0.8735,
+      "step": 19709
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.010805648259859e-07,
+      "loss": 0.9627,
+      "step": 19710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.007497631690578e-07,
+      "loss": 1.0025,
+      "step": 19711
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.004190467466065e-07,
+      "loss": 0.8637,
+      "step": 19712
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.000884155650371e-07,
+      "loss": 0.8732,
+      "step": 19713
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.997578696307577e-07,
+      "loss": 0.882,
+      "step": 19714
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.9942740895017e-07,
+      "loss": 0.8139,
+      "step": 19715
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.990970335296792e-07,
+      "loss": 1.0155,
+      "step": 19716
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.987667433756844e-07,
+      "loss": 0.9836,
+      "step": 19717
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.984365384945862e-07,
+      "loss": 0.9948,
+      "step": 19718
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.981064188927799e-07,
+      "loss": 0.9875,
+      "step": 19719
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.977763845766649e-07,
+      "loss": 0.9469,
+      "step": 19720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.974464355526328e-07,
+      "loss": 1.0029,
+      "step": 19721
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.971165718270788e-07,
+      "loss": 0.8904,
+      "step": 19722
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.96786793406391e-07,
+      "loss": 0.9031,
+      "step": 19723
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.964571002969615e-07,
+      "loss": 0.7843,
+      "step": 19724
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.961274925051757e-07,
+      "loss": 0.7738,
+      "step": 19725
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.957979700374223e-07,
+      "loss": 1.014,
+      "step": 19726
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.954685329000826e-07,
+      "loss": 0.9647,
+      "step": 19727
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.951391810995422e-07,
+      "loss": 0.9633,
+      "step": 19728
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.948099146421799e-07,
+      "loss": 0.8752,
+      "step": 19729
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.944807335343772e-07,
+      "loss": 1.1201,
+      "step": 19730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.941516377825102e-07,
+      "loss": 1.0122,
+      "step": 19731
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.938226273929554e-07,
+      "loss": 1.14,
+      "step": 19732
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.934937023720888e-07,
+      "loss": 0.9215,
+      "step": 19733
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.931648627262815e-07,
+      "loss": 1.0848,
+      "step": 19734
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.928361084619061e-07,
+      "loss": 0.8838,
+      "step": 19735
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.925074395853303e-07,
+      "loss": 0.8468,
+      "step": 19736
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.92178856102924e-07,
+      "loss": 0.8762,
+      "step": 19737
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.918503580210517e-07,
+      "loss": 1.0281,
+      "step": 19738
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.915219453460796e-07,
+      "loss": 0.9457,
+      "step": 19739
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.911936180843691e-07,
+      "loss": 0.8538,
+      "step": 19740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.908653762422828e-07,
+      "loss": 0.8395,
+      "step": 19741
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.905372198261777e-07,
+      "loss": 0.9215,
+      "step": 19742
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.902091488424161e-07,
+      "loss": 0.8757,
+      "step": 19743
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.898811632973505e-07,
+      "loss": 0.8908,
+      "step": 19744
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.895532631973383e-07,
+      "loss": 0.8841,
+      "step": 19745
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.892254485487303e-07,
+      "loss": 0.8585,
+      "step": 19746
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.888977193578793e-07,
+      "loss": 1.1652,
+      "step": 19747
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.885700756311335e-07,
+      "loss": 1.0471,
+      "step": 19748
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.88242517374843e-07,
+      "loss": 0.8514,
+      "step": 19749
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.879150445953519e-07,
+      "loss": 0.8863,
+      "step": 19750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.875876572990069e-07,
+      "loss": 0.7906,
+      "step": 19751
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.872603554921491e-07,
+      "loss": 0.9074,
+      "step": 19752
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.869331391811223e-07,
+      "loss": 0.826,
+      "step": 19753
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.866060083722624e-07,
+      "loss": 0.8598,
+      "step": 19754
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.862789630719124e-07,
+      "loss": 0.9622,
+      "step": 19755
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.859520032864046e-07,
+      "loss": 0.9214,
+      "step": 19756
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.856251290220771e-07,
+      "loss": 0.9293,
+      "step": 19757
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.852983402852602e-07,
+      "loss": 0.8843,
+      "step": 19758
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.849716370822878e-07,
+      "loss": 0.8731,
+      "step": 19759
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.84645019419487e-07,
+      "loss": 0.8959,
+      "step": 19760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.843184873031888e-07,
+      "loss": 0.8768,
+      "step": 19761
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.839920407397165e-07,
+      "loss": 1.0443,
+      "step": 19762
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.836656797353985e-07,
+      "loss": 0.9343,
+      "step": 19763
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.833394042965546e-07,
+      "loss": 0.7919,
+      "step": 19764
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.830132144295081e-07,
+      "loss": 0.7691,
+      "step": 19765
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.826871101405795e-07,
+      "loss": 0.9165,
+      "step": 19766
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.823610914360855e-07,
+      "loss": 1.0184,
+      "step": 19767
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.820351583223444e-07,
+      "loss": 0.9085,
+      "step": 19768
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.817093108056682e-07,
+      "loss": 1.008,
+      "step": 19769
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.813835488923736e-07,
+      "loss": 1.0094,
+      "step": 19770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.810578725887689e-07,
+      "loss": 0.9387,
+      "step": 19771
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.807322819011668e-07,
+      "loss": 0.94,
+      "step": 19772
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.804067768358729e-07,
+      "loss": 0.9735,
+      "step": 19773
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.800813573991965e-07,
+      "loss": 0.9279,
+      "step": 19774
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.797560235974404e-07,
+      "loss": 0.8331,
+      "step": 19775
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.794307754369094e-07,
+      "loss": 0.8371,
+      "step": 19776
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.791056129239026e-07,
+      "loss": 0.8791,
+      "step": 19777
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.78780536064723e-07,
+      "loss": 0.7693,
+      "step": 19778
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.784555448656676e-07,
+      "loss": 0.9008,
+      "step": 19779
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.781306393330332e-07,
+      "loss": 0.9026,
+      "step": 19780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.77805819473114e-07,
+      "loss": 1.0245,
+      "step": 19781
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.774810852922053e-07,
+      "loss": 0.9657,
+      "step": 19782
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.77156436796596e-07,
+      "loss": 0.9132,
+      "step": 19783
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.768318739925782e-07,
+      "loss": 0.7823,
+      "step": 19784
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.765073968864388e-07,
+      "loss": 0.9811,
+      "step": 19785
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.761830054844664e-07,
+      "loss": 1.0231,
+      "step": 19786
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.758586997929433e-07,
+      "loss": 0.9457,
+      "step": 19787
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.755344798181555e-07,
+      "loss": 0.885,
+      "step": 19788
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.752103455663826e-07,
+      "loss": 0.8962,
+      "step": 19789
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.748862970439057e-07,
+      "loss": 0.9365,
+      "step": 19790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.745623342570039e-07,
+      "loss": 1.1434,
+      "step": 19791
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.742384572119519e-07,
+      "loss": 0.8111,
+      "step": 19792
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.739146659150257e-07,
+      "loss": 0.8539,
+      "step": 19793
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.735909603724998e-07,
+      "loss": 0.9833,
+      "step": 19794
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.732673405906447e-07,
+      "loss": 0.854,
+      "step": 19795
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.729438065757309e-07,
+      "loss": 0.7765,
+      "step": 19796
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.726203583340261e-07,
+      "loss": 0.9391,
+      "step": 19797
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.722969958717983e-07,
+      "loss": 1.0647,
+      "step": 19798
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.719737191953112e-07,
+      "loss": 0.9714,
+      "step": 19799
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.716505283108287e-07,
+      "loss": 1.0101,
+      "step": 19800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.713274232246135e-07,
+      "loss": 0.9247,
+      "step": 19801
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.710044039429241e-07,
+      "loss": 1.0228,
+      "step": 19802
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.706814704720203e-07,
+      "loss": 0.8125,
+      "step": 19803
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.703586228181574e-07,
+      "loss": 0.992,
+      "step": 19804
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.700358609875928e-07,
+      "loss": 1.0752,
+      "step": 19805
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.697131849865773e-07,
+      "loss": 0.852,
+      "step": 19806
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.693905948213646e-07,
+      "loss": 0.9375,
+      "step": 19807
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.690680904982032e-07,
+      "loss": 0.9178,
+      "step": 19808
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.687456720233436e-07,
+      "loss": 0.8794,
+      "step": 19809
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.684233394030309e-07,
+      "loss": 0.8263,
+      "step": 19810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.68101092643511e-07,
+      "loss": 0.797,
+      "step": 19811
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.677789317510252e-07,
+      "loss": 0.9738,
+      "step": 19812
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.674568567318195e-07,
+      "loss": 0.9551,
+      "step": 19813
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.671348675921307e-07,
+      "loss": 0.8718,
+      "step": 19814
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.668129643381992e-07,
+      "loss": 1.0622,
+      "step": 19815
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.664911469762596e-07,
+      "loss": 1.0962,
+      "step": 19816
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.661694155125497e-07,
+      "loss": 0.9305,
+      "step": 19817
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.658477699533004e-07,
+      "loss": 0.9973,
+      "step": 19818
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.65526210304746e-07,
+      "loss": 0.8754,
+      "step": 19819
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.652047365731145e-07,
+      "loss": 0.8351,
+      "step": 19820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.648833487646354e-07,
+      "loss": 0.9061,
+      "step": 19821
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.645620468855351e-07,
+      "loss": 0.9435,
+      "step": 19822
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.6424083094204e-07,
+      "loss": 0.8972,
+      "step": 19823
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.639197009403713e-07,
+      "loss": 0.9271,
+      "step": 19824
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.635986568867524e-07,
+      "loss": 0.977,
+      "step": 19825
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.632776987874039e-07,
+      "loss": 0.7818,
+      "step": 19826
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.62956826648543e-07,
+      "loss": 0.9645,
+      "step": 19827
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.626360404763875e-07,
+      "loss": 0.9369,
+      "step": 19828
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.62315340277152e-07,
+      "loss": 0.9638,
+      "step": 19829
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.619947260570502e-07,
+      "loss": 0.8118,
+      "step": 19830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.616741978222934e-07,
+      "loss": 1.0489,
+      "step": 19831
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.613537555790932e-07,
+      "loss": 0.8951,
+      "step": 19832
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.610333993336558e-07,
+      "loss": 0.9754,
+      "step": 19833
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.607131290921908e-07,
+      "loss": 0.9381,
+      "step": 19834
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.603929448608997e-07,
+      "loss": 0.9145,
+      "step": 19835
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.600728466459903e-07,
+      "loss": 1.0196,
+      "step": 19836
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.597528344536612e-07,
+      "loss": 0.8864,
+      "step": 19837
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.594329082901151e-07,
+      "loss": 1.0736,
+      "step": 19838
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.591130681615481e-07,
+      "loss": 0.8366,
+      "step": 19839
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.587933140741586e-07,
+      "loss": 0.7805,
+      "step": 19840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.584736460341406e-07,
+      "loss": 1.001,
+      "step": 19841
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.58154064047689e-07,
+      "loss": 1.0045,
+      "step": 19842
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.578345681209935e-07,
+      "loss": 0.9473,
+      "step": 19843
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.575151582602473e-07,
+      "loss": 0.8391,
+      "step": 19844
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.571958344716355e-07,
+      "loss": 0.9257,
+      "step": 19845
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.568765967613482e-07,
+      "loss": 1.0402,
+      "step": 19846
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.565574451355666e-07,
+      "loss": 1.0022,
+      "step": 19847
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.562383796004783e-07,
+      "loss": 0.7506,
+      "step": 19848
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.55919400162263e-07,
+      "loss": 0.8714,
+      "step": 19849
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.55600506827102e-07,
+      "loss": 0.8448,
+      "step": 19850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.552816996011717e-07,
+      "loss": 1.0244,
+      "step": 19851
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.549629784906518e-07,
+      "loss": 0.9865,
+      "step": 19852
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.546443435017146e-07,
+      "loss": 0.9028,
+      "step": 19853
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.543257946405362e-07,
+      "loss": 0.9499,
+      "step": 19854
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.540073319132855e-07,
+      "loss": 0.9761,
+      "step": 19855
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.536889553261355e-07,
+      "loss": 1.0927,
+      "step": 19856
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.533706648852522e-07,
+      "loss": 0.9502,
+      "step": 19857
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.53052460596804e-07,
+      "loss": 0.7957,
+      "step": 19858
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.527343424669562e-07,
+      "loss": 1.0183,
+      "step": 19859
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.524163105018709e-07,
+      "loss": 0.8981,
+      "step": 19860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.520983647077122e-07,
+      "loss": 0.9116,
+      "step": 19861
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.517805050906371e-07,
+      "loss": 1.0247,
+      "step": 19862
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.514627316568072e-07,
+      "loss": 0.8136,
+      "step": 19863
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.511450444123761e-07,
+      "loss": 0.884,
+      "step": 19864
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.508274433635019e-07,
+      "loss": 0.9036,
+      "step": 19865
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.505099285163357e-07,
+      "loss": 0.9097,
+      "step": 19866
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.501924998770314e-07,
+      "loss": 1.0747,
+      "step": 19867
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.498751574517375e-07,
+      "loss": 0.8532,
+      "step": 19868
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.495579012466034e-07,
+      "loss": 0.9555,
+      "step": 19869
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.492407312677733e-07,
+      "loss": 0.9305,
+      "step": 19870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.489236475213966e-07,
+      "loss": 0.9172,
+      "step": 19871
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.48606650013614e-07,
+      "loss": 0.9109,
+      "step": 19872
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.482897387505681e-07,
+      "loss": 0.9327,
+      "step": 19873
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.479729137383982e-07,
+      "loss": 0.9182,
+      "step": 19874
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.476561749832438e-07,
+      "loss": 0.9852,
+      "step": 19875
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.473395224912403e-07,
+      "loss": 0.9508,
+      "step": 19876
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.470229562685248e-07,
+      "loss": 1.0474,
+      "step": 19877
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.467064763212282e-07,
+      "loss": 0.9268,
+      "step": 19878
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.463900826554841e-07,
+      "loss": 0.961,
+      "step": 19879
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.460737752774214e-07,
+      "loss": 1.0219,
+      "step": 19880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.457575541931698e-07,
+      "loss": 0.8622,
+      "step": 19881
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.454414194088542e-07,
+      "loss": 0.8415,
+      "step": 19882
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.451253709306004e-07,
+      "loss": 1.0213,
+      "step": 19883
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.448094087645334e-07,
+      "loss": 0.9962,
+      "step": 19884
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.444935329167722e-07,
+      "loss": 0.7916,
+      "step": 19885
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.441777433934392e-07,
+      "loss": 1.0203,
+      "step": 19886
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.438620402006506e-07,
+      "loss": 0.9198,
+      "step": 19887
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.435464233445254e-07,
+      "loss": 0.9516,
+      "step": 19888
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.432308928311758e-07,
+      "loss": 0.778,
+      "step": 19889
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.429154486667182e-07,
+      "loss": 0.9018,
+      "step": 19890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.426000908572615e-07,
+      "loss": 1.0512,
+      "step": 19891
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.422848194089175e-07,
+      "loss": 0.9698,
+      "step": 19892
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.41969634327793e-07,
+      "loss": 0.9068,
+      "step": 19893
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.416545356199976e-07,
+      "loss": 1.1531,
+      "step": 19894
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.413395232916324e-07,
+      "loss": 0.8834,
+      "step": 19895
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.410245973488038e-07,
+      "loss": 0.7796,
+      "step": 19896
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.407097577976117e-07,
+      "loss": 0.8569,
+      "step": 19897
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.403950046441569e-07,
+      "loss": 0.9051,
+      "step": 19898
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.400803378945363e-07,
+      "loss": 1.0726,
+      "step": 19899
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.397657575548487e-07,
+      "loss": 1.0318,
+      "step": 19900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.394512636311866e-07,
+      "loss": 0.9497,
+      "step": 19901
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.391368561296456e-07,
+      "loss": 0.9784,
+      "step": 19902
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.388225350563153e-07,
+      "loss": 1.0321,
+      "step": 19903
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.38508300417287e-07,
+      "loss": 0.9678,
+      "step": 19904
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.381941522186463e-07,
+      "loss": 0.8486,
+      "step": 19905
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.378800904664844e-07,
+      "loss": 1.0562,
+      "step": 19906
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.375661151668821e-07,
+      "loss": 0.7837,
+      "step": 19907
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.37252226325925e-07,
+      "loss": 0.8507,
+      "step": 19908
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.369384239496927e-07,
+      "loss": 0.8462,
+      "step": 19909
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.366247080442671e-07,
+      "loss": 1.0272,
+      "step": 19910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.36311078615725e-07,
+      "loss": 1.0244,
+      "step": 19911
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.359975356701435e-07,
+      "loss": 1.0492,
+      "step": 19912
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.356840792135958e-07,
+      "loss": 1.0915,
+      "step": 19913
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.353707092521581e-07,
+      "loss": 0.9229,
+      "step": 19914
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.350574257918989e-07,
+      "loss": 1.0661,
+      "step": 19915
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.347442288388904e-07,
+      "loss": 0.9774,
+      "step": 19916
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.344311183991985e-07,
+      "loss": 1.0668,
+      "step": 19917
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.341180944788904e-07,
+      "loss": 0.835,
+      "step": 19918
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.338051570840325e-07,
+      "loss": 0.8347,
+      "step": 19919
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.33492306220686e-07,
+      "loss": 1.0298,
+      "step": 19920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.331795418949131e-07,
+      "loss": 0.9869,
+      "step": 19921
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.328668641127727e-07,
+      "loss": 0.9505,
+      "step": 19922
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.325542728803245e-07,
+      "loss": 0.9539,
+      "step": 19923
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.32241768203623e-07,
+      "loss": 0.9533,
+      "step": 19924
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.319293500887246e-07,
+      "loss": 0.8512,
+      "step": 19925
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.316170185416802e-07,
+      "loss": 1.0658,
+      "step": 19926
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.313047735685439e-07,
+      "loss": 0.8366,
+      "step": 19927
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.30992615175362e-07,
+      "loss": 0.8396,
+      "step": 19928
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.306805433681861e-07,
+      "loss": 1.0021,
+      "step": 19929
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.303685581530605e-07,
+      "loss": 1.0221,
+      "step": 19930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.300566595360307e-07,
+      "loss": 1.1079,
+      "step": 19931
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.297448475231382e-07,
+      "loss": 0.9089,
+      "step": 19932
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.294331221204263e-07,
+      "loss": 0.9276,
+      "step": 19933
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.291214833339325e-07,
+      "loss": 0.9049,
+      "step": 19934
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.288099311696971e-07,
+      "loss": 1.0323,
+      "step": 19935
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.284984656337533e-07,
+      "loss": 0.9548,
+      "step": 19936
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.281870867321392e-07,
+      "loss": 1.0138,
+      "step": 19937
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.278757944708846e-07,
+      "loss": 0.8723,
+      "step": 19938
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.275645888560233e-07,
+      "loss": 0.961,
+      "step": 19939
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.272534698935811e-07,
+      "loss": 0.8774,
+      "step": 19940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.269424375895904e-07,
+      "loss": 0.8511,
+      "step": 19941
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.266314919500748e-07,
+      "loss": 1.0418,
+      "step": 19942
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2632063298106e-07,
+      "loss": 1.0252,
+      "step": 19943
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.260098606885672e-07,
+      "loss": 0.895,
+      "step": 19944
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.256991750786189e-07,
+      "loss": 0.963,
+      "step": 19945
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.253885761572341e-07,
+      "loss": 0.9894,
+      "step": 19946
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.25078063930431e-07,
+      "loss": 0.7999,
+      "step": 19947
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.247676384042244e-07,
+      "loss": 0.8989,
+      "step": 19948
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.244572995846308e-07,
+      "loss": 0.8199,
+      "step": 19949
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.241470474776606e-07,
+      "loss": 0.8523,
+      "step": 19950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.238368820893263e-07,
+      "loss": 0.9063,
+      "step": 19951
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.23526803425638e-07,
+      "loss": 0.804,
+      "step": 19952
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.232168114926018e-07,
+      "loss": 0.8729,
+      "step": 19953
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.22906906296225e-07,
+      "loss": 0.7889,
+      "step": 19954
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.225970878425102e-07,
+      "loss": 0.8572,
+      "step": 19955
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.22287356137462e-07,
+      "loss": 0.8829,
+      "step": 19956
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.2197771118708e-07,
+      "loss": 0.9675,
+      "step": 19957
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.21668152997365e-07,
+      "loss": 0.9564,
+      "step": 19958
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.213586815743127e-07,
+      "loss": 0.9207,
+      "step": 19959
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.210492969239206e-07,
+      "loss": 0.9822,
+      "step": 19960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.207399990521817e-07,
+      "loss": 1.2171,
+      "step": 19961
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.204307879650905e-07,
+      "loss": 1.0268,
+      "step": 19962
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.201216636686341e-07,
+      "loss": 0.9254,
+      "step": 19963
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.198126261688063e-07,
+      "loss": 0.9724,
+      "step": 19964
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.195036754715921e-07,
+      "loss": 0.9529,
+      "step": 19965
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.191948115829786e-07,
+      "loss": 0.8663,
+      "step": 19966
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.188860345089481e-07,
+      "loss": 0.9969,
+      "step": 19967
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.185773442554853e-07,
+      "loss": 0.8242,
+      "step": 19968
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.182687408285692e-07,
+      "loss": 0.8249,
+      "step": 19969
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.179602242341802e-07,
+      "loss": 0.986,
+      "step": 19970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.176517944782939e-07,
+      "loss": 0.8902,
+      "step": 19971
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.17343451566889e-07,
+      "loss": 1.1467,
+      "step": 19972
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.170351955059367e-07,
+      "loss": 0.8687,
+      "step": 19973
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.167270263014112e-07,
+      "loss": 0.8339,
+      "step": 19974
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.16418943959282e-07,
+      "loss": 0.8699,
+      "step": 19975
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.161109484855182e-07,
+      "loss": 0.9524,
+      "step": 19976
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.158030398860891e-07,
+      "loss": 0.7596,
+      "step": 19977
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.154952181669581e-07,
+      "loss": 0.9488,
+      "step": 19978
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.151874833340903e-07,
+      "loss": 0.7951,
+      "step": 19979
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.148798353934476e-07,
+      "loss": 0.992,
+      "step": 19980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.145722743509907e-07,
+      "loss": 0.9239,
+      "step": 19981
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.142648002126782e-07,
+      "loss": 0.8939,
+      "step": 19982
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.139574129844682e-07,
+      "loss": 0.8551,
+      "step": 19983
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.136501126723154e-07,
+      "loss": 0.9779,
+      "step": 19984
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.133428992821743e-07,
+      "loss": 1.0189,
+      "step": 19985
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.130357728199953e-07,
+      "loss": 0.9443,
+      "step": 19986
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.127287332917319e-07,
+      "loss": 0.9545,
+      "step": 19987
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.12421780703331e-07,
+      "loss": 0.848,
+      "step": 19988
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.121149150607408e-07,
+      "loss": 0.6516,
+      "step": 19989
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.118081363699051e-07,
+      "loss": 0.8865,
+      "step": 19990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.115014446367695e-07,
+      "loss": 1.06,
+      "step": 19991
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.111948398672745e-07,
+      "loss": 0.8033,
+      "step": 19992
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.108883220673621e-07,
+      "loss": 0.8585,
+      "step": 19993
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.105818912429694e-07,
+      "loss": 0.7908,
+      "step": 19994
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.102755474000348e-07,
+      "loss": 0.9377,
+      "step": 19995
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.099692905444925e-07,
+      "loss": 0.9497,
+      "step": 19996
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.096631206822772e-07,
+      "loss": 0.919,
+      "step": 19997
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.093570378193185e-07,
+      "loss": 1.1143,
+      "step": 19998
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.090510419615507e-07,
+      "loss": 0.979,
+      "step": 19999
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.08745133114899e-07,
+      "loss": 1.0391,
+      "step": 20000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.084393112852931e-07,
+      "loss": 0.8277,
+      "step": 20001
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.081335764786549e-07,
+      "loss": 0.9865,
+      "step": 20002
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.078279287009113e-07,
+      "loss": 1.0317,
+      "step": 20003
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.075223679579816e-07,
+      "loss": 0.8411,
+      "step": 20004
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.072168942557876e-07,
+      "loss": 0.9555,
+      "step": 20005
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.069115076002462e-07,
+      "loss": 0.9436,
+      "step": 20006
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.066062079972767e-07,
+      "loss": 1.0086,
+      "step": 20007
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.063009954527909e-07,
+      "loss": 1.0105,
+      "step": 20008
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.059958699727053e-07,
+      "loss": 1.0536,
+      "step": 20009
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.056908315629294e-07,
+      "loss": 0.969,
+      "step": 20010
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.053858802293743e-07,
+      "loss": 0.9862,
+      "step": 20011
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.05081015977949e-07,
+      "loss": 0.8964,
+      "step": 20012
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.047762388145582e-07,
+      "loss": 1.0633,
+      "step": 20013
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.044715487451091e-07,
+      "loss": 1.0487,
+      "step": 20014
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.041669457755033e-07,
+      "loss": 0.8588,
+      "step": 20015
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.03862429911644e-07,
+      "loss": 0.9847,
+      "step": 20016
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.035580011594293e-07,
+      "loss": 0.9954,
+      "step": 20017
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.032536595247589e-07,
+      "loss": 0.9574,
+      "step": 20018
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.029494050135281e-07,
+      "loss": 0.9712,
+      "step": 20019
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.026452376316338e-07,
+      "loss": 0.9732,
+      "step": 20020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.023411573849657e-07,
+      "loss": 0.9434,
+      "step": 20021
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.020371642794192e-07,
+      "loss": 0.8432,
+      "step": 20022
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.017332583208811e-07,
+      "loss": 0.7572,
+      "step": 20023
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.014294395152425e-07,
+      "loss": 0.9535,
+      "step": 20024
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.011257078683868e-07,
+      "loss": 0.9845,
+      "step": 20025
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.008220633862005e-07,
+      "loss": 1.0398,
+      "step": 20026
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.005185060745654e-07,
+      "loss": 0.953,
+      "step": 20027
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.002150359393648e-07,
+      "loss": 0.885,
+      "step": 20028
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.999116529864762e-07,
+      "loss": 0.9265,
+      "step": 20029
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.996083572217791e-07,
+      "loss": 1.0783,
+      "step": 20030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.993051486511485e-07,
+      "loss": 1.0238,
+      "step": 20031
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.990020272804608e-07,
+      "loss": 0.8578,
+      "step": 20032
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.986989931155862e-07,
+      "loss": 0.8836,
+      "step": 20033
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.983960461623988e-07,
+      "loss": 0.935,
+      "step": 20034
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.980931864267663e-07,
+      "loss": 0.9561,
+      "step": 20035
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.977904139145579e-07,
+      "loss": 1.0222,
+      "step": 20036
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.974877286316376e-07,
+      "loss": 1.0036,
+      "step": 20037
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.971851305838727e-07,
+      "loss": 0.9905,
+      "step": 20038
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.968826197771237e-07,
+      "loss": 0.9475,
+      "step": 20039
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.965801962172528e-07,
+      "loss": 0.9601,
+      "step": 20040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.962778599101186e-07,
+      "loss": 0.9315,
+      "step": 20041
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.959756108615805e-07,
+      "loss": 0.7631,
+      "step": 20042
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.956734490774917e-07,
+      "loss": 0.951,
+      "step": 20043
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.953713745637079e-07,
+      "loss": 0.9025,
+      "step": 20044
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.950693873260831e-07,
+      "loss": 0.8932,
+      "step": 20045
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.947674873704661e-07,
+      "loss": 0.8631,
+      "step": 20046
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.944656747027083e-07,
+      "loss": 0.7935,
+      "step": 20047
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.941639493286549e-07,
+      "loss": 0.8522,
+      "step": 20048
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.938623112541541e-07,
+      "loss": 0.8346,
+      "step": 20049
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.935607604850473e-07,
+      "loss": 0.8505,
+      "step": 20050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.932592970271799e-07,
+      "loss": 0.8984,
+      "step": 20051
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.929579208863899e-07,
+      "loss": 0.8311,
+      "step": 20052
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.926566320685194e-07,
+      "loss": 0.9829,
+      "step": 20053
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.923554305794032e-07,
+      "loss": 0.8788,
+      "step": 20054
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.920543164248787e-07,
+      "loss": 1.0141,
+      "step": 20055
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.917532896107774e-07,
+      "loss": 0.9113,
+      "step": 20056
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.914523501429352e-07,
+      "loss": 0.7315,
+      "step": 20057
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.911514980271803e-07,
+      "loss": 0.8868,
+      "step": 20058
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.908507332693435e-07,
+      "loss": 0.9822,
+      "step": 20059
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.905500558752496e-07,
+      "loss": 0.9795,
+      "step": 20060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.902494658507262e-07,
+      "loss": 0.7716,
+      "step": 20061
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.899489632015958e-07,
+      "loss": 0.9994,
+      "step": 20062
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.896485479336822e-07,
+      "loss": 0.912,
+      "step": 20063
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.893482200528038e-07,
+      "loss": 0.9474,
+      "step": 20064
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.890479795647812e-07,
+      "loss": 0.8363,
+      "step": 20065
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.887478264754303e-07,
+      "loss": 1.0392,
+      "step": 20066
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.884477607905674e-07,
+      "loss": 1.0685,
+      "step": 20067
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.881477825160052e-07,
+      "loss": 0.8966,
+      "step": 20068
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.878478916575558e-07,
+      "loss": 0.9138,
+      "step": 20069
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.875480882210315e-07,
+      "loss": 0.9385,
+      "step": 20070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.872483722122379e-07,
+      "loss": 1.0668,
+      "step": 20071
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.869487436369846e-07,
+      "loss": 0.7339,
+      "step": 20072
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.866492025010749e-07,
+      "loss": 0.7833,
+      "step": 20073
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.863497488103136e-07,
+      "loss": 0.7655,
+      "step": 20074
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.860503825705016e-07,
+      "loss": 1.0507,
+      "step": 20075
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.857511037874402e-07,
+      "loss": 0.8516,
+      "step": 20076
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.854519124669261e-07,
+      "loss": 0.6964,
+      "step": 20077
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.851528086147583e-07,
+      "loss": 0.9728,
+      "step": 20078
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.848537922367286e-07,
+      "loss": 0.8532,
+      "step": 20079
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.845548633386343e-07,
+      "loss": 0.9785,
+      "step": 20080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.842560219262643e-07,
+      "loss": 0.9914,
+      "step": 20081
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.839572680054106e-07,
+      "loss": 0.9747,
+      "step": 20082
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.836586015818595e-07,
+      "loss": 0.8859,
+      "step": 20083
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.833600226613999e-07,
+      "loss": 1.0271,
+      "step": 20084
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.830615312498138e-07,
+      "loss": 0.8965,
+      "step": 20085
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.827631273528871e-07,
+      "loss": 1.0295,
+      "step": 20086
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.824648109763991e-07,
+      "loss": 0.9272,
+      "step": 20087
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.821665821261312e-07,
+      "loss": 1.0416,
+      "step": 20088
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.818684408078605e-07,
+      "loss": 1.0178,
+      "step": 20089
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.815703870273647e-07,
+      "loss": 0.8006,
+      "step": 20090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.812724207904157e-07,
+      "loss": 0.8298,
+      "step": 20091
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.809745421027906e-07,
+      "loss": 0.8759,
+      "step": 20092
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.806767509702576e-07,
+      "loss": 0.8404,
+      "step": 20093
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.803790473985886e-07,
+      "loss": 1.0206,
+      "step": 20094
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.800814313935492e-07,
+      "loss": 0.997,
+      "step": 20095
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.797839029609075e-07,
+      "loss": 1.0159,
+      "step": 20096
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.794864621064266e-07,
+      "loss": 0.936,
+      "step": 20097
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.791891088358714e-07,
+      "loss": 0.9874,
+      "step": 20098
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.78891843155e-07,
+      "loss": 1.0094,
+      "step": 20099
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.78594665069575e-07,
+      "loss": 0.8449,
+      "step": 20100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.782975745853519e-07,
+      "loss": 0.8643,
+      "step": 20101
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.780005717080871e-07,
+      "loss": 0.9324,
+      "step": 20102
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.777036564435367e-07,
+      "loss": 0.8866,
+      "step": 20103
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.774068287974515e-07,
+      "loss": 0.8087,
+      "step": 20104
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.771100887755836e-07,
+      "loss": 0.9117,
+      "step": 20105
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7681343638368074e-07,
+      "loss": 0.9134,
+      "step": 20106
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.765168716274926e-07,
+      "loss": 0.8879,
+      "step": 20107
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7622039451276305e-07,
+      "loss": 1.0107,
+      "step": 20108
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.759240050452385e-07,
+      "loss": 0.8278,
+      "step": 20109
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.756277032306589e-07,
+      "loss": 0.9032,
+      "step": 20110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7533148907476676e-07,
+      "loss": 1.0649,
+      "step": 20111
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.750353625833004e-07,
+      "loss": 1.0171,
+      "step": 20112
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7473932376199794e-07,
+      "loss": 1.0258,
+      "step": 20113
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7444337261659314e-07,
+      "loss": 0.8932,
+      "step": 20114
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.74147509152823e-07,
+      "loss": 0.8645,
+      "step": 20115
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.73851733376417e-07,
+      "loss": 0.9718,
+      "step": 20116
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.735560452931082e-07,
+      "loss": 1.0035,
+      "step": 20117
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7326044490862364e-07,
+      "loss": 0.9932,
+      "step": 20118
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7296493222869166e-07,
+      "loss": 0.9526,
+      "step": 20119
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7266950725903646e-07,
+      "loss": 0.9828,
+      "step": 20120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7237417000538356e-07,
+      "loss": 0.9092,
+      "step": 20121
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.720789204734527e-07,
+      "loss": 1.0065,
+      "step": 20122
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7178375866896717e-07,
+      "loss": 0.8264,
+      "step": 20123
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.71488684597643e-07,
+      "loss": 0.7934,
+      "step": 20124
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7119369826519934e-07,
+      "loss": 1.0126,
+      "step": 20125
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.708987996773484e-07,
+      "loss": 1.0261,
+      "step": 20126
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7060398883980787e-07,
+      "loss": 0.9825,
+      "step": 20127
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.703092657582864e-07,
+      "loss": 0.9202,
+      "step": 20128
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.700146304384967e-07,
+      "loss": 0.9164,
+      "step": 20129
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.697200828861448e-07,
+      "loss": 0.9938,
+      "step": 20130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.694256231069394e-07,
+      "loss": 0.9602,
+      "step": 20131
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6913125110658433e-07,
+      "loss": 0.9927,
+      "step": 20132
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.688369668907844e-07,
+      "loss": 0.7684,
+      "step": 20133
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.68542770465239e-07,
+      "loss": 0.8944,
+      "step": 20134
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6824866183565077e-07,
+      "loss": 0.7689,
+      "step": 20135
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6795464100771626e-07,
+      "loss": 0.9952,
+      "step": 20136
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6766070798713205e-07,
+      "loss": 1.0125,
+      "step": 20137
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6736686277959465e-07,
+      "loss": 0.9214,
+      "step": 20138
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6707310539079453e-07,
+      "loss": 0.8541,
+      "step": 20139
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.667794358264266e-07,
+      "loss": 1.0199,
+      "step": 20140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.664858540921774e-07,
+      "loss": 1.0427,
+      "step": 20141
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6619236019373735e-07,
+      "loss": 0.8083,
+      "step": 20142
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6589895413679087e-07,
+      "loss": 0.8398,
+      "step": 20143
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6560563592702446e-07,
+      "loss": 0.9453,
+      "step": 20144
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.653124055701197e-07,
+      "loss": 0.7934,
+      "step": 20145
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6501926307175925e-07,
+      "loss": 0.8804,
+      "step": 20146
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.647262084376214e-07,
+      "loss": 0.9575,
+      "step": 20147
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.644332416733849e-07,
+      "loss": 0.8522,
+      "step": 20148
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.641403627847235e-07,
+      "loss": 1.0668,
+      "step": 20149
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6384757177731556e-07,
+      "loss": 0.6878,
+      "step": 20150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.635548686568314e-07,
+      "loss": 0.8896,
+      "step": 20151
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6326225342894326e-07,
+      "loss": 0.9123,
+      "step": 20152
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6296972609931936e-07,
+      "loss": 0.8812,
+      "step": 20153
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6267728667362843e-07,
+      "loss": 0.9935,
+      "step": 20154
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.623849351575349e-07,
+      "loss": 0.9261,
+      "step": 20155
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6209267155670476e-07,
+      "loss": 0.91,
+      "step": 20156
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.618004958767991e-07,
+      "loss": 0.8522,
+      "step": 20157
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6150840812348e-07,
+      "loss": 0.7255,
+      "step": 20158
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.612164083024051e-07,
+      "loss": 0.8643,
+      "step": 20159
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6092449641923385e-07,
+      "loss": 1.0698,
+      "step": 20160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.606326724796195e-07,
+      "loss": 0.9642,
+      "step": 20161
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.603409364892181e-07,
+      "loss": 0.9575,
+      "step": 20162
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.600492884536817e-07,
+      "loss": 0.7278,
+      "step": 20163
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.597577283786597e-07,
+      "loss": 0.9206,
+      "step": 20164
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5946625626980323e-07,
+      "loss": 0.9144,
+      "step": 20165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.591748721327566e-07,
+      "loss": 0.9788,
+      "step": 20166
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5888357597316747e-07,
+      "loss": 0.7912,
+      "step": 20167
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.585923677966786e-07,
+      "loss": 0.9305,
+      "step": 20168
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5830124760893325e-07,
+      "loss": 1.017,
+      "step": 20169
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5801021541556966e-07,
+      "loss": 1.019,
+      "step": 20170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5771927122222947e-07,
+      "loss": 1.001,
+      "step": 20171
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5742841503454593e-07,
+      "loss": 0.858,
+      "step": 20172
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.571376468581579e-07,
+      "loss": 1.0554,
+      "step": 20173
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5684696669869686e-07,
+      "loss": 1.03,
+      "step": 20174
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.565563745617962e-07,
+      "loss": 0.8262,
+      "step": 20175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.562658704530842e-07,
+      "loss": 0.7549,
+      "step": 20176
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5597545437819133e-07,
+      "loss": 0.752,
+      "step": 20177
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5568512634274196e-07,
+      "loss": 1.0664,
+      "step": 20178
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5539488635236385e-07,
+      "loss": 0.8274,
+      "step": 20179
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5510473441267743e-07,
+      "loss": 1.0151,
+      "step": 20180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5481467052930716e-07,
+      "loss": 0.8499,
+      "step": 20181
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5452469470787073e-07,
+      "loss": 1.0266,
+      "step": 20182
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5423480695398814e-07,
+      "loss": 1.0017,
+      "step": 20183
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5394500727327316e-07,
+      "loss": 0.9903,
+      "step": 20184
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5365529567134416e-07,
+      "loss": 1.005,
+      "step": 20185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5336567215381156e-07,
+      "loss": 0.9193,
+      "step": 20186
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.530761367262887e-07,
+      "loss": 0.6981,
+      "step": 20187
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5278668939438333e-07,
+      "loss": 1.0099,
+      "step": 20188
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5249733016370533e-07,
+      "loss": 0.9613,
+      "step": 20189
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.522080590398592e-07,
+      "loss": 0.9293,
+      "step": 20190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.51918876028451e-07,
+      "loss": 0.8707,
+      "step": 20191
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5162978113508225e-07,
+      "loss": 0.915,
+      "step": 20192
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.513407743653558e-07,
+      "loss": 0.9253,
+      "step": 20193
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.510518557248689e-07,
+      "loss": 0.8193,
+      "step": 20194
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5076302521922024e-07,
+      "loss": 0.9886,
+      "step": 20195
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5047428285400705e-07,
+      "loss": 0.9686,
+      "step": 20196
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.501856286348216e-07,
+      "loss": 0.8446,
+      "step": 20197
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.498970625672588e-07,
+      "loss": 0.7704,
+      "step": 20198
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4960858465690704e-07,
+      "loss": 0.9501,
+      "step": 20199
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4932019490935786e-07,
+      "loss": 1.0166,
+      "step": 20200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.490318933301968e-07,
+      "loss": 0.873,
+      "step": 20201
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.487436799250111e-07,
+      "loss": 1.0616,
+      "step": 20202
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.484555546993835e-07,
+      "loss": 0.8969,
+      "step": 20203
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.481675176588979e-07,
+      "loss": 0.7344,
+      "step": 20204
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4787956880913307e-07,
+      "loss": 0.9325,
+      "step": 20205
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.475917081556702e-07,
+      "loss": 0.8374,
+      "step": 20206
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4730393570408314e-07,
+      "loss": 0.9313,
+      "step": 20207
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4701625145995186e-07,
+      "loss": 0.9573,
+      "step": 20208
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4672865542884693e-07,
+      "loss": 0.8834,
+      "step": 20209
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4644114761634215e-07,
+      "loss": 0.9237,
+      "step": 20210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.461537280280065e-07,
+      "loss": 0.9651,
+      "step": 20211
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.458663966694099e-07,
+      "loss": 0.9221,
+      "step": 20212
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4557915354611845e-07,
+      "loss": 1.13,
+      "step": 20213
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.452919986636983e-07,
+      "loss": 1.1593,
+      "step": 20214
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4500493202771166e-07,
+      "loss": 0.8177,
+      "step": 20215
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.447179536437224e-07,
+      "loss": 0.7755,
+      "step": 20216
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4443106351728827e-07,
+      "loss": 0.9446,
+      "step": 20217
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4414426165396984e-07,
+      "loss": 0.8773,
+      "step": 20218
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.43857548059321e-07,
+      "loss": 0.8728,
+      "step": 20219
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4357092273890067e-07,
+      "loss": 0.9501,
+      "step": 20220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.432843856982588e-07,
+      "loss": 1.0716,
+      "step": 20221
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.429979369429499e-07,
+      "loss": 0.9539,
+      "step": 20222
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.427115764785211e-07,
+      "loss": 0.744,
+      "step": 20223
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4242530431052253e-07,
+      "loss": 0.9053,
+      "step": 20224
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.421391204444991e-07,
+      "loss": 1.0376,
+      "step": 20225
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.41853024885997e-07,
+      "loss": 0.9636,
+      "step": 20226
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4156701764055785e-07,
+      "loss": 0.9093,
+      "step": 20227
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4128109871372447e-07,
+      "loss": 0.9283,
+      "step": 20228
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.409952681110352e-07,
+      "loss": 0.7851,
+      "step": 20229
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4070952583802783e-07,
+      "loss": 1.0537,
+      "step": 20230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4042387190024073e-07,
+      "loss": 0.8815,
+      "step": 20231
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.401383063032055e-07,
+      "loss": 1.0278,
+      "step": 20232
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.398528290524573e-07,
+      "loss": 1.022,
+      "step": 20233
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.395674401535255e-07,
+      "loss": 1.0579,
+      "step": 20234
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.392821396119407e-07,
+      "loss": 0.8422,
+      "step": 20235
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.389969274332284e-07,
+      "loss": 0.9198,
+      "step": 20236
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3871180362291763e-07,
+      "loss": 1.0363,
+      "step": 20237
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3842676818653007e-07,
+      "loss": 1.0356,
+      "step": 20238
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3814182112959013e-07,
+      "loss": 0.9925,
+      "step": 20239
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3785696245761624e-07,
+      "loss": 0.9469,
+      "step": 20240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3757219217613e-07,
+      "loss": 1.0004,
+      "step": 20241
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.37287510290646e-07,
+      "loss": 0.8554,
+      "step": 20242
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3700291680668305e-07,
+      "loss": 1.0532,
+      "step": 20243
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3671841172975236e-07,
+      "loss": 0.9384,
+      "step": 20244
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3643399506536843e-07,
+      "loss": 0.9828,
+      "step": 20245
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3614966681904015e-07,
+      "loss": 1.0043,
+      "step": 20246
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.35865426996277e-07,
+      "loss": 0.8489,
+      "step": 20247
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3558127560258565e-07,
+      "loss": 0.9366,
+      "step": 20248
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3529721264347233e-07,
+      "loss": 1.0146,
+      "step": 20249
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3501323812443927e-07,
+      "loss": 0.9936,
+      "step": 20250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3472935205098986e-07,
+      "loss": 1.0184,
+      "step": 20251
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.34445554428623e-07,
+      "loss": 1.0743,
+      "step": 20252
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3416184526283876e-07,
+      "loss": 1.0043,
+      "step": 20253
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.338782245591322e-07,
+      "loss": 0.9719,
+      "step": 20254
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.335946923229989e-07,
+      "loss": 0.9905,
+      "step": 20255
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.333112485599339e-07,
+      "loss": 0.8625,
+      "step": 20256
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.330278932754267e-07,
+      "loss": 0.9976,
+      "step": 20257
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3274462647496853e-07,
+      "loss": 0.7327,
+      "step": 20258
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3246144816404657e-07,
+      "loss": 1.0139,
+      "step": 20259
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3217835834814925e-07,
+      "loss": 0.9431,
+      "step": 20260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.318953570327589e-07,
+      "loss": 0.8533,
+      "step": 20261
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.316124442233605e-07,
+      "loss": 0.8633,
+      "step": 20262
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.313296199254341e-07,
+      "loss": 0.933,
+      "step": 20263
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3104688414446037e-07,
+      "loss": 0.8797,
+      "step": 20264
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.307642368859155e-07,
+      "loss": 1.0129,
+      "step": 20265
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3048167815527895e-07,
+      "loss": 0.8314,
+      "step": 20266
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3019920795802196e-07,
+      "loss": 0.9813,
+      "step": 20267
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2991682629962007e-07,
+      "loss": 1.0209,
+      "step": 20268
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.296345331855417e-07,
+      "loss": 1.1129,
+      "step": 20269
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2935232862125864e-07,
+      "loss": 0.9224,
+      "step": 20270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2907021261223646e-07,
+      "loss": 0.8991,
+      "step": 20271
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2878818516394307e-07,
+      "loss": 0.8736,
+      "step": 20272
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.285062462818407e-07,
+      "loss": 1.081,
+      "step": 20273
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2822439597139386e-07,
+      "loss": 0.8642,
+      "step": 20274
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.279426342380616e-07,
+      "loss": 1.0154,
+      "step": 20275
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2766096108730395e-07,
+      "loss": 1.0541,
+      "step": 20276
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.273793765245765e-07,
+      "loss": 0.9687,
+      "step": 20277
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.270978805553383e-07,
+      "loss": 1.0823,
+      "step": 20278
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.268164731850405e-07,
+      "loss": 1.0404,
+      "step": 20279
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.265351544191365e-07,
+      "loss": 1.0656,
+      "step": 20280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2625392426307646e-07,
+      "loss": 0.8931,
+      "step": 20281
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2597278272230923e-07,
+      "loss": 0.9963,
+      "step": 20282
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2569172980228115e-07,
+      "loss": 1.0733,
+      "step": 20283
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.254107655084394e-07,
+      "loss": 0.8745,
+      "step": 20284
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2512988984622474e-07,
+      "loss": 0.8337,
+      "step": 20285
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2484910282108215e-07,
+      "loss": 0.8996,
+      "step": 20286
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.245684044384496e-07,
+      "loss": 1.0121,
+      "step": 20287
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.24287794703766e-07,
+      "loss": 0.8602,
+      "step": 20288
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2400727362246927e-07,
+      "loss": 0.8788,
+      "step": 20289
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2372684119999287e-07,
+      "loss": 1.0912,
+      "step": 20290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.234464974417718e-07,
+      "loss": 0.9236,
+      "step": 20291
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.231662423532357e-07,
+      "loss": 1.0472,
+      "step": 20292
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.228860759398168e-07,
+      "loss": 0.9114,
+      "step": 20293
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2260599820694025e-07,
+      "loss": 0.976,
+      "step": 20294
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.223260091600356e-07,
+      "loss": 0.8912,
+      "step": 20295
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2204610880452525e-07,
+      "loss": 1.0107,
+      "step": 20296
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.217662971458336e-07,
+      "loss": 0.9691,
+      "step": 20297
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.214865741893809e-07,
+      "loss": 0.9428,
+      "step": 20298
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.212069399405877e-07,
+      "loss": 0.8324,
+      "step": 20299
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.209273944048703e-07,
+      "loss": 0.9503,
+      "step": 20300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.206479375876471e-07,
+      "loss": 0.9156,
+      "step": 20301
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.203685694943305e-07,
+      "loss": 0.923,
+      "step": 20302
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.200892901303355e-07,
+      "loss": 0.9868,
+      "step": 20303
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.198100995010701e-07,
+      "loss": 0.8436,
+      "step": 20304
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1953099761194606e-07,
+      "loss": 0.8838,
+      "step": 20305
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1925198446836903e-07,
+      "loss": 0.9921,
+      "step": 20306
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.189730600757469e-07,
+      "loss": 0.8936,
+      "step": 20307
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.186942244394815e-07,
+      "loss": 0.9215,
+      "step": 20308
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.184154775649768e-07,
+      "loss": 1.0179,
+      "step": 20309
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.181368194576324e-07,
+      "loss": 0.834,
+      "step": 20310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1785825012284896e-07,
+      "loss": 0.9268,
+      "step": 20311
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.175797695660211e-07,
+      "loss": 1.0477,
+      "step": 20312
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.173013777925461e-07,
+      "loss": 0.9214,
+      "step": 20313
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.17023074807818e-07,
+      "loss": 0.9423,
+      "step": 20314
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.167448606172275e-07,
+      "loss": 0.9585,
+      "step": 20315
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.164667352261664e-07,
+      "loss": 0.9235,
+      "step": 20316
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.161886986400221e-07,
+      "loss": 0.8671,
+      "step": 20317
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1591075086418186e-07,
+      "loss": 1.0227,
+      "step": 20318
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1563289190403146e-07,
+      "loss": 0.9888,
+      "step": 20319
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.153551217649532e-07,
+      "loss": 1.1107,
+      "step": 20320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1507744045233054e-07,
+      "loss": 1.0795,
+      "step": 20321
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.147998479715415e-07,
+      "loss": 0.9829,
+      "step": 20322
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1452234432796556e-07,
+      "loss": 1.0021,
+      "step": 20323
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.142449295269796e-07,
+      "loss": 0.9468,
+      "step": 20324
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.13967603573957e-07,
+      "loss": 1.1346,
+      "step": 20325
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1369036647427254e-07,
+      "loss": 1.0163,
+      "step": 20326
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.134132182332962e-07,
+      "loss": 0.9025,
+      "step": 20327
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.131361588563992e-07,
+      "loss": 0.9026,
+      "step": 20328
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.128591883489474e-07,
+      "loss": 0.8513,
+      "step": 20329
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1258230671630983e-07,
+      "loss": 1.0431,
+      "step": 20330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.123055139638482e-07,
+      "loss": 1.0509,
+      "step": 20331
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1202881009692777e-07,
+      "loss": 0.9238,
+      "step": 20332
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.117521951209069e-07,
+      "loss": 0.856,
+      "step": 20333
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.114756690411481e-07,
+      "loss": 0.9574,
+      "step": 20334
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.111992318630054e-07,
+      "loss": 0.9175,
+      "step": 20335
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.109228835918383e-07,
+      "loss": 0.721,
+      "step": 20336
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.106466242329982e-07,
+      "loss": 0.7622,
+      "step": 20337
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1037045379184016e-07,
+      "loss": 1.0829,
+      "step": 20338
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.100943722737122e-07,
+      "loss": 1.0473,
+      "step": 20339
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0981837968396555e-07,
+      "loss": 1.1666,
+      "step": 20340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.095424760279454e-07,
+      "loss": 0.8502,
+      "step": 20341
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0926666131099966e-07,
+      "loss": 0.753,
+      "step": 20342
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0899093553847024e-07,
+      "loss": 0.9166,
+      "step": 20343
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0871529871570114e-07,
+      "loss": 1.0042,
+      "step": 20344
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0843975084803034e-07,
+      "loss": 0.9412,
+      "step": 20345
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0816429194079857e-07,
+      "loss": 0.8319,
+      "step": 20346
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.078889219993415e-07,
+      "loss": 0.9351,
+      "step": 20347
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.076136410289949e-07,
+      "loss": 0.9838,
+      "step": 20348
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0733844903509287e-07,
+      "loss": 1.0367,
+      "step": 20349
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0706334602296614e-07,
+      "loss": 0.8173,
+      "step": 20350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.06788331997946e-07,
+      "loss": 1.0848,
+      "step": 20351
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0651340696535866e-07,
+      "loss": 1.017,
+      "step": 20352
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.062385709305333e-07,
+      "loss": 1.0869,
+      "step": 20353
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0596382389879287e-07,
+      "loss": 0.8176,
+      "step": 20354
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0568916587546193e-07,
+      "loss": 1.0673,
+      "step": 20355
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.054145968658607e-07,
+      "loss": 0.788,
+      "step": 20356
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0514011687530996e-07,
+      "loss": 0.9007,
+      "step": 20357
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0486572590912544e-07,
+      "loss": 0.9466,
+      "step": 20358
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0459142397262673e-07,
+      "loss": 1.0714,
+      "step": 20359
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.043172110711263e-07,
+      "loss": 1.2476,
+      "step": 20360
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.040430872099382e-07,
+      "loss": 0.9813,
+      "step": 20361
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.037690523943716e-07,
+      "loss": 1.0533,
+      "step": 20362
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0349510662973825e-07,
+      "loss": 0.9932,
+      "step": 20363
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0322124992134337e-07,
+      "loss": 0.9288,
+      "step": 20364
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0294748227449497e-07,
+      "loss": 1.0056,
+      "step": 20365
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0267380369449547e-07,
+      "loss": 1.0231,
+      "step": 20366
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.02400214186649e-07,
+      "loss": 1.0017,
+      "step": 20367
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.021267137562546e-07,
+      "loss": 0.9411,
+      "step": 20368
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0185330240861296e-07,
+      "loss": 0.9675,
+      "step": 20369
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.015799801490189e-07,
+      "loss": 0.8721,
+      "step": 20370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.013067469827714e-07,
+      "loss": 0.9495,
+      "step": 20371
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.010336029151618e-07,
+      "loss": 0.812,
+      "step": 20372
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0076054795148367e-07,
+      "loss": 0.9786,
+      "step": 20373
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.004875820970261e-07,
+      "loss": 0.9361,
+      "step": 20374
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0021470535707873e-07,
+      "loss": 0.9466,
+      "step": 20375
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.999419177369268e-07,
+      "loss": 1.0077,
+      "step": 20376
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.996692192418583e-07,
+      "loss": 0.9506,
+      "step": 20377
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.99396609877154e-07,
+      "loss": 0.9705,
+      "step": 20378
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.99124089648098e-07,
+      "loss": 1.0027,
+      "step": 20379
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9885165855996774e-07,
+      "loss": 1.0164,
+      "step": 20380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9857931661804296e-07,
+      "loss": 0.7782,
+      "step": 20381
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9830706382760096e-07,
+      "loss": 0.8642,
+      "step": 20382
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9803490019391545e-07,
+      "loss": 0.8833,
+      "step": 20383
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9776282572226045e-07,
+      "loss": 0.7017,
+      "step": 20384
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9749084041790565e-07,
+      "loss": 0.7904,
+      "step": 20385
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.972189442861224e-07,
+      "loss": 1.0825,
+      "step": 20386
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.969471373321776e-07,
+      "loss": 0.9941,
+      "step": 20387
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.966754195613387e-07,
+      "loss": 0.8592,
+      "step": 20388
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.964037909788687e-07,
+      "loss": 0.8616,
+      "step": 20389
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9613225159003124e-07,
+      "loss": 0.7027,
+      "step": 20390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.958608014000864e-07,
+      "loss": 0.8745,
+      "step": 20391
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.955894404142957e-07,
+      "loss": 1.0243,
+      "step": 20392
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9531816863791263e-07,
+      "loss": 0.8846,
+      "step": 20393
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.950469860761974e-07,
+      "loss": 0.9851,
+      "step": 20394
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.947758927344014e-07,
+      "loss": 0.8542,
+      "step": 20395
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.945048886177788e-07,
+      "loss": 0.7699,
+      "step": 20396
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9423397373157814e-07,
+      "loss": 0.9979,
+      "step": 20397
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9396314808105073e-07,
+      "loss": 0.8548,
+      "step": 20398
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9369241167144136e-07,
+      "loss": 0.8819,
+      "step": 20399
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.93421764507998e-07,
+      "loss": 0.9784,
+      "step": 20400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.931512065959614e-07,
+      "loss": 1.0427,
+      "step": 20401
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9288073794057634e-07,
+      "loss": 0.8839,
+      "step": 20402
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.926103585470814e-07,
+      "loss": 1.0656,
+      "step": 20403
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9234006842071625e-07,
+      "loss": 1.0812,
+      "step": 20404
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9206986756671616e-07,
+      "loss": 0.8901,
+      "step": 20405
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9179975599031695e-07,
+      "loss": 0.9734,
+      "step": 20406
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.915297336967533e-07,
+      "loss": 1.0279,
+      "step": 20407
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9125980069125436e-07,
+      "loss": 0.763,
+      "step": 20408
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.909899569790526e-07,
+      "loss": 1.0443,
+      "step": 20409
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9072020256537447e-07,
+      "loss": 0.9823,
+      "step": 20410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.904505374554468e-07,
+      "loss": 0.8503,
+      "step": 20411
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.901809616544944e-07,
+      "loss": 1.0085,
+      "step": 20412
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.899114751677402e-07,
+      "loss": 0.843,
+      "step": 20413
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.896420780004051e-07,
+      "loss": 1.004,
+      "step": 20414
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.893727701577088e-07,
+      "loss": 0.8353,
+      "step": 20415
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8910355164486934e-07,
+      "loss": 1.0082,
+      "step": 20416
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.888344224671031e-07,
+      "loss": 1.0431,
+      "step": 20417
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8856538262962375e-07,
+      "loss": 0.991,
+      "step": 20418
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.882964321376442e-07,
+      "loss": 0.9486,
+      "step": 20419
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.880275709963749e-07,
+      "loss": 0.9742,
+      "step": 20420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.87758799211026e-07,
+      "loss": 0.9678,
+      "step": 20421
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.874901167868028e-07,
+      "loss": 0.9172,
+      "step": 20422
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.872215237289134e-07,
+      "loss": 0.9065,
+      "step": 20423
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8695302004255976e-07,
+      "loss": 0.8429,
+      "step": 20424
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8668460573294597e-07,
+      "loss": 0.9489,
+      "step": 20425
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.864162808052707e-07,
+      "loss": 0.7155,
+      "step": 20426
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.861480452647343e-07,
+      "loss": 0.8074,
+      "step": 20427
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.858798991165308e-07,
+      "loss": 0.9522,
+      "step": 20428
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8561184236585945e-07,
+      "loss": 0.8784,
+      "step": 20429
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8534387501791116e-07,
+      "loss": 0.8283,
+      "step": 20430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.850759970778795e-07,
+      "loss": 0.9491,
+      "step": 20431
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8480820855095245e-07,
+      "loss": 1.0488,
+      "step": 20432
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8454050944232047e-07,
+      "loss": 0.824,
+      "step": 20433
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8427289975716817e-07,
+      "loss": 0.8295,
+      "step": 20434
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8400537950068263e-07,
+      "loss": 1.0075,
+      "step": 20435
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8373794867804457e-07,
+      "loss": 0.8708,
+      "step": 20436
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8347060729443777e-07,
+      "loss": 0.8783,
+      "step": 20437
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.832033553550396e-07,
+      "loss": 0.8787,
+      "step": 20438
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.829361928650299e-07,
+      "loss": 0.965,
+      "step": 20439
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8266911982958397e-07,
+      "loss": 0.9601,
+      "step": 20440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.82402136253876e-07,
+      "loss": 0.9125,
+      "step": 20441
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8213524214307965e-07,
+      "loss": 0.9039,
+      "step": 20442
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8186843750236514e-07,
+      "loss": 0.874,
+      "step": 20443
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8160172233690283e-07,
+      "loss": 0.9324,
+      "step": 20444
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8133509665185866e-07,
+      "loss": 0.895,
+      "step": 20445
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8106856045240006e-07,
+      "loss": 0.9579,
+      "step": 20446
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.808021137436896e-07,
+      "loss": 0.8111,
+      "step": 20447
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8053575653089145e-07,
+      "loss": 0.9341,
+      "step": 20448
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.802694888191638e-07,
+      "loss": 0.9162,
+      "step": 20449
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.80003310613668e-07,
+      "loss": 0.9735,
+      "step": 20450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.797372219195583e-07,
+      "loss": 0.8452,
+      "step": 20451
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.794712227419939e-07,
+      "loss": 0.9701,
+      "step": 20452
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7920531308612505e-07,
+      "loss": 1.038,
+      "step": 20453
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.789394929571061e-07,
+      "loss": 1.0681,
+      "step": 20454
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.786737623600856e-07,
+      "loss": 0.8541,
+      "step": 20455
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7840812130021333e-07,
+      "loss": 1.0582,
+      "step": 20456
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7814256978263465e-07,
+      "loss": 0.9145,
+      "step": 20457
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.77877107812496e-07,
+      "loss": 1.0621,
+      "step": 20458
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7761173539493933e-07,
+      "loss": 1.0,
+      "step": 20459
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7734645253510726e-07,
+      "loss": 0.8501,
+      "step": 20460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7708125923813844e-07,
+      "loss": 0.8887,
+      "step": 20461
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7681615550917203e-07,
+      "loss": 0.9415,
+      "step": 20462
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.765511413533429e-07,
+      "loss": 1.0485,
+      "step": 20463
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7628621677578803e-07,
+      "loss": 1.0052,
+      "step": 20464
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.760213817816377e-07,
+      "loss": 0.8803,
+      "step": 20465
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7575663637602513e-07,
+      "loss": 0.9386,
+      "step": 20466
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7549198056407833e-07,
+      "loss": 1.1297,
+      "step": 20467
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.75227414350926e-07,
+      "loss": 0.8657,
+      "step": 20468
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7496293774169247e-07,
+      "loss": 0.6804,
+      "step": 20469
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.746985507415041e-07,
+      "loss": 0.8342,
+      "step": 20470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.744342533554807e-07,
+      "loss": 0.8928,
+      "step": 20471
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7417004558874595e-07,
+      "loss": 0.8294,
+      "step": 20472
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7390592744641573e-07,
+      "loss": 0.7589,
+      "step": 20473
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7364189893360935e-07,
+      "loss": 0.9777,
+      "step": 20474
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7337796005544215e-07,
+      "loss": 0.8301,
+      "step": 20475
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7311411081702666e-07,
+      "loss": 1.0537,
+      "step": 20476
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.728503512234766e-07,
+      "loss": 0.9274,
+      "step": 20477
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.725866812799006e-07,
+      "loss": 0.9497,
+      "step": 20478
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.723231009914091e-07,
+      "loss": 0.9299,
+      "step": 20479
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.720596103631069e-07,
+      "loss": 0.998,
+      "step": 20480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.717962094001004e-07,
+      "loss": 1.0293,
+      "step": 20481
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7153289810749173e-07,
+      "loss": 0.8403,
+      "step": 20482
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.712696764903845e-07,
+      "loss": 0.8489,
+      "step": 20483
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7100654455387574e-07,
+      "loss": 0.9479,
+      "step": 20484
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.707435023030664e-07,
+      "loss": 1.0201,
+      "step": 20485
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7048054974305023e-07,
+      "loss": 0.996,
+      "step": 20486
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.702176868789248e-07,
+      "loss": 0.8913,
+      "step": 20487
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.699549137157804e-07,
+      "loss": 0.8387,
+      "step": 20488
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6969223025871024e-07,
+      "loss": 0.8953,
+      "step": 20489
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6942963651280195e-07,
+      "loss": 0.9036,
+      "step": 20490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.691671324831447e-07,
+      "loss": 0.8025,
+      "step": 20491
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.689047181748234e-07,
+      "loss": 0.9061,
+      "step": 20492
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6864239359292343e-07,
+      "loss": 0.8873,
+      "step": 20493
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.683801587425251e-07,
+      "loss": 0.8371,
+      "step": 20494
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6811801362871215e-07,
+      "loss": 0.8071,
+      "step": 20495
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.678559582565605e-07,
+      "loss": 0.9066,
+      "step": 20496
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6759399263115003e-07,
+      "loss": 0.9851,
+      "step": 20497
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6733211675755445e-07,
+      "loss": 0.9139,
+      "step": 20498
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.670703306408485e-07,
+      "loss": 0.9329,
+      "step": 20499
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6680863428610434e-07,
+      "loss": 1.0513,
+      "step": 20500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6654702769839124e-07,
+      "loss": 1.1805,
+      "step": 20501
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6628551088277895e-07,
+      "loss": 0.9622,
+      "step": 20502
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6602408384433354e-07,
+      "loss": 0.9148,
+      "step": 20503
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6576274658812083e-07,
+      "loss": 1.0293,
+      "step": 20504
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6550149911920297e-07,
+      "loss": 1.2599,
+      "step": 20505
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.652403414426431e-07,
+      "loss": 0.8136,
+      "step": 20506
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.649792735634994e-07,
+      "loss": 0.8711,
+      "step": 20507
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6471829548683167e-07,
+      "loss": 0.9829,
+      "step": 20508
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6445740721769375e-07,
+      "loss": 0.9721,
+      "step": 20509
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.641966087611437e-07,
+      "loss": 0.9921,
+      "step": 20510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.63935900122232e-07,
+      "loss": 1.0517,
+      "step": 20511
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.636752813060118e-07,
+      "loss": 0.9342,
+      "step": 20512
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.634147523175302e-07,
+      "loss": 1.0401,
+      "step": 20513
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.631543131618365e-07,
+      "loss": 1.051,
+      "step": 20514
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.62893963843976e-07,
+      "loss": 0.9541,
+      "step": 20515
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.626337043689937e-07,
+      "loss": 0.8375,
+      "step": 20516
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.623735347419305e-07,
+      "loss": 0.9939,
+      "step": 20517
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6211345496782846e-07,
+      "loss": 1.0059,
+      "step": 20518
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.618534650517258e-07,
+      "loss": 0.7947,
+      "step": 20519
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.615935649986613e-07,
+      "loss": 0.9361,
+      "step": 20520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.613337548136675e-07,
+      "loss": 0.9416,
+      "step": 20521
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.61074034501781e-07,
+      "loss": 1.0858,
+      "step": 20522
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.608144040680328e-07,
+      "loss": 0.9804,
+      "step": 20523
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6055486351745327e-07,
+      "loss": 0.8894,
+      "step": 20524
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6029541285507065e-07,
+      "loss": 0.8471,
+      "step": 20525
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.600360520859125e-07,
+      "loss": 1.0027,
+      "step": 20526
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.597767812150027e-07,
+      "loss": 0.9827,
+      "step": 20527
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.595176002473655e-07,
+      "loss": 0.8625,
+      "step": 20528
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.592585091880219e-07,
+      "loss": 0.9378,
+      "step": 20529
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.589995080419928e-07,
+      "loss": 0.8994,
+      "step": 20530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.587405968142943e-07,
+      "loss": 0.9107,
+      "step": 20531
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5848177550994454e-07,
+      "loss": 0.8257,
+      "step": 20532
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.582230441339574e-07,
+      "loss": 0.9388,
+      "step": 20533
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5796440269134537e-07,
+      "loss": 0.9395,
+      "step": 20534
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.577058511871212e-07,
+      "loss": 0.9078,
+      "step": 20535
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.57447389626292e-07,
+      "loss": 0.8429,
+      "step": 20536
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5718901801386706e-07,
+      "loss": 0.9374,
+      "step": 20537
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.569307363548513e-07,
+      "loss": 0.9508,
+      "step": 20538
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5667254465425015e-07,
+      "loss": 1.0105,
+      "step": 20539
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.564144429170646e-07,
+      "loss": 0.8798,
+      "step": 20540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5615643114829626e-07,
+      "loss": 0.9752,
+      "step": 20541
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.558985093529427e-07,
+      "loss": 1.1024,
+      "step": 20542
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.556406775360033e-07,
+      "loss": 0.7671,
+      "step": 20543
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5538293570247074e-07,
+      "loss": 1.0047,
+      "step": 20544
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5512528385734157e-07,
+      "loss": 0.9495,
+      "step": 20545
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5486772200560513e-07,
+      "loss": 1.0679,
+      "step": 20546
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5461025015225415e-07,
+      "loss": 0.8499,
+      "step": 20547
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.543528683022751e-07,
+      "loss": 0.9657,
+      "step": 20548
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.540955764606563e-07,
+      "loss": 0.9968,
+      "step": 20549
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.538383746323809e-07,
+      "loss": 0.9661,
+      "step": 20550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.535812628224339e-07,
+      "loss": 0.9634,
+      "step": 20551
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.533242410357951e-07,
+      "loss": 0.8924,
+      "step": 20552
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.530673092774456e-07,
+      "loss": 0.9006,
+      "step": 20553
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5281046755236246e-07,
+      "loss": 0.9574,
+      "step": 20554
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.525537158655229e-07,
+      "loss": 1.0116,
+      "step": 20555
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.522970542218995e-07,
+      "loss": 0.7697,
+      "step": 20556
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.520404826264678e-07,
+      "loss": 1.0113,
+      "step": 20557
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.517840010841972e-07,
+      "loss": 0.8107,
+      "step": 20558
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.515276096000575e-07,
+      "loss": 1.0328,
+      "step": 20559
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.512713081790148e-07,
+      "loss": 1.1077,
+      "step": 20560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.510150968260373e-07,
+      "loss": 1.0421,
+      "step": 20561
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5075897554608663e-07,
+      "loss": 0.9441,
+      "step": 20562
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5050294434412715e-07,
+      "loss": 0.986,
+      "step": 20563
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.502470032251176e-07,
+      "loss": 1.0665,
+      "step": 20564
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.499911521940186e-07,
+      "loss": 0.7573,
+      "step": 20565
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.497353912557855e-07,
+      "loss": 1.1848,
+      "step": 20566
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.494797204153738e-07,
+      "loss": 0.9302,
+      "step": 20567
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4922413967773905e-07,
+      "loss": 0.9848,
+      "step": 20568
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.489686490478306e-07,
+      "loss": 0.9318,
+      "step": 20569
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4871324853059995e-07,
+      "loss": 0.9629,
+      "step": 20570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4845793813099436e-07,
+      "loss": 0.7844,
+      "step": 20571
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.482027178539621e-07,
+      "loss": 0.8033,
+      "step": 20572
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4794758770444636e-07,
+      "loss": 0.9394,
+      "step": 20573
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4769254768739105e-07,
+      "loss": 0.945,
+      "step": 20574
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4743759780773714e-07,
+      "loss": 1.0192,
+      "step": 20575
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.471827380704246e-07,
+      "loss": 0.7617,
+      "step": 20576
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4692796848039054e-07,
+      "loss": 1.0253,
+      "step": 20577
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.466732890425728e-07,
+      "loss": 1.0564,
+      "step": 20578
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.464186997619023e-07,
+      "loss": 0.9888,
+      "step": 20579
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.461642006433158e-07,
+      "loss": 0.9986,
+      "step": 20580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4590979169174146e-07,
+      "loss": 0.8773,
+      "step": 20581
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4565547291210977e-07,
+      "loss": 0.8345,
+      "step": 20582
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4540124430934683e-07,
+      "loss": 1.0309,
+      "step": 20583
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.451471058883798e-07,
+      "loss": 1.0484,
+      "step": 20584
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4489305765413097e-07,
+      "loss": 0.8219,
+      "step": 20585
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4463909961152397e-07,
+      "loss": 0.855,
+      "step": 20586
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.443852317654772e-07,
+      "loss": 0.9305,
+      "step": 20587
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4413145412091176e-07,
+      "loss": 0.8621,
+      "step": 20588
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.438777666827425e-07,
+      "loss": 0.8371,
+      "step": 20589
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.436241694558856e-07,
+      "loss": 0.9731,
+      "step": 20590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.433706624452537e-07,
+      "loss": 0.8532,
+      "step": 20591
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.431172456557591e-07,
+      "loss": 0.8567,
+      "step": 20592
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.428639190923122e-07,
+      "loss": 0.7467,
+      "step": 20593
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.426106827598197e-07,
+      "loss": 0.9868,
+      "step": 20594
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.423575366631893e-07,
+      "loss": 0.8263,
+      "step": 20595
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.421044808073243e-07,
+      "loss": 0.993,
+      "step": 20596
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4185151519712966e-07,
+      "loss": 1.0574,
+      "step": 20597
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4159863983750375e-07,
+      "loss": 1.0964,
+      "step": 20598
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.413458547333487e-07,
+      "loss": 0.771,
+      "step": 20599
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.410931598895606e-07,
+      "loss": 0.8374,
+      "step": 20600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4084055531103556e-07,
+      "loss": 0.9002,
+      "step": 20601
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.405880410026668e-07,
+      "loss": 1.0255,
+      "step": 20602
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4033561696934945e-07,
+      "loss": 0.9727,
+      "step": 20603
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.400832832159717e-07,
+      "loss": 0.8067,
+      "step": 20604
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3983103974742415e-07,
+      "loss": 0.937,
+      "step": 20605
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3957888656859226e-07,
+      "loss": 0.8935,
+      "step": 20606
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.393268236843633e-07,
+      "loss": 0.8136,
+      "step": 20607
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3907485109961833e-07,
+      "loss": 1.1453,
+      "step": 20608
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3882296881924236e-07,
+      "loss": 0.9332,
+      "step": 20609
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3857117684811315e-07,
+      "loss": 0.8871,
+      "step": 20610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.383194751911106e-07,
+      "loss": 0.8292,
+      "step": 20611
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.380678638531104e-07,
+      "loss": 0.928,
+      "step": 20612
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.37816342838988e-07,
+      "loss": 0.8983,
+      "step": 20613
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.375649121536151e-07,
+      "loss": 0.9562,
+      "step": 20614
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3731357180186606e-07,
+      "loss": 0.9461,
+      "step": 20615
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.370623217886082e-07,
+      "loss": 0.8646,
+      "step": 20616
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.368111621187109e-07,
+      "loss": 1.093,
+      "step": 20617
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.365600927970386e-07,
+      "loss": 0.9764,
+      "step": 20618
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3630911382845745e-07,
+      "loss": 0.9739,
+      "step": 20619
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3605822521782847e-07,
+      "loss": 0.9928,
+      "step": 20620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.35807426970014e-07,
+      "loss": 0.96,
+      "step": 20621
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.355567190898723e-07,
+      "loss": 0.9736,
+      "step": 20622
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.353061015822617e-07,
+      "loss": 1.0271,
+      "step": 20623
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.350555744520362e-07,
+      "loss": 0.8386,
+      "step": 20624
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.348051377040512e-07,
+      "loss": 1.134,
+      "step": 20625
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3455479134315904e-07,
+      "loss": 0.8412,
+      "step": 20626
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.343045353742086e-07,
+      "loss": 0.8037,
+      "step": 20627
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3405436980205044e-07,
+      "loss": 0.811,
+      "step": 20628
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.338042946315295e-07,
+      "loss": 0.8343,
+      "step": 20629
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.335543098674926e-07,
+      "loss": 1.0697,
+      "step": 20630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3330441551478187e-07,
+      "loss": 0.8782,
+      "step": 20631
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3305461157824015e-07,
+      "loss": 0.7851,
+      "step": 20632
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.328048980627058e-07,
+      "loss": 0.8786,
+      "step": 20633
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3255527497301874e-07,
+      "loss": 0.9499,
+      "step": 20634
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.323057423140136e-07,
+      "loss": 0.8354,
+      "step": 20635
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.320563000905269e-07,
+      "loss": 0.9097,
+      "step": 20636
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.318069483073888e-07,
+      "loss": 1.0974,
+      "step": 20637
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3155768696943366e-07,
+      "loss": 0.9277,
+      "step": 20638
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3130851608148884e-07,
+      "loss": 1.0051,
+      "step": 20639
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3105943564838314e-07,
+      "loss": 0.997,
+      "step": 20640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.308104456749406e-07,
+      "loss": 0.9181,
+      "step": 20641
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3056154616598777e-07,
+      "loss": 0.9381,
+      "step": 20642
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3031273712634484e-07,
+      "loss": 0.9856,
+      "step": 20643
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.300640185608345e-07,
+      "loss": 1.0814,
+      "step": 20644
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2981539047427347e-07,
+      "loss": 1.073,
+      "step": 20645
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2956685287148014e-07,
+      "loss": 0.9278,
+      "step": 20646
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.293184057572696e-07,
+      "loss": 0.8479,
+      "step": 20647
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2907004913645567e-07,
+      "loss": 0.8912,
+      "step": 20648
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2882178301384847e-07,
+      "loss": 0.7409,
+      "step": 20649
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.285736073942614e-07,
+      "loss": 0.8473,
+      "step": 20650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2832552228250045e-07,
+      "loss": 0.9738,
+      "step": 20651
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.280775276833731e-07,
+      "loss": 0.8099,
+      "step": 20652
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.278296236016837e-07,
+      "loss": 0.9825,
+      "step": 20653
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.275818100422357e-07,
+      "loss": 0.9582,
+      "step": 20654
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.273340870098296e-07,
+      "loss": 0.8361,
+      "step": 20655
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.270864545092667e-07,
+      "loss": 0.9549,
+      "step": 20656
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2683891254534306e-07,
+      "loss": 1.0223,
+      "step": 20657
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.26591461122856e-07,
+      "loss": 0.9841,
+      "step": 20658
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2634410024659834e-07,
+      "loss": 0.8844,
+      "step": 20659
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.260968299213635e-07,
+      "loss": 0.9302,
+      "step": 20660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.258496501519431e-07,
+      "loss": 0.9605,
+      "step": 20661
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2560256094312516e-07,
+      "loss": 0.8847,
+      "step": 20662
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.253555622996979e-07,
+      "loss": 0.8476,
+      "step": 20663
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2510865422644533e-07,
+      "loss": 1.1996,
+      "step": 20664
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.248618367281531e-07,
+      "loss": 0.8342,
+      "step": 20665
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.246151098096012e-07,
+      "loss": 0.9083,
+      "step": 20666
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.24368473475572e-07,
+      "loss": 1.094,
+      "step": 20667
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.241219277308422e-07,
+      "loss": 0.9264,
+      "step": 20668
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2387547258018956e-07,
+      "loss": 1.0624,
+      "step": 20669
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2362910802838877e-07,
+      "loss": 0.9773,
+      "step": 20670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.233828340802142e-07,
+      "loss": 0.9781,
+      "step": 20671
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.231366507404343e-07,
+      "loss": 0.9323,
+      "step": 20672
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2289055801382306e-07,
+      "loss": 0.8239,
+      "step": 20673
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.22644555905145e-07,
+      "loss": 0.9559,
+      "step": 20674
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.223986444191685e-07,
+      "loss": 1.0474,
+      "step": 20675
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.221528235606569e-07,
+      "loss": 0.8417,
+      "step": 20676
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2190709333437377e-07,
+      "loss": 0.8194,
+      "step": 20677
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.216614537450785e-07,
+      "loss": 0.9045,
+      "step": 20678
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.214159047975324e-07,
+      "loss": 0.8722,
+      "step": 20679
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.21170446496491e-07,
+      "loss": 0.9213,
+      "step": 20680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.209250788467111e-07,
+      "loss": 0.903,
+      "step": 20681
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2067980185294614e-07,
+      "loss": 0.8202,
+      "step": 20682
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2043461551994946e-07,
+      "loss": 0.9274,
+      "step": 20683
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.201895198524696e-07,
+      "loss": 0.9001,
+      "step": 20684
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.199445148552566e-07,
+      "loss": 1.1281,
+      "step": 20685
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.196996005330577e-07,
+      "loss": 0.8858,
+      "step": 20686
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.194547768906164e-07,
+      "loss": 0.8219,
+      "step": 20687
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1921004393267783e-07,
+      "loss": 0.9775,
+      "step": 20688
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1896540166398203e-07,
+      "loss": 0.9284,
+      "step": 20689
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1872085008927076e-07,
+      "loss": 0.7877,
+      "step": 20690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1847638921328026e-07,
+      "loss": 1.0145,
+      "step": 20691
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.182320190407484e-07,
+      "loss": 0.8719,
+      "step": 20692
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1798773957640863e-07,
+      "loss": 1.0999,
+      "step": 20693
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1774355082499484e-07,
+      "loss": 0.8024,
+      "step": 20694
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.174994527912362e-07,
+      "loss": 0.9074,
+      "step": 20695
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.172554454798649e-07,
+      "loss": 0.8923,
+      "step": 20696
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.170115288956066e-07,
+      "loss": 0.8528,
+      "step": 20697
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.167677030431887e-07,
+      "loss": 1.0133,
+      "step": 20698
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1652396792733296e-07,
+      "loss": 0.9832,
+      "step": 20699
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1628032355276394e-07,
+      "loss": 1.0135,
+      "step": 20700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.160367699242001e-07,
+      "loss": 0.8944,
+      "step": 20701
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.157933070463626e-07,
+      "loss": 0.9512,
+      "step": 20702
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.15549934923966e-07,
+      "loss": 0.9118,
+      "step": 20703
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1530665356172776e-07,
+      "loss": 0.9359,
+      "step": 20704
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1506346296436006e-07,
+      "loss": 1.0311,
+      "step": 20705
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.148203631365754e-07,
+      "loss": 0.947,
+      "step": 20706
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1457735408308153e-07,
+      "loss": 0.8187,
+      "step": 20707
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.143344358085909e-07,
+      "loss": 0.9025,
+      "step": 20708
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.140916083178064e-07,
+      "loss": 0.9393,
+      "step": 20709
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1384887161543477e-07,
+      "loss": 0.9967,
+      "step": 20710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.136062257061778e-07,
+      "loss": 0.9603,
+      "step": 20711
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1336367059473796e-07,
+      "loss": 0.909,
+      "step": 20712
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1312120628581297e-07,
+      "loss": 0.9357,
+      "step": 20713
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1287883278410203e-07,
+      "loss": 1.029,
+      "step": 20714
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.126365500943002e-07,
+      "loss": 0.9182,
+      "step": 20715
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1239435822110253e-07,
+      "loss": 1.0368,
+      "step": 20716
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1215225716919984e-07,
+      "loss": 0.9852,
+      "step": 20717
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.119102469432839e-07,
+      "loss": 0.9647,
+      "step": 20718
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.116683275480442e-07,
+      "loss": 0.9427,
+      "step": 20719
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1142649898816656e-07,
+      "loss": 0.8594,
+      "step": 20720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1118476126833773e-07,
+      "loss": 1.2413,
+      "step": 20721
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.109431143932395e-07,
+      "loss": 0.8974,
+      "step": 20722
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.10701558367556e-07,
+      "loss": 0.8987,
+      "step": 20723
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1046009319596505e-07,
+      "loss": 0.8649,
+      "step": 20724
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.102187188831468e-07,
+      "loss": 0.8765,
+      "step": 20725
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.099774354337759e-07,
+      "loss": 0.9405,
+      "step": 20726
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0973624285252965e-07,
+      "loss": 0.8315,
+      "step": 20727
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0949514114407884e-07,
+      "loss": 0.8942,
+      "step": 20728
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0925413031309635e-07,
+      "loss": 0.9497,
+      "step": 20729
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0901321036424956e-07,
+      "loss": 1.038,
+      "step": 20730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.087723813022092e-07,
+      "loss": 0.9141,
+      "step": 20731
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.085316431316393e-07,
+      "loss": 0.9896,
+      "step": 20732
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.082909958572056e-07,
+      "loss": 0.9886,
+      "step": 20733
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0805043948356816e-07,
+      "loss": 0.9679,
+      "step": 20734
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0780997401539e-07,
+      "loss": 1.023,
+      "step": 20735
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.075695994573291e-07,
+      "loss": 0.8447,
+      "step": 20736
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.073293158140428e-07,
+      "loss": 1.0211,
+      "step": 20737
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.070891230901862e-07,
+      "loss": 1.1432,
+      "step": 20738
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0684902129041405e-07,
+      "loss": 0.8436,
+      "step": 20739
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.066090104193764e-07,
+      "loss": 0.8152,
+      "step": 20740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.063690904817257e-07,
+      "loss": 0.8899,
+      "step": 20741
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.061292614821082e-07,
+      "loss": 1.1179,
+      "step": 20742
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.058895234251713e-07,
+      "loss": 0.766,
+      "step": 20743
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0564987631556076e-07,
+      "loss": 0.8364,
+      "step": 20744
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.054103201579195e-07,
+      "loss": 0.9363,
+      "step": 20745
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0517085495688714e-07,
+      "loss": 0.9761,
+      "step": 20746
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0493148071710554e-07,
+      "loss": 1.0177,
+      "step": 20747
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.04692197443211e-07,
+      "loss": 1.0246,
+      "step": 20748
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.044530051398403e-07,
+      "loss": 1.103,
+      "step": 20749
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.042139038116271e-07,
+      "loss": 0.9456,
+      "step": 20750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.039748934632053e-07,
+      "loss": 0.8716,
+      "step": 20751
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0373597409920355e-07,
+      "loss": 1.1156,
+      "step": 20752
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.034971457242525e-07,
+      "loss": 0.9051,
+      "step": 20753
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.03258408342979e-07,
+      "loss": 0.9832,
+      "step": 20754
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.030197619600078e-07,
+      "loss": 0.7699,
+      "step": 20755
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0278120657996457e-07,
+      "loss": 0.9739,
+      "step": 20756
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.025427422074689e-07,
+      "loss": 0.9033,
+      "step": 20757
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0230436884714276e-07,
+      "loss": 0.8802,
+      "step": 20758
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.020660865036029e-07,
+      "loss": 0.9348,
+      "step": 20759
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0182789518146795e-07,
+      "loss": 0.808,
+      "step": 20760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.015897948853508e-07,
+      "loss": 0.9741,
+      "step": 20761
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0135178561986665e-07,
+      "loss": 0.931,
+      "step": 20762
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0111386738962465e-07,
+      "loss": 0.789,
+      "step": 20763
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0087604019923667e-07,
+      "loss": 1.0258,
+      "step": 20764
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.006383040533084e-07,
+      "loss": 0.9411,
+      "step": 20765
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0040065895644787e-07,
+      "loss": 0.9911,
+      "step": 20766
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0016310491325805e-07,
+      "loss": 0.9632,
+      "step": 20767
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9992564192834253e-07,
+      "loss": 1.0648,
+      "step": 20768
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9968827000630084e-07,
+      "loss": 0.7968,
+      "step": 20769
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9945098915173334e-07,
+      "loss": 0.8319,
+      "step": 20770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9921379936923623e-07,
+      "loss": 0.89,
+      "step": 20771
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.989767006634059e-07,
+      "loss": 1.0335,
+      "step": 20772
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9873969303883475e-07,
+      "loss": 1.0363,
+      "step": 20773
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9850277650011694e-07,
+      "loss": 0.9775,
+      "step": 20774
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9826595105184044e-07,
+      "loss": 0.7511,
+      "step": 20775
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9802921669859486e-07,
+      "loss": 0.8863,
+      "step": 20776
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9779257344496603e-07,
+      "loss": 1.046,
+      "step": 20777
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9755602129553974e-07,
+      "loss": 1.0123,
+      "step": 20778
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.973195602548995e-07,
+      "loss": 0.8423,
+      "step": 20779
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9708319032762556e-07,
+      "loss": 0.8817,
+      "step": 20780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.968469115182981e-07,
+      "loss": 0.9735,
+      "step": 20781
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.966107238314947e-07,
+      "loss": 0.8587,
+      "step": 20782
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.963746272717921e-07,
+      "loss": 0.9452,
+      "step": 20783
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9613862184376343e-07,
+      "loss": 0.9539,
+      "step": 20784
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.959027075519827e-07,
+      "loss": 0.9394,
+      "step": 20785
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.956668844010191e-07,
+      "loss": 0.9094,
+      "step": 20786
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.954311523954434e-07,
+      "loss": 0.8998,
+      "step": 20787
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9519551153982084e-07,
+      "loss": 0.9271,
+      "step": 20788
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.949599618387189e-07,
+      "loss": 1.0464,
+      "step": 20789
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.947245032967e-07,
+      "loss": 0.8192,
+      "step": 20790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9448913591832715e-07,
+      "loss": 0.8051,
+      "step": 20791
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.94253859708159e-07,
+      "loss": 0.8157,
+      "step": 20792
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9401867467075575e-07,
+      "loss": 1.0748,
+      "step": 20793
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9378358081067204e-07,
+      "loss": 1.2062,
+      "step": 20794
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.935485781324654e-07,
+      "loss": 1.024,
+      "step": 20795
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9331366664068607e-07,
+      "loss": 1.0155,
+      "step": 20796
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9307884633988814e-07,
+      "loss": 0.9773,
+      "step": 20797
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9284411723461857e-07,
+      "loss": 0.8211,
+      "step": 20798
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.926094793294276e-07,
+      "loss": 0.9351,
+      "step": 20799
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.923749326288583e-07,
+      "loss": 0.8856,
+      "step": 20800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9214047713745863e-07,
+      "loss": 1.0255,
+      "step": 20801
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9190611285976836e-07,
+      "loss": 0.8817,
+      "step": 20802
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9167183980032984e-07,
+      "loss": 0.9523,
+      "step": 20803
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.914376579636807e-07,
+      "loss": 0.9961,
+      "step": 20804
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.912035673543601e-07,
+      "loss": 0.937,
+      "step": 20805
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.909695679769009e-07,
+      "loss": 1.03,
+      "step": 20806
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.907356598358391e-07,
+      "loss": 0.9599,
+      "step": 20807
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9050184293570483e-07,
+      "loss": 1.0582,
+      "step": 20808
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.902681172810301e-07,
+      "loss": 0.8881,
+      "step": 20809
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9003448287634073e-07,
+      "loss": 0.8632,
+      "step": 20810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.898009397261659e-07,
+      "loss": 1.0118,
+      "step": 20811
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8956748783502965e-07,
+      "loss": 1.0933,
+      "step": 20812
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8933412720745404e-07,
+      "loss": 0.9249,
+      "step": 20813
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.891008578479615e-07,
+      "loss": 0.8872,
+      "step": 20814
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8886767976107124e-07,
+      "loss": 1.0167,
+      "step": 20815
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8863459295130127e-07,
+      "loss": 0.9478,
+      "step": 20816
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8840159742316685e-07,
+      "loss": 1.0497,
+      "step": 20817
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8816869318118335e-07,
+      "loss": 0.9638,
+      "step": 20818
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.879358802298621e-07,
+      "loss": 0.9642,
+      "step": 20819
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8770315857371446e-07,
+      "loss": 0.9297,
+      "step": 20820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.874705282172491e-07,
+      "loss": 0.8079,
+      "step": 20821
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8723798916497346e-07,
+      "loss": 0.853,
+      "step": 20822
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.870055414213918e-07,
+      "loss": 1.228,
+      "step": 20823
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8677318499100994e-07,
+      "loss": 0.9673,
+      "step": 20824
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8654091987832755e-07,
+      "loss": 0.9475,
+      "step": 20825
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8630874608784667e-07,
+      "loss": 0.7569,
+      "step": 20826
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.860766636240636e-07,
+      "loss": 0.8855,
+      "step": 20827
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.858446724914765e-07,
+      "loss": 0.9701,
+      "step": 20828
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.85612772694579e-07,
+      "loss": 0.8728,
+      "step": 20829
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.853809642378658e-07,
+      "loss": 0.8622,
+      "step": 20830
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.851492471258255e-07,
+      "loss": 1.0324,
+      "step": 20831
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.849176213629501e-07,
+      "loss": 0.9905,
+      "step": 20832
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8468608695372546e-07,
+      "loss": 1.0072,
+      "step": 20833
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.844546439026391e-07,
+      "loss": 0.6518,
+      "step": 20834
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8422329221417357e-07,
+      "loss": 0.9493,
+      "step": 20835
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.839920318928119e-07,
+      "loss": 1.0707,
+      "step": 20836
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.837608629430355e-07,
+      "loss": 0.8786,
+      "step": 20837
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8352978536932194e-07,
+      "loss": 0.867,
+      "step": 20838
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.832987991761499e-07,
+      "loss": 0.8863,
+      "step": 20839
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.830679043679929e-07,
+      "loss": 0.8858,
+      "step": 20840
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.828371009493258e-07,
+      "loss": 0.9465,
+      "step": 20841
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.826063889246189e-07,
+      "loss": 0.924,
+      "step": 20842
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8237576829834355e-07,
+      "loss": 0.9643,
+      "step": 20843
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.821452390749685e-07,
+      "loss": 0.9691,
+      "step": 20844
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.819148012589579e-07,
+      "loss": 0.9478,
+      "step": 20845
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8168445485477814e-07,
+      "loss": 0.9561,
+      "step": 20846
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.814541998668924e-07,
+      "loss": 0.8064,
+      "step": 20847
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8122403629976036e-07,
+      "loss": 0.9961,
+      "step": 20848
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8099396415784296e-07,
+      "loss": 1.0817,
+      "step": 20849
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8076398344559664e-07,
+      "loss": 1.1159,
+      "step": 20850
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8053409416747833e-07,
+      "loss": 0.8759,
+      "step": 20851
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.803042963279401e-07,
+      "loss": 0.9703,
+      "step": 20852
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8007458993143667e-07,
+      "loss": 0.8434,
+      "step": 20853
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.798449749824167e-07,
+      "loss": 1.0296,
+      "step": 20854
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7961545148533053e-07,
+      "loss": 0.8426,
+      "step": 20855
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.793860194446235e-07,
+      "loss": 1.0468,
+      "step": 20856
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7915667886474153e-07,
+      "loss": 0.9271,
+      "step": 20857
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.789274297501271e-07,
+      "loss": 1.0461,
+      "step": 20858
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.786982721052245e-07,
+      "loss": 0.9165,
+      "step": 20859
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.784692059344707e-07,
+      "loss": 0.8045,
+      "step": 20860
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7824023124230603e-07,
+      "loss": 1.0698,
+      "step": 20861
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.780113480331653e-07,
+      "loss": 0.8648,
+      "step": 20862
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7778255631148387e-07,
+      "loss": 0.8849,
+      "step": 20863
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7755385608169374e-07,
+      "loss": 1.0319,
+      "step": 20864
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7732524734822686e-07,
+      "loss": 1.0199,
+      "step": 20865
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7709673011551143e-07,
+      "loss": 1.0024,
+      "step": 20866
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.768683043879766e-07,
+      "loss": 0.8537,
+      "step": 20867
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7663997017004617e-07,
+      "loss": 0.8734,
+      "step": 20868
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.764117274661454e-07,
+      "loss": 0.9868,
+      "step": 20869
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7618357628069526e-07,
+      "loss": 1.0628,
+      "step": 20870
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.759555166181166e-07,
+      "loss": 0.9773,
+      "step": 20871
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.757275484828292e-07,
+      "loss": 1.0551,
+      "step": 20872
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7549967187924796e-07,
+      "loss": 0.9538,
+      "step": 20873
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.752718868117893e-07,
+      "loss": 0.9614,
+      "step": 20874
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.750441932848663e-07,
+      "loss": 0.8996,
+      "step": 20875
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7481659130288997e-07,
+      "loss": 0.7517,
+      "step": 20876
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7458908087027005e-07,
+      "loss": 0.8861,
+      "step": 20877
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7436166199141577e-07,
+      "loss": 0.8008,
+      "step": 20878
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.741343346707315e-07,
+      "loss": 0.9646,
+      "step": 20879
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.73907098912623e-07,
+      "loss": 1.1033,
+      "step": 20880
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7367995472149133e-07,
+      "loss": 0.9446,
+      "step": 20881
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7345290210173956e-07,
+      "loss": 0.9095,
+      "step": 20882
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.732259410577648e-07,
+      "loss": 0.9551,
+      "step": 20883
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.729990715939662e-07,
+      "loss": 0.9435,
+      "step": 20884
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7277229371473755e-07,
+      "loss": 1.0978,
+      "step": 20885
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7254560742447366e-07,
+      "loss": 0.859,
+      "step": 20886
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.72319012727566e-07,
+      "loss": 1.0141,
+      "step": 20887
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7209250962840606e-07,
+      "loss": 0.8075,
+      "step": 20888
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.718660981313803e-07,
+      "loss": 1.0774,
+      "step": 20889
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7163977824087696e-07,
+      "loss": 1.0524,
+      "step": 20890
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7141354996127967e-07,
+      "loss": 0.9412,
+      "step": 20891
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7118741329697274e-07,
+      "loss": 0.9442,
+      "step": 20892
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.70961368252336e-07,
+      "loss": 0.7752,
+      "step": 20893
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.707354148317509e-07,
+      "loss": 0.9141,
+      "step": 20894
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7050955303959405e-07,
+      "loss": 0.8839,
+      "step": 20895
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.702837828802429e-07,
+      "loss": 0.9799,
+      "step": 20896
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.700581043580691e-07,
+      "loss": 0.7451,
+      "step": 20897
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6983251747744743e-07,
+      "loss": 0.8833,
+      "step": 20898
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.696070222427477e-07,
+      "loss": 1.059,
+      "step": 20899
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6938161865833866e-07,
+      "loss": 0.9085,
+      "step": 20900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.691563067285874e-07,
+      "loss": 0.9659,
+      "step": 20901
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6893108645786036e-07,
+      "loss": 1.0475,
+      "step": 20902
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.687059578505197e-07,
+      "loss": 0.9729,
+      "step": 20903
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.684809209109279e-07,
+      "loss": 0.938,
+      "step": 20904
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.682559756434455e-07,
+      "loss": 1.0121,
+      "step": 20905
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6803112205242953e-07,
+      "loss": 0.8933,
+      "step": 20906
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.678063601422376e-07,
+      "loss": 0.901,
+      "step": 20907
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.67581689917224e-07,
+      "loss": 0.9007,
+      "step": 20908
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.673571113817419e-07,
+      "loss": 0.8989,
+      "step": 20909
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6713262454014124e-07,
+      "loss": 0.999,
+      "step": 20910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.669082293967734e-07,
+      "loss": 1.0636,
+      "step": 20911
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6668392595598447e-07,
+      "loss": 0.9634,
+      "step": 20912
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6645971422212146e-07,
+      "loss": 0.9131,
+      "step": 20913
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.662355941995265e-07,
+      "loss": 1.0365,
+      "step": 20914
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.660115658925444e-07,
+      "loss": 0.8926,
+      "step": 20915
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6578762930551285e-07,
+      "loss": 0.9806,
+      "step": 20916
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.655637844427733e-07,
+      "loss": 0.9077,
+      "step": 20917
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.653400313086607e-07,
+      "loss": 0.8479,
+      "step": 20918
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.651163699075121e-07,
+      "loss": 0.9916,
+      "step": 20919
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.648928002436585e-07,
+      "loss": 1.0276,
+      "step": 20920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6466932232143416e-07,
+      "loss": 0.8442,
+      "step": 20921
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6444593614516624e-07,
+      "loss": 0.8226,
+      "step": 20922
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.642226417191851e-07,
+      "loss": 0.75,
+      "step": 20923
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6399943904781566e-07,
+      "loss": 1.0568,
+      "step": 20924
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.637763281353828e-07,
+      "loss": 0.9877,
+      "step": 20925
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.635533089862091e-07,
+      "loss": 1.0429,
+      "step": 20926
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.633303816046162e-07,
+      "loss": 0.8537,
+      "step": 20927
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6310754599492174e-07,
+      "loss": 0.848,
+      "step": 20928
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.628848021614444e-07,
+      "loss": 0.9757,
+      "step": 20929
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.626621501084997e-07,
+      "loss": 1.2137,
+      "step": 20930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6243958984040084e-07,
+      "loss": 0.7589,
+      "step": 20931
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6221712136146105e-07,
+      "loss": 0.7848,
+      "step": 20932
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.619947446759891e-07,
+      "loss": 1.0145,
+      "step": 20933
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.617724597882948e-07,
+      "loss": 0.9991,
+      "step": 20934
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.615502667026837e-07,
+      "loss": 0.8851,
+      "step": 20935
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.613281654234617e-07,
+      "loss": 0.9056,
+      "step": 20936
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6110615595493104e-07,
+      "loss": 1.0882,
+      "step": 20937
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.608842383013943e-07,
+      "loss": 0.9551,
+      "step": 20938
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6066241246714917e-07,
+      "loss": 1.08,
+      "step": 20939
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.604406784564961e-07,
+      "loss": 0.8792,
+      "step": 20940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.602190362737295e-07,
+      "loss": 0.8741,
+      "step": 20941
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5999748592314413e-07,
+      "loss": 1.0547,
+      "step": 20942
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5977602740903163e-07,
+      "loss": 0.832,
+      "step": 20943
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5955466073568415e-07,
+      "loss": 0.7601,
+      "step": 20944
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.593333859073888e-07,
+      "loss": 1.2025,
+      "step": 20945
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.591122029284343e-07,
+      "loss": 0.8909,
+      "step": 20946
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.588911118031051e-07,
+      "loss": 0.9323,
+      "step": 20947
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.586701125356861e-07,
+      "loss": 0.8976,
+      "step": 20948
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.584492051304571e-07,
+      "loss": 0.9047,
+      "step": 20949
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.582283895916998e-07,
+      "loss": 0.8243,
+      "step": 20950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5800766592369077e-07,
+      "loss": 0.9861,
+      "step": 20951
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.577870341307087e-07,
+      "loss": 0.9725,
+      "step": 20952
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.57566494217027e-07,
+      "loss": 0.9175,
+      "step": 20953
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.573460461869187e-07,
+      "loss": 0.9443,
+      "step": 20954
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.571256900446545e-07,
+      "loss": 0.88,
+      "step": 20955
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5690542579450474e-07,
+      "loss": 1.063,
+      "step": 20956
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.56685253440736e-07,
+      "loss": 0.8478,
+      "step": 20957
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.56465172987615e-07,
+      "loss": 0.962,
+      "step": 20958
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5624518443940483e-07,
+      "loss": 1.0063,
+      "step": 20959
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5602528780036826e-07,
+      "loss": 0.982,
+      "step": 20960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5580548307476525e-07,
+      "loss": 0.7399,
+      "step": 20961
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.555857702668557e-07,
+      "loss": 0.847,
+      "step": 20962
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.553661493808951e-07,
+      "loss": 0.9894,
+      "step": 20963
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5514662042113893e-07,
+      "loss": 0.9293,
+      "step": 20964
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5492718339184095e-07,
+      "loss": 0.9053,
+      "step": 20965
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.547078382972523e-07,
+      "loss": 1.0888,
+      "step": 20966
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5448858514162343e-07,
+      "loss": 0.9799,
+      "step": 20967
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5426942392920095e-07,
+      "loss": 1.0539,
+      "step": 20968
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5405035466423253e-07,
+      "loss": 0.9903,
+      "step": 20969
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5383137735096155e-07,
+      "loss": 1.0066,
+      "step": 20970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5361249199363115e-07,
+      "loss": 1.0479,
+      "step": 20971
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5339369859648186e-07,
+      "loss": 1.1129,
+      "step": 20972
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.531749971637532e-07,
+      "loss": 0.8058,
+      "step": 20973
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.52956387699681e-07,
+      "loss": 0.9052,
+      "step": 20974
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.527378702085037e-07,
+      "loss": 0.9667,
+      "step": 20975
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5251944469445233e-07,
+      "loss": 0.8992,
+      "step": 20976
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.523011111617607e-07,
+      "loss": 0.9307,
+      "step": 20977
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5208286961465713e-07,
+      "loss": 0.8379,
+      "step": 20978
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.518647200573715e-07,
+      "loss": 1.0445,
+      "step": 20979
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.516466624941294e-07,
+      "loss": 0.7707,
+      "step": 20980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.514286969291568e-07,
+      "loss": 0.8147,
+      "step": 20981
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.512108233666755e-07,
+      "loss": 0.9069,
+      "step": 20982
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.50993041810908e-07,
+      "loss": 0.9676,
+      "step": 20983
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.507753522660722e-07,
+      "loss": 1.049,
+      "step": 20984
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.505577547363869e-07,
+      "loss": 1.006,
+      "step": 20985
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.50340249226067e-07,
+      "loss": 0.8517,
+      "step": 20986
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.501228357393287e-07,
+      "loss": 0.9956,
+      "step": 20987
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.499055142803819e-07,
+      "loss": 0.9612,
+      "step": 20988
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.496882848534393e-07,
+      "loss": 0.8504,
+      "step": 20989
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4947114746270816e-07,
+      "loss": 1.0382,
+      "step": 20990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.492541021123962e-07,
+      "loss": 0.8562,
+      "step": 20991
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.490371488067078e-07,
+      "loss": 1.0534,
+      "step": 20992
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.488202875498474e-07,
+      "loss": 0.8043,
+      "step": 20993
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4860351834601613e-07,
+      "loss": 0.8145,
+      "step": 20994
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4838684119941394e-07,
+      "loss": 1.0528,
+      "step": 20995
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.481702561142385e-07,
+      "loss": 0.8678,
+      "step": 20996
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4795376309468657e-07,
+      "loss": 0.955,
+      "step": 20997
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4773736214495305e-07,
+      "loss": 0.8201,
+      "step": 20998
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4752105326922904e-07,
+      "loss": 0.9791,
+      "step": 20999
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.473048364717079e-07,
+      "loss": 0.9561,
+      "step": 21000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4708871175657624e-07,
+      "loss": 0.9705,
+      "step": 21001
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.46872679128024e-07,
+      "loss": 1.0145,
+      "step": 21002
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.466567385902341e-07,
+      "loss": 1.0433,
+      "step": 21003
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.464408901473925e-07,
+      "loss": 1.002,
+      "step": 21004
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4622513380367977e-07,
+      "loss": 0.9153,
+      "step": 21005
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4600946956327754e-07,
+      "loss": 0.8528,
+      "step": 21006
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4579389743036255e-07,
+      "loss": 0.8354,
+      "step": 21007
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4557841740911303e-07,
+      "loss": 0.8535,
+      "step": 21008
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.453630295037024e-07,
+      "loss": 0.764,
+      "step": 21009
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4514773371830504e-07,
+      "loss": 0.9708,
+      "step": 21010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.449325300570915e-07,
+      "loss": 1.0536,
+      "step": 21011
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.447174185242324e-07,
+      "loss": 0.97,
+      "step": 21012
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4450239912389373e-07,
+      "loss": 0.9297,
+      "step": 21013
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.442874718602428e-07,
+      "loss": 0.9167,
+      "step": 21014
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4407263673744295e-07,
+      "loss": 1.0111,
+      "step": 21015
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4385789375965805e-07,
+      "loss": 0.9796,
+      "step": 21016
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.436432429310465e-07,
+      "loss": 0.8708,
+      "step": 21017
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.434286842557687e-07,
+      "loss": 0.9876,
+      "step": 21018
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.432142177379804e-07,
+      "loss": 0.8768,
+      "step": 21019
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.429998433818387e-07,
+      "loss": 0.8406,
+      "step": 21020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.427855611914948e-07,
+      "loss": 1.0088,
+      "step": 21021
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4257137117110196e-07,
+      "loss": 0.8818,
+      "step": 21022
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.423572733248103e-07,
+      "loss": 0.9572,
+      "step": 21023
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4214326765676644e-07,
+      "loss": 0.9833,
+      "step": 21024
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4192935417111816e-07,
+      "loss": 0.9203,
+      "step": 21025
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.417155328720089e-07,
+      "loss": 0.8199,
+      "step": 21026
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.415018037635819e-07,
+      "loss": 0.9326,
+      "step": 21027
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.412881668499778e-07,
+      "loss": 1.0105,
+      "step": 21028
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.410746221353366e-07,
+      "loss": 0.945,
+      "step": 21029
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4086116962379504e-07,
+      "loss": 1.2514,
+      "step": 21030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.406478093194886e-07,
+      "loss": 1.1744,
+      "step": 21031
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.404345412265502e-07,
+      "loss": 0.9878,
+      "step": 21032
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.402213653491142e-07,
+      "loss": 0.9189,
+      "step": 21033
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4000828169130906e-07,
+      "loss": 0.7807,
+      "step": 21034
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.397952902572642e-07,
+      "loss": 0.9103,
+      "step": 21035
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.395823910511053e-07,
+      "loss": 0.9024,
+      "step": 21036
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.393695840769583e-07,
+      "loss": 0.7886,
+      "step": 21037
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.391568693389446e-07,
+      "loss": 0.9888,
+      "step": 21038
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.389442468411879e-07,
+      "loss": 1.0644,
+      "step": 21039
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.387317165878056e-07,
+      "loss": 0.8638,
+      "step": 21040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3851927858291714e-07,
+      "loss": 1.024,
+      "step": 21041
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3830693283063644e-07,
+      "loss": 0.937,
+      "step": 21042
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3809467933507969e-07,
+      "loss": 0.9448,
+      "step": 21043
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.378825181003569e-07,
+      "loss": 0.958,
+      "step": 21044
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3767044913058146e-07,
+      "loss": 0.9698,
+      "step": 21045
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.374584724298601e-07,
+      "loss": 1.0262,
+      "step": 21046
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3724658800230173e-07,
+      "loss": 0.9412,
+      "step": 21047
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3703479585200916e-07,
+      "loss": 0.8798,
+      "step": 21048
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.368230959830875e-07,
+      "loss": 0.9613,
+      "step": 21049
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3661148839963731e-07,
+      "loss": 0.7649,
+      "step": 21050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.363999731057598e-07,
+      "loss": 0.9642,
+      "step": 21051
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3618855010555164e-07,
+      "loss": 0.9589,
+      "step": 21052
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3597721940311068e-07,
+      "loss": 0.8695,
+      "step": 21053
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3576598100252922e-07,
+      "loss": 0.8268,
+      "step": 21054
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3555483490790177e-07,
+      "loss": 0.9515,
+      "step": 21055
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3534378112331835e-07,
+      "loss": 0.7782,
+      "step": 21056
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3513281965286848e-07,
+      "loss": 0.8321,
+      "step": 21057
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3492195050063948e-07,
+      "loss": 0.8773,
+      "step": 21058
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3471117367071638e-07,
+      "loss": 0.9138,
+      "step": 21059
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3450048916718426e-07,
+      "loss": 0.7809,
+      "step": 21060
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.342898969941232e-07,
+      "loss": 0.7988,
+      "step": 21061
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3407939715561546e-07,
+      "loss": 1.1096,
+      "step": 21062
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3386898965573723e-07,
+      "loss": 0.8673,
+      "step": 21063
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3365867449856695e-07,
+      "loss": 0.8307,
+      "step": 21064
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.33448451688178e-07,
+      "loss": 0.9652,
+      "step": 21065
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3323832122864433e-07,
+      "loss": 0.9615,
+      "step": 21066
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3302828312403602e-07,
+      "loss": 1.1631,
+      "step": 21067
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3281833737842485e-07,
+      "loss": 0.7957,
+      "step": 21068
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3260848399587587e-07,
+      "loss": 1.0748,
+      "step": 21069
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.323987229804575e-07,
+      "loss": 1.182,
+      "step": 21070
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3218905433623095e-07,
+      "loss": 0.9515,
+      "step": 21071
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.319794780672613e-07,
+      "loss": 0.7881,
+      "step": 21072
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3176999417760637e-07,
+      "loss": 1.0049,
+      "step": 21073
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.315606026713274e-07,
+      "loss": 0.9022,
+      "step": 21074
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.313513035524789e-07,
+      "loss": 1.1495,
+      "step": 21075
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3114209682511824e-07,
+      "loss": 0.8357,
+      "step": 21076
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.309329824932971e-07,
+      "loss": 0.827,
+      "step": 21077
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3072396056106783e-07,
+      "loss": 1.073,
+      "step": 21078
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3051503103247885e-07,
+      "loss": 0.937,
+      "step": 21079
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.303061939115808e-07,
+      "loss": 0.8986,
+      "step": 21080
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3009744920241717e-07,
+      "loss": 0.924,
+      "step": 21081
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2988879690903465e-07,
+      "loss": 1.0018,
+      "step": 21082
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2968023703547394e-07,
+      "loss": 0.9991,
+      "step": 21083
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2947176958577677e-07,
+      "loss": 0.9191,
+      "step": 21084
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2926339456398162e-07,
+      "loss": 0.9052,
+      "step": 21085
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2905511197412634e-07,
+      "loss": 0.961,
+      "step": 21086
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2884692182024605e-07,
+      "loss": 0.9747,
+      "step": 21087
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2863882410637417e-07,
+      "loss": 0.7015,
+      "step": 21088
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2843081883654304e-07,
+      "loss": 0.9786,
+      "step": 21089
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.282229060147817e-07,
+      "loss": 0.8345,
+      "step": 21090
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.280150856451202e-07,
+      "loss": 0.9486,
+      "step": 21091
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.278073577315826e-07,
+      "loss": 0.8514,
+      "step": 21092
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2759972227819616e-07,
+      "loss": 0.9836,
+      "step": 21093
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2739217928898217e-07,
+      "loss": 0.9612,
+      "step": 21094
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.271847287679624e-07,
+      "loss": 0.8555,
+      "step": 21095
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2697737071915472e-07,
+      "loss": 1.0286,
+      "step": 21096
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2677010514657926e-07,
+      "loss": 1.0866,
+      "step": 21097
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2656293205424895e-07,
+      "loss": 0.8871,
+      "step": 21098
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2635585144617999e-07,
+      "loss": 0.9926,
+      "step": 21099
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2614886332638308e-07,
+      "loss": 0.8932,
+      "step": 21100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2594196769886945e-07,
+      "loss": 0.9339,
+      "step": 21101
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2573516456764587e-07,
+      "loss": 1.0942,
+      "step": 21102
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2552845393672197e-07,
+      "loss": 0.9841,
+      "step": 21103
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2532183581010058e-07,
+      "loss": 0.9435,
+      "step": 21104
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2511531019178634e-07,
+      "loss": 0.9407,
+      "step": 21105
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.249088770857788e-07,
+      "loss": 0.946,
+      "step": 21106
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2470253649607976e-07,
+      "loss": 0.8174,
+      "step": 21107
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2449628842668492e-07,
+      "loss": 0.8948,
+      "step": 21108
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.242901328815922e-07,
+      "loss": 0.9427,
+      "step": 21109
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2408406986479337e-07,
+      "loss": 0.9935,
+      "step": 21110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.238780993802836e-07,
+      "loss": 0.9418,
+      "step": 21111
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2367222143205136e-07,
+      "loss": 0.9953,
+      "step": 21112
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2346643602408736e-07,
+      "loss": 0.8018,
+      "step": 21113
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.232607431603767e-07,
+      "loss": 0.9159,
+      "step": 21114
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.230551428449057e-07,
+      "loss": 0.9168,
+      "step": 21115
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2284963508165836e-07,
+      "loss": 1.0245,
+      "step": 21116
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2264421987461538e-07,
+      "loss": 0.8847,
+      "step": 21117
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.224388972277569e-07,
+      "loss": 0.9912,
+      "step": 21118
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.222336671450609e-07,
+      "loss": 0.9252,
+      "step": 21119
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.220285296305047e-07,
+      "loss": 0.8618,
+      "step": 21120
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.218234846880607e-07,
+      "loss": 0.8613,
+      "step": 21121
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.216185323217035e-07,
+      "loss": 1.065,
+      "step": 21122
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2141367253540325e-07,
+      "loss": 0.9948,
+      "step": 21123
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.21208905333129e-07,
+      "loss": 0.8386,
+      "step": 21124
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2100423071884759e-07,
+      "loss": 0.9762,
+      "step": 21125
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2079964869652582e-07,
+      "loss": 1.0226,
+      "step": 21126
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2059515927012663e-07,
+      "loss": 0.8666,
+      "step": 21127
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2039076244361245e-07,
+      "loss": 0.7911,
+      "step": 21128
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2018645822094286e-07,
+      "loss": 0.8312,
+      "step": 21129
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1998224660607636e-07,
+      "loss": 0.8951,
+      "step": 21130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1977812760296923e-07,
+      "loss": 0.9906,
+      "step": 21131
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1957410121557722e-07,
+      "loss": 1.0345,
+      "step": 21132
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.193701674478521e-07,
+      "loss": 0.9975,
+      "step": 21133
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1916632630374579e-07,
+      "loss": 0.9106,
+      "step": 21134
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.189625777872073e-07,
+      "loss": 0.9909,
+      "step": 21135
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1875892190218462e-07,
+      "loss": 1.1385,
+      "step": 21136
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1855535865262178e-07,
+      "loss": 0.8884,
+      "step": 21137
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1835188804246622e-07,
+      "loss": 0.9131,
+      "step": 21138
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1814851007565696e-07,
+      "loss": 1.0534,
+      "step": 21139
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1794522475613644e-07,
+      "loss": 0.9309,
+      "step": 21140
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.177420320878415e-07,
+      "loss": 0.9593,
+      "step": 21141
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.175389320747112e-07,
+      "loss": 0.8825,
+      "step": 21142
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.17335924720678e-07,
+      "loss": 0.9417,
+      "step": 21143
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1713301002967702e-07,
+      "loss": 0.9001,
+      "step": 21144
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1693018800563903e-07,
+      "loss": 1.094,
+      "step": 21145
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1672745865249367e-07,
+      "loss": 0.9663,
+      "step": 21146
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1652482197416836e-07,
+      "loss": 0.8907,
+      "step": 21147
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1632227797458938e-07,
+      "loss": 1.0029,
+      "step": 21148
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1611982665768194e-07,
+      "loss": 0.8231,
+      "step": 21149
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1591746802736736e-07,
+      "loss": 0.9886,
+      "step": 21150
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1571520208756692e-07,
+      "loss": 0.9963,
+      "step": 21151
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1551302884219861e-07,
+      "loss": 1.0759,
+      "step": 21152
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.153109482951804e-07,
+      "loss": 0.7825,
+      "step": 21153
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1510896045042696e-07,
+      "loss": 0.8489,
+      "step": 21154
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1490706531185234e-07,
+      "loss": 0.7913,
+      "step": 21155
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1470526288336734e-07,
+      "loss": 0.9581,
+      "step": 21156
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1450355316888327e-07,
+      "loss": 0.7786,
+      "step": 21157
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1430193617230645e-07,
+      "loss": 0.9733,
+      "step": 21158
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1410041189754483e-07,
+      "loss": 0.8993,
+      "step": 21159
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1389898034850086e-07,
+      "loss": 0.899,
+      "step": 21160
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.136976415290798e-07,
+      "loss": 0.9593,
+      "step": 21161
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1349639544318013e-07,
+      "loss": 0.8764,
+      "step": 21162
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1329524209470322e-07,
+      "loss": 1.0049,
+      "step": 21163
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1309418148754423e-07,
+      "loss": 0.9679,
+      "step": 21164
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1289321362560067e-07,
+      "loss": 0.7956,
+      "step": 21165
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1269233851276383e-07,
+      "loss": 1.1222,
+      "step": 21166
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1249155615292838e-07,
+      "loss": 0.9558,
+      "step": 21167
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1229086654998177e-07,
+      "loss": 1.0601,
+      "step": 21168
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1209026970781476e-07,
+      "loss": 0.9918,
+      "step": 21169
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1188976563031204e-07,
+      "loss": 1.027,
+      "step": 21170
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.116893543213594e-07,
+      "loss": 1.0945,
+      "step": 21171
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1148903578483814e-07,
+      "loss": 0.883,
+      "step": 21172
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.112888100246313e-07,
+      "loss": 1.069,
+      "step": 21173
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1108867704461745e-07,
+      "loss": 1.0028,
+      "step": 21174
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.108886368486751e-07,
+      "loss": 0.8806,
+      "step": 21175
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.106886894406779e-07,
+      "loss": 0.9727,
+      "step": 21176
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1048883482450156e-07,
+      "loss": 0.9572,
+      "step": 21177
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1028907300401746e-07,
+      "loss": 0.7329,
+      "step": 21178
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.100894039830964e-07,
+      "loss": 0.8603,
+      "step": 21179
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0988982776560586e-07,
+      "loss": 0.8272,
+      "step": 21180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.096903443554138e-07,
+      "loss": 0.8586,
+      "step": 21181
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.094909537563844e-07,
+      "loss": 0.9191,
+      "step": 21182
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.092916559723812e-07,
+      "loss": 0.9918,
+      "step": 21183
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0909245100726616e-07,
+      "loss": 0.9845,
+      "step": 21184
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0889333886489727e-07,
+      "loss": 0.7994,
+      "step": 21185
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0869431954913365e-07,
+      "loss": 0.9613,
+      "step": 21186
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0849539306383003e-07,
+      "loss": 0.9619,
+      "step": 21187
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.082965594128422e-07,
+      "loss": 0.884,
+      "step": 21188
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0809781860002098e-07,
+      "loss": 1.0259,
+      "step": 21189
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0789917062921772e-07,
+      "loss": 1.0245,
+      "step": 21190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.07700615504281e-07,
+      "loss": 0.9833,
+      "step": 21191
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0750215322905775e-07,
+      "loss": 0.9592,
+      "step": 21192
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0730378380739326e-07,
+      "loss": 1.0225,
+      "step": 21193
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0710550724313105e-07,
+      "loss": 0.9819,
+      "step": 21194
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.069073235401109e-07,
+      "loss": 0.9419,
+      "step": 21195
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.067092327021758e-07,
+      "loss": 0.905,
+      "step": 21196
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0651123473316103e-07,
+      "loss": 0.938,
+      "step": 21197
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0631332963690464e-07,
+      "loss": 0.8078,
+      "step": 21198
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0611551741723912e-07,
+      "loss": 0.9512,
+      "step": 21199
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0591779807799917e-07,
+      "loss": 0.7875,
+      "step": 21200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0572017162301338e-07,
+      "loss": 1.0197,
+      "step": 21201
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0552263805611261e-07,
+      "loss": 0.8137,
+      "step": 21202
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.053251973811221e-07,
+      "loss": 0.7449,
+      "step": 21203
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.051278496018694e-07,
+      "loss": 0.8501,
+      "step": 21204
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0493059472217635e-07,
+      "loss": 0.8439,
+      "step": 21205
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0473343274586555e-07,
+      "loss": 0.6736,
+      "step": 21206
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0453636367675666e-07,
+      "loss": 0.9808,
+      "step": 21207
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0433938751866722e-07,
+      "loss": 0.8831,
+      "step": 21208
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0414250427541526e-07,
+      "loss": 0.8829,
+      "step": 21209
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0394571395081387e-07,
+      "loss": 0.7784,
+      "step": 21210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0374901654867717e-07,
+      "loss": 0.9792,
+      "step": 21211
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.035524120728144e-07,
+      "loss": 0.9413,
+      "step": 21212
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0335590052703636e-07,
+      "loss": 1.0478,
+      "step": 21213
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.031594819151489e-07,
+      "loss": 0.8988,
+      "step": 21214
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0296315624095898e-07,
+      "loss": 1.2375,
+      "step": 21215
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.027669235082691e-07,
+      "loss": 0.9712,
+      "step": 21216
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.025707837208829e-07,
+      "loss": 0.8341,
+      "step": 21217
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.023747368825979e-07,
+      "loss": 0.8871,
+      "step": 21218
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.021787829972155e-07,
+      "loss": 0.7915,
+      "step": 21219
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0198292206852989e-07,
+      "loss": 0.7539,
+      "step": 21220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.01787154100338e-07,
+      "loss": 0.9615,
+      "step": 21221
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0159147909643074e-07,
+      "loss": 0.9408,
+      "step": 21222
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0139589706060058e-07,
+      "loss": 0.7522,
+      "step": 21223
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0120040799663564e-07,
+      "loss": 0.8009,
+      "step": 21224
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.010050119083251e-07,
+      "loss": 0.9757,
+      "step": 21225
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0080970879945316e-07,
+      "loss": 1.012,
+      "step": 21226
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.006144986738051e-07,
+      "loss": 0.9395,
+      "step": 21227
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0041938153516182e-07,
+      "loss": 0.9853,
+      "step": 21228
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.002243573873053e-07,
+      "loss": 0.9979,
+      "step": 21229
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0002942623401134e-07,
+      "loss": 0.9827,
+      "step": 21230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9983458807905976e-07,
+      "loss": 1.0961,
+      "step": 21231
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9963984292622363e-07,
+      "loss": 0.8152,
+      "step": 21232
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9944519077927772e-07,
+      "loss": 1.043,
+      "step": 21233
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.992506316419912e-07,
+      "loss": 1.0072,
+      "step": 21234
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.99056165518135e-07,
+      "loss": 0.9427,
+      "step": 21235
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9886179241147663e-07,
+      "loss": 0.9001,
+      "step": 21236
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9866751232578197e-07,
+      "loss": 1.105,
+      "step": 21237
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9847332526481465e-07,
+      "loss": 0.9943,
+      "step": 21238
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.982792312323384e-07,
+      "loss": 0.8903,
+      "step": 21239
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9808523023211234e-07,
+      "loss": 0.7737,
+      "step": 21240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.978913222678952e-07,
+      "loss": 0.7295,
+      "step": 21241
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9769750734344506e-07,
+      "loss": 0.9166,
+      "step": 21242
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9750378546251613e-07,
+      "loss": 0.9723,
+      "step": 21243
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.973101566288621e-07,
+      "loss": 0.8832,
+      "step": 21244
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.971166208462344e-07,
+      "loss": 0.9104,
+      "step": 21245
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.969231781183828e-07,
+      "loss": 0.7964,
+      "step": 21246
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.967298284490543e-07,
+      "loss": 0.9325,
+      "step": 21247
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9653657184199703e-07,
+      "loss": 0.9441,
+      "step": 21248
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.96343408300953e-07,
+      "loss": 0.9478,
+      "step": 21249
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9615033782966641e-07,
+      "loss": 0.8164,
+      "step": 21250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9595736043187708e-07,
+      "loss": 0.9857,
+      "step": 21251
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.957644761113242e-07,
+      "loss": 1.0014,
+      "step": 21252
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9557168487174428e-07,
+      "loss": 0.9606,
+      "step": 21253
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9537898671687373e-07,
+      "loss": 1.0331,
+      "step": 21254
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9518638165044511e-07,
+      "loss": 0.8775,
+      "step": 21255
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9499386967619104e-07,
+      "loss": 0.81,
+      "step": 21256
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9480145079784018e-07,
+      "loss": 0.8322,
+      "step": 21257
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9460912501912176e-07,
+      "loss": 1.0192,
+      "step": 21258
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9441689234376117e-07,
+      "loss": 0.9077,
+      "step": 21259
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9422475277548368e-07,
+      "loss": 0.9014,
+      "step": 21260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9403270631801086e-07,
+      "loss": 1.0451,
+      "step": 21261
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9384075297506467e-07,
+      "loss": 0.9479,
+      "step": 21262
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9364889275036269e-07,
+      "loss": 0.8078,
+      "step": 21263
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.934571256476242e-07,
+      "loss": 0.8647,
+      "step": 21264
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9326545167056287e-07,
+      "loss": 0.9577,
+      "step": 21265
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9307387082289352e-07,
+      "loss": 0.973,
+      "step": 21266
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9288238310832763e-07,
+      "loss": 1.0121,
+      "step": 21267
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9269098853057445e-07,
+      "loss": 0.8457,
+      "step": 21268
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.924996870933432e-07,
+      "loss": 1.1134,
+      "step": 21269
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9230847880034043e-07,
+      "loss": 0.9224,
+      "step": 21270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.921173636552698e-07,
+      "loss": 0.8583,
+      "step": 21271
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.91926341661835e-07,
+      "loss": 0.8522,
+      "step": 21272
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9173541282373642e-07,
+      "loss": 0.9964,
+      "step": 21273
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.915445771446739e-07,
+      "loss": 0.9934,
+      "step": 21274
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.913538346283439e-07,
+      "loss": 1.0317,
+      "step": 21275
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.911631852784429e-07,
+      "loss": 0.9206,
+      "step": 21276
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9097262909866465e-07,
+      "loss": 0.9026,
+      "step": 21277
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9078216609270007e-07,
+      "loss": 0.7743,
+      "step": 21278
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9059179626424118e-07,
+      "loss": 0.9913,
+      "step": 21279
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9040151961697394e-07,
+      "loss": 0.9983,
+      "step": 21280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9021133615458765e-07,
+      "loss": 1.0587,
+      "step": 21281
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9002124588076486e-07,
+      "loss": 0.968,
+      "step": 21282
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.898312487991899e-07,
+      "loss": 1.0594,
+      "step": 21283
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8964134491354258e-07,
+      "loss": 0.6677,
+      "step": 21284
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8945153422750384e-07,
+      "loss": 0.8458,
+      "step": 21285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8926181674474963e-07,
+      "loss": 0.886,
+      "step": 21286
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8907219246895703e-07,
+      "loss": 1.0045,
+      "step": 21287
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8888266140379808e-07,
+      "loss": 0.9815,
+      "step": 21288
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8869322355294816e-07,
+      "loss": 0.9526,
+      "step": 21289
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8850387892007438e-07,
+      "loss": 1.069,
+      "step": 21290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.883146275088471e-07,
+      "loss": 1.036,
+      "step": 21291
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.881254693229323e-07,
+      "loss": 0.9763,
+      "step": 21292
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8793640436599537e-07,
+      "loss": 0.7062,
+      "step": 21293
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8774743264169837e-07,
+      "loss": 0.975,
+      "step": 21294
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8755855415370394e-07,
+      "loss": 0.9928,
+      "step": 21295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8736976890567028e-07,
+      "loss": 0.8848,
+      "step": 21296
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8718107690125608e-07,
+      "loss": 0.9571,
+      "step": 21297
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8699247814411625e-07,
+      "loss": 0.9087,
+      "step": 21298
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8680397263790616e-07,
+      "loss": 0.7682,
+      "step": 21299
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8661556038627627e-07,
+      "loss": 1.0036,
+      "step": 21300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.864272413928786e-07,
+      "loss": 0.902,
+      "step": 21301
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8623901566136137e-07,
+      "loss": 0.845,
+      "step": 21302
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8605088319537112e-07,
+      "loss": 0.86,
+      "step": 21303
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8586284399855326e-07,
+      "loss": 0.874,
+      "step": 21304
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8567489807455041e-07,
+      "loss": 0.8492,
+      "step": 21305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.854870454270047e-07,
+      "loss": 0.8841,
+      "step": 21306
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8529928605955484e-07,
+      "loss": 1.0108,
+      "step": 21307
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8511161997584015e-07,
+      "loss": 0.8999,
+      "step": 21308
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8492404717949441e-07,
+      "loss": 0.9164,
+      "step": 21309
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8473656767415414e-07,
+      "loss": 0.892,
+      "step": 21310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8454918146344925e-07,
+      "loss": 0.8348,
+      "step": 21311
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8436188855101233e-07,
+      "loss": 0.9146,
+      "step": 21312
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8417468894047164e-07,
+      "loss": 1.0582,
+      "step": 21313
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8398758263545368e-07,
+      "loss": 0.9007,
+      "step": 21314
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8380056963958393e-07,
+      "loss": 0.7865,
+      "step": 21315
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.836136499564861e-07,
+      "loss": 0.9773,
+      "step": 21316
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.834268235897807e-07,
+      "loss": 0.9099,
+      "step": 21317
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8324009054308812e-07,
+      "loss": 0.9959,
+      "step": 21318
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8305345082002603e-07,
+      "loss": 1.1129,
+      "step": 21319
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.828669044242115e-07,
+      "loss": 0.8947,
+      "step": 21320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8268045135925673e-07,
+      "loss": 0.8969,
+      "step": 21321
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8249409162877652e-07,
+      "loss": 0.841,
+      "step": 21322
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8230782523637912e-07,
+      "loss": 0.8961,
+      "step": 21323
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8212165218567614e-07,
+      "loss": 0.9547,
+      "step": 21324
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8193557248027237e-07,
+      "loss": 0.9505,
+      "step": 21325
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8174958612377502e-07,
+      "loss": 0.9394,
+      "step": 21326
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8156369311978562e-07,
+      "loss": 0.9654,
+      "step": 21327
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8137789347190737e-07,
+      "loss": 0.918,
+      "step": 21328
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8119218718373853e-07,
+      "loss": 1.0217,
+      "step": 21329
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8100657425887901e-07,
+      "loss": 0.8214,
+      "step": 21330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8082105470092258e-07,
+      "loss": 0.7389,
+      "step": 21331
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.806356285134664e-07,
+      "loss": 0.9791,
+      "step": 21332
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8045029570010086e-07,
+      "loss": 1.0044,
+      "step": 21333
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8026505626441758e-07,
+      "loss": 0.9932,
+      "step": 21334
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8007991021000594e-07,
+      "loss": 0.7843,
+      "step": 21335
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7989485754045244e-07,
+      "loss": 0.8638,
+      "step": 21336
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7970989825934316e-07,
+      "loss": 1.134,
+      "step": 21337
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7952503237026076e-07,
+      "loss": 0.9194,
+      "step": 21338
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7934025987678795e-07,
+      "loss": 1.0045,
+      "step": 21339
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7915558078250294e-07,
+      "loss": 0.9105,
+      "step": 21340
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7897099509098624e-07,
+      "loss": 0.8512,
+      "step": 21341
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.787865028058122e-07,
+      "loss": 1.2753,
+      "step": 21342
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7860210393055633e-07,
+      "loss": 0.9993,
+      "step": 21343
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7841779846879071e-07,
+      "loss": 0.9252,
+      "step": 21344
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7823358642408694e-07,
+      "loss": 1.0332,
+      "step": 21345
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7804946780001275e-07,
+      "loss": 0.9591,
+      "step": 21346
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7786544260013695e-07,
+      "loss": 0.8513,
+      "step": 21347
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7768151082802443e-07,
+      "loss": 0.8996,
+      "step": 21348
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7749767248723905e-07,
+      "loss": 0.8704,
+      "step": 21349
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7731392758134182e-07,
+      "loss": 1.0318,
+      "step": 21350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7713027611389376e-07,
+      "loss": 0.86,
+      "step": 21351
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.769467180884521e-07,
+      "loss": 1.0703,
+      "step": 21352
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7676325350857393e-07,
+      "loss": 0.8303,
+      "step": 21353
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7657988237781365e-07,
+      "loss": 1.0038,
+      "step": 21354
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7639660469972397e-07,
+      "loss": 0.9197,
+      "step": 21355
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7621342047785595e-07,
+      "loss": 0.9255,
+      "step": 21356
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7603032971575896e-07,
+      "loss": 1.0222,
+      "step": 21357
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.758473324169796e-07,
+      "loss": 0.9941,
+      "step": 21358
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7566442858506339e-07,
+      "loss": 0.7955,
+      "step": 21359
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7548161822355582e-07,
+      "loss": 0.8293,
+      "step": 21360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7529890133599682e-07,
+      "loss": 0.8866,
+      "step": 21361
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7511627792592744e-07,
+      "loss": 0.882,
+      "step": 21362
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7493374799688489e-07,
+      "loss": 1.0793,
+      "step": 21363
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7475131155240743e-07,
+      "loss": 0.996,
+      "step": 21364
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7456896859602833e-07,
+      "loss": 0.858,
+      "step": 21365
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7438671913128148e-07,
+      "loss": 0.8529,
+      "step": 21366
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.742045631616962e-07,
+      "loss": 0.9362,
+      "step": 21367
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7402250069080308e-07,
+      "loss": 1.0284,
+      "step": 21368
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7384053172212978e-07,
+      "loss": 0.9057,
+      "step": 21369
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.736586562592013e-07,
+      "loss": 1.0893,
+      "step": 21370
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.734768743055415e-07,
+      "loss": 0.8066,
+      "step": 21371
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7329518586467254e-07,
+      "loss": 0.7555,
+      "step": 21372
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7311359094011437e-07,
+      "loss": 1.0559,
+      "step": 21373
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7293208953538587e-07,
+      "loss": 1.0919,
+      "step": 21374
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7275068165400255e-07,
+      "loss": 0.8181,
+      "step": 21375
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7256936729948049e-07,
+      "loss": 0.7744,
+      "step": 21376
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7238814647533074e-07,
+      "loss": 1.0048,
+      "step": 21377
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7220701918506666e-07,
+      "loss": 0.9657,
+      "step": 21378
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7202598543219595e-07,
+      "loss": 0.8881,
+      "step": 21379
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7184504522022694e-07,
+      "loss": 0.8817,
+      "step": 21380
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7166419855266404e-07,
+      "loss": 1.0248,
+      "step": 21381
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7148344543301277e-07,
+      "loss": 0.9997,
+      "step": 21382
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7130278586477423e-07,
+      "loss": 0.83,
+      "step": 21383
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.711222198514495e-07,
+      "loss": 0.8002,
+      "step": 21384
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.709417473965358e-07,
+      "loss": 1.0931,
+      "step": 21385
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7076136850353032e-07,
+      "loss": 0.9088,
+      "step": 21386
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7058108317592802e-07,
+      "loss": 0.9438,
+      "step": 21387
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7040089141722227e-07,
+      "loss": 0.85,
+      "step": 21388
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7022079323090302e-07,
+      "loss": 1.0424,
+      "step": 21389
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.700407886204608e-07,
+      "loss": 0.814,
+      "step": 21390
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6986087758938174e-07,
+      "loss": 1.1784,
+      "step": 21391
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.696810601411536e-07,
+      "loss": 0.9244,
+      "step": 21392
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.695013362792586e-07,
+      "loss": 0.8635,
+      "step": 21393
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6932170600717946e-07,
+      "loss": 0.8283,
+      "step": 21394
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.691421693283968e-07,
+      "loss": 1.0324,
+      "step": 21395
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6896272624638832e-07,
+      "loss": 0.9824,
+      "step": 21396
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6878337676463186e-07,
+      "loss": 0.9682,
+      "step": 21397
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6860412088660127e-07,
+      "loss": 0.9804,
+      "step": 21398
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.684249586157699e-07,
+      "loss": 0.9153,
+      "step": 21399
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6824588995560886e-07,
+      "loss": 0.8495,
+      "step": 21400
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.680669149095876e-07,
+      "loss": 0.7839,
+      "step": 21401
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.678880334811739e-07,
+      "loss": 0.9654,
+      "step": 21402
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6770924567383385e-07,
+      "loss": 0.6866,
+      "step": 21403
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.675305514910297e-07,
+      "loss": 1.0001,
+      "step": 21404
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6735195093622648e-07,
+      "loss": 1.0284,
+      "step": 21405
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6717344401288195e-07,
+      "loss": 0.7268,
+      "step": 21406
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.669950307244561e-07,
+      "loss": 0.8807,
+      "step": 21407
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.668167110744051e-07,
+      "loss": 0.8575,
+      "step": 21408
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.666384850661845e-07,
+      "loss": 0.8258,
+      "step": 21409
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.66460352703246e-07,
+      "loss": 1.0406,
+      "step": 21410
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6628231398904237e-07,
+      "loss": 0.942,
+      "step": 21411
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6610436892702198e-07,
+      "loss": 0.8453,
+      "step": 21412
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6592651752063372e-07,
+      "loss": 0.8611,
+      "step": 21413
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6574875977332206e-07,
+      "loss": 0.9189,
+      "step": 21414
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6557109568853203e-07,
+      "loss": 1.0293,
+      "step": 21415
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6539352526970421e-07,
+      "loss": 0.8434,
+      "step": 21416
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6521604852028083e-07,
+      "loss": 0.7803,
+      "step": 21417
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6503866544369974e-07,
+      "loss": 0.9683,
+      "step": 21418
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6486137604339813e-07,
+      "loss": 0.8415,
+      "step": 21419
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6468418032281051e-07,
+      "loss": 0.9562,
+      "step": 21420
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6450707828537027e-07,
+      "loss": 0.8894,
+      "step": 21421
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6433006993450795e-07,
+      "loss": 0.81,
+      "step": 21422
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6415315527365417e-07,
+      "loss": 0.8919,
+      "step": 21423
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6397633430623505e-07,
+      "loss": 0.9751,
+      "step": 21424
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6379960703567843e-07,
+      "loss": 1.0224,
+      "step": 21425
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6362297346540711e-07,
+      "loss": 0.9557,
+      "step": 21426
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6344643359884338e-07,
+      "loss": 0.8332,
+      "step": 21427
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6326998743940836e-07,
+      "loss": 0.8751,
+      "step": 21428
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6309363499051934e-07,
+      "loss": 0.9909,
+      "step": 21429
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6291737625559523e-07,
+      "loss": 0.996,
+      "step": 21430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6274121123804833e-07,
+      "loss": 1.0273,
+      "step": 21431
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.625651399412942e-07,
+      "loss": 0.8035,
+      "step": 21432
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6238916236874292e-07,
+      "loss": 1.0345,
+      "step": 21433
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6221327852380454e-07,
+      "loss": 0.9449,
+      "step": 21434
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6203748840988576e-07,
+      "loss": 0.9304,
+      "step": 21435
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6186179203039388e-07,
+      "loss": 0.8541,
+      "step": 21436
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.616861893887317e-07,
+      "loss": 1.0104,
+      "step": 21437
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6151068048830266e-07,
+      "loss": 0.8404,
+      "step": 21438
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6133526533250566e-07,
+      "loss": 0.8761,
+      "step": 21439
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.611599439247413e-07,
+      "loss": 0.9564,
+      "step": 21440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6098471626840474e-07,
+      "loss": 0.7726,
+      "step": 21441
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.608095823668926e-07,
+      "loss": 0.9368,
+      "step": 21442
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.606345422235961e-07,
+      "loss": 0.9004,
+      "step": 21443
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6045959584190806e-07,
+      "loss": 0.8411,
+      "step": 21444
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.602847432252175e-07,
+      "loss": 0.901,
+      "step": 21445
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6010998437691216e-07,
+      "loss": 0.8121,
+      "step": 21446
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5993531930037776e-07,
+      "loss": 0.9544,
+      "step": 21447
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5976074799899932e-07,
+      "loss": 0.6784,
+      "step": 21448
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5958627047615805e-07,
+      "loss": 0.891,
+      "step": 21449
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5941188673523512e-07,
+      "loss": 1.039,
+      "step": 21450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.592375967796078e-07,
+      "loss": 0.9261,
+      "step": 21451
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5906340061265458e-07,
+      "loss": 0.8932,
+      "step": 21452
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5888929823775046e-07,
+      "loss": 0.9242,
+      "step": 21453
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5871528965826723e-07,
+      "loss": 0.9832,
+      "step": 21454
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.585413748775777e-07,
+      "loss": 0.9404,
+      "step": 21455
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5836755389905035e-07,
+      "loss": 0.9651,
+      "step": 21456
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5819382672605356e-07,
+      "loss": 0.9684,
+      "step": 21457
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5802019336195295e-07,
+      "loss": 0.9763,
+      "step": 21458
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.578466538101131e-07,
+      "loss": 0.9474,
+      "step": 21459
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5767320807389518e-07,
+      "loss": 0.9255,
+      "step": 21460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5749985615666153e-07,
+      "loss": 0.8969,
+      "step": 21461
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.573265980617683e-07,
+      "loss": 0.914,
+      "step": 21462
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.571534337925751e-07,
+      "loss": 0.9466,
+      "step": 21463
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5698036335243473e-07,
+      "loss": 0.9183,
+      "step": 21464
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5680738674470174e-07,
+      "loss": 0.8728,
+      "step": 21465
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5663450397272628e-07,
+      "loss": 0.7902,
+      "step": 21466
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5646171503985953e-07,
+      "loss": 0.913,
+      "step": 21467
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5628901994944767e-07,
+      "loss": 0.9573,
+      "step": 21468
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5611641870483806e-07,
+      "loss": 0.9374,
+      "step": 21469
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5594391130937357e-07,
+      "loss": 1.0189,
+      "step": 21470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.557714977663971e-07,
+      "loss": 0.9472,
+      "step": 21471
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5559917807924875e-07,
+      "loss": 1.0097,
+      "step": 21472
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5542695225126803e-07,
+      "loss": 1.001,
+      "step": 21473
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5525482028579009e-07,
+      "loss": 0.9848,
+      "step": 21474
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5508278218615168e-07,
+      "loss": 0.8671,
+      "step": 21475
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5491083795568517e-07,
+      "loss": 0.801,
+      "step": 21476
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5473898759772288e-07,
+      "loss": 1.0944,
+      "step": 21477
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5456723111559268e-07,
+      "loss": 0.8467,
+      "step": 21478
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5439556851262415e-07,
+      "loss": 1.0563,
+      "step": 21479
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.542239997921413e-07,
+      "loss": 0.975,
+      "step": 21480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5405252495747035e-07,
+      "loss": 0.883,
+      "step": 21481
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5388114401193143e-07,
+      "loss": 0.9663,
+      "step": 21482
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5370985695884688e-07,
+      "loss": 0.9143,
+      "step": 21483
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5353866380153405e-07,
+      "loss": 0.8519,
+      "step": 21484
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5336756454331082e-07,
+      "loss": 0.8348,
+      "step": 21485
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5319655918749064e-07,
+      "loss": 0.9544,
+      "step": 21486
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5302564773738816e-07,
+      "loss": 0.9543,
+      "step": 21487
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5285483019631452e-07,
+      "loss": 0.7696,
+      "step": 21488
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.526841065675788e-07,
+      "loss": 0.9109,
+      "step": 21489
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5251347685448947e-07,
+      "loss": 1.1344,
+      "step": 21490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5234294106035107e-07,
+      "loss": 0.949,
+      "step": 21491
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5217249918846934e-07,
+      "loss": 0.7723,
+      "step": 21492
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5200215124214491e-07,
+      "loss": 0.8444,
+      "step": 21493
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5183189722468018e-07,
+      "loss": 0.9971,
+      "step": 21494
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5166173713937194e-07,
+      "loss": 0.817,
+      "step": 21495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5149167098951811e-07,
+      "loss": 0.8661,
+      "step": 21496
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.513216987784122e-07,
+      "loss": 0.8475,
+      "step": 21497
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5115182050934985e-07,
+      "loss": 1.1686,
+      "step": 21498
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5098203618562012e-07,
+      "loss": 1.0834,
+      "step": 21499
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5081234581051484e-07,
+      "loss": 0.9981,
+      "step": 21500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5064274938731915e-07,
+      "loss": 1.0467,
+      "step": 21501
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5047324691932098e-07,
+      "loss": 0.9272,
+      "step": 21502
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5030383840980268e-07,
+      "loss": 0.9636,
+      "step": 21503
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.501345238620483e-07,
+      "loss": 1.1011,
+      "step": 21504
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.499653032793369e-07,
+      "loss": 0.8674,
+      "step": 21505
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4979617666494805e-07,
+      "loss": 0.7954,
+      "step": 21506
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.496271440221575e-07,
+      "loss": 0.6962,
+      "step": 21507
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.494582053542415e-07,
+      "loss": 1.046,
+      "step": 21508
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4928936066447076e-07,
+      "loss": 0.8826,
+      "step": 21509
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4912060995611987e-07,
+      "loss": 1.0243,
+      "step": 21510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4895195323245627e-07,
+      "loss": 0.8223,
+      "step": 21511
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4878339049674896e-07,
+      "loss": 1.0683,
+      "step": 21512
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.48614921752262e-07,
+      "loss": 0.9532,
+      "step": 21513
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.484465470022617e-07,
+      "loss": 0.8834,
+      "step": 21514
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4827826625000764e-07,
+      "loss": 0.8397,
+      "step": 21515
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4811007949876277e-07,
+      "loss": 0.8417,
+      "step": 21516
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.479419867517834e-07,
+      "loss": 1.0401,
+      "step": 21517
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4777398801232856e-07,
+      "loss": 0.873,
+      "step": 21518
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4760608328365123e-07,
+      "loss": 0.9837,
+      "step": 21519
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4743827256900545e-07,
+      "loss": 0.9046,
+      "step": 21520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.472705558716425e-07,
+      "loss": 0.9334,
+      "step": 21521
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.471029331948115e-07,
+      "loss": 0.9103,
+      "step": 21522
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4693540454176092e-07,
+      "loss": 0.938,
+      "step": 21523
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.467679699157354e-07,
+      "loss": 0.7808,
+      "step": 21524
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.466006293199801e-07,
+      "loss": 0.9082,
+      "step": 21525
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4643338275773633e-07,
+      "loss": 0.8848,
+      "step": 21526
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.462662302322454e-07,
+      "loss": 0.9345,
+      "step": 21527
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4609917174674414e-07,
+      "loss": 0.9517,
+      "step": 21528
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4593220730447165e-07,
+      "loss": 0.9295,
+      "step": 21529
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4576533690866035e-07,
+      "loss": 0.8858,
+      "step": 21530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.455985605625454e-07,
+      "loss": 0.9122,
+      "step": 21531
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4543187826935646e-07,
+      "loss": 0.8651,
+      "step": 21532
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4526529003232425e-07,
+      "loss": 0.9242,
+      "step": 21533
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4509879585467514e-07,
+      "loss": 0.8101,
+      "step": 21534
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4493239573963647e-07,
+      "loss": 1.0203,
+      "step": 21535
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.447660896904307e-07,
+      "loss": 0.7635,
+      "step": 21536
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4459987771028084e-07,
+      "loss": 0.8939,
+      "step": 21537
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4443375980240702e-07,
+      "loss": 0.8623,
+      "step": 21538
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4426773597002785e-07,
+      "loss": 0.8179,
+      "step": 21539
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4410180621635904e-07,
+      "loss": 0.8867,
+      "step": 21540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4393597054461749e-07,
+      "loss": 1.0985,
+      "step": 21541
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4377022895801396e-07,
+      "loss": 0.9836,
+      "step": 21542
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4360458145976086e-07,
+      "loss": 0.9824,
+      "step": 21543
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4343902805306732e-07,
+      "loss": 0.8179,
+      "step": 21544
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4327356874114074e-07,
+      "loss": 0.8635,
+      "step": 21545
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4310820352718802e-07,
+      "loss": 0.9284,
+      "step": 21546
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4294293241441103e-07,
+      "loss": 0.8702,
+      "step": 21547
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.427777554060139e-07,
+      "loss": 0.9226,
+      "step": 21548
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4261267250519518e-07,
+      "loss": 0.9367,
+      "step": 21549
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4244768371515505e-07,
+      "loss": 0.9421,
+      "step": 21550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4228278903908822e-07,
+      "loss": 0.9775,
+      "step": 21551
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4211798848019098e-07,
+      "loss": 1.1662,
+      "step": 21552
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4195328204165527e-07,
+      "loss": 0.9649,
+      "step": 21553
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.417886697266735e-07,
+      "loss": 1.0745,
+      "step": 21554
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.416241515384331e-07,
+      "loss": 0.9562,
+      "step": 21555
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4145972748012383e-07,
+      "loss": 0.9802,
+      "step": 21556
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4129539755492917e-07,
+      "loss": 1.0676,
+      "step": 21557
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4113116176603492e-07,
+      "loss": 0.8272,
+      "step": 21558
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.409670201166219e-07,
+      "loss": 0.7905,
+      "step": 21559
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4080297260987087e-07,
+      "loss": 0.8752,
+      "step": 21560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4063901924895985e-07,
+      "loss": 1.0793,
+      "step": 21561
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4047516003706573e-07,
+      "loss": 1.005,
+      "step": 21562
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4031139497736267e-07,
+      "loss": 0.6423,
+      "step": 21563
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4014772407302424e-07,
+      "loss": 1.0235,
+      "step": 21564
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.399841473272212e-07,
+      "loss": 0.9119,
+      "step": 21565
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3982066474312272e-07,
+      "loss": 0.9581,
+      "step": 21566
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.396572763238957e-07,
+      "loss": 0.8492,
+      "step": 21567
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.394939820727076e-07,
+      "loss": 0.964,
+      "step": 21568
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3933078199272089e-07,
+      "loss": 0.9348,
+      "step": 21569
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3916767608709747e-07,
+      "loss": 0.9345,
+      "step": 21570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3900466435899762e-07,
+      "loss": 0.994,
+      "step": 21571
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.388417468115799e-07,
+      "loss": 0.9702,
+      "step": 21572
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.386789234480007e-07,
+      "loss": 0.9989,
+      "step": 21573
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3851619427141472e-07,
+      "loss": 0.9805,
+      "step": 21574
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.383535592849744e-07,
+      "loss": 0.9594,
+      "step": 21575
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3819101849183115e-07,
+      "loss": 0.8496,
+      "step": 21576
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3802857189513407e-07,
+      "loss": 0.8177,
+      "step": 21577
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3786621949803015e-07,
+      "loss": 0.8493,
+      "step": 21578
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3770396130366625e-07,
+      "loss": 0.9733,
+      "step": 21579
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3754179731518435e-07,
+      "loss": 0.8527,
+      "step": 21580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3737972753572747e-07,
+      "loss": 0.8788,
+      "step": 21581
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3721775196843535e-07,
+      "loss": 1.0608,
+      "step": 21582
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3705587061644653e-07,
+      "loss": 0.9404,
+      "step": 21583
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3689408348289633e-07,
+      "loss": 0.876,
+      "step": 21584
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3673239057092114e-07,
+      "loss": 0.9257,
+      "step": 21585
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3657079188365174e-07,
+      "loss": 0.8926,
+      "step": 21586
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3640928742422067e-07,
+      "loss": 0.9677,
+      "step": 21587
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3624787719575595e-07,
+      "loss": 0.9066,
+      "step": 21588
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3608656120138565e-07,
+      "loss": 0.8177,
+      "step": 21589
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3592533944423393e-07,
+      "loss": 1.0134,
+      "step": 21590
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.357642119274266e-07,
+      "loss": 0.9842,
+      "step": 21591
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3560317865408346e-07,
+      "loss": 0.812,
+      "step": 21592
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.354422396273264e-07,
+      "loss": 0.9249,
+      "step": 21593
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3528139485027127e-07,
+      "loss": 0.9178,
+      "step": 21594
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3512064432603665e-07,
+      "loss": 0.868,
+      "step": 21595
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3495998805773514e-07,
+      "loss": 1.069,
+      "step": 21596
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3479942604848138e-07,
+      "loss": 0.9285,
+      "step": 21597
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3463895830138406e-07,
+      "loss": 0.9917,
+      "step": 21598
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.34478584819554e-07,
+      "loss": 0.9015,
+      "step": 21599
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3431830560609704e-07,
+      "loss": 0.8487,
+      "step": 21600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3415812066411958e-07,
+      "loss": 1.1308,
+      "step": 21601
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.339980299967242e-07,
+      "loss": 0.9435,
+      "step": 21602
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3383803360701385e-07,
+      "loss": 0.8847,
+      "step": 21603
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3367813149808728e-07,
+      "loss": 0.8028,
+      "step": 21604
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3351832367304362e-07,
+      "loss": 0.8067,
+      "step": 21605
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.333586101349782e-07,
+      "loss": 0.8386,
+      "step": 21606
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3319899088698574e-07,
+      "loss": 0.8877,
+      "step": 21607
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3303946593215878e-07,
+      "loss": 0.9965,
+      "step": 21608
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3288003527358873e-07,
+      "loss": 0.9048,
+      "step": 21609
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3272069891436312e-07,
+      "loss": 0.6817,
+      "step": 21610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.325614568575706e-07,
+      "loss": 1.0407,
+      "step": 21611
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3240230910629536e-07,
+      "loss": 0.9721,
+      "step": 21612
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3224325566362107e-07,
+      "loss": 0.9904,
+      "step": 21613
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3208429653263022e-07,
+      "loss": 0.9316,
+      "step": 21614
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3192543171640092e-07,
+      "loss": 0.8552,
+      "step": 21615
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3176666121801297e-07,
+      "loss": 0.842,
+      "step": 21616
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3160798504054106e-07,
+      "loss": 0.802,
+      "step": 21617
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3144940318706056e-07,
+      "loss": 1.0358,
+      "step": 21618
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3129091566064344e-07,
+      "loss": 0.8889,
+      "step": 21619
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.311325224643606e-07,
+      "loss": 0.8439,
+      "step": 21620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3097422360128008e-07,
+      "loss": 0.9507,
+      "step": 21621
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3081601907447007e-07,
+      "loss": 0.9593,
+      "step": 21622
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.306579088869947e-07,
+      "loss": 0.6242,
+      "step": 21623
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3049989304191822e-07,
+      "loss": 0.9055,
+      "step": 21624
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3034197154230045e-07,
+      "loss": 0.8551,
+      "step": 21625
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3018414439120385e-07,
+      "loss": 1.0078,
+      "step": 21626
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.300264115916844e-07,
+      "loss": 0.8914,
+      "step": 21627
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2986877314679847e-07,
+      "loss": 1.054,
+      "step": 21628
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.297112290596003e-07,
+      "loss": 1.0243,
+      "step": 21629
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.295537793331425e-07,
+      "loss": 0.9499,
+      "step": 21630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.293964239704748e-07,
+      "loss": 0.8128,
+      "step": 21631
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2923916297464756e-07,
+      "loss": 0.9451,
+      "step": 21632
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.290819963487061e-07,
+      "loss": 0.8888,
+      "step": 21633
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2892492409569634e-07,
+      "loss": 0.8186,
+      "step": 21634
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2876794621866028e-07,
+      "loss": 1.0295,
+      "step": 21635
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.286110627206416e-07,
+      "loss": 0.9803,
+      "step": 21636
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2845427360467732e-07,
+      "loss": 0.8387,
+      "step": 21637
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2829757887380721e-07,
+      "loss": 0.9981,
+      "step": 21638
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.281409785310661e-07,
+      "loss": 1.097,
+      "step": 21639
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2798447257948876e-07,
+      "loss": 0.9365,
+      "step": 21640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.278280610221072e-07,
+      "loss": 1.0215,
+      "step": 21641
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2767174386195124e-07,
+      "loss": 0.7991,
+      "step": 21642
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2751552110205013e-07,
+      "loss": 1.0025,
+      "step": 21643
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2735939274543086e-07,
+      "loss": 0.8778,
+      "step": 21644
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.272033587951177e-07,
+      "loss": 1.1005,
+      "step": 21645
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2704741925413434e-07,
+      "loss": 0.8084,
+      "step": 21646
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2689157412550224e-07,
+      "loss": 1.0204,
+      "step": 21647
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.26735823412239e-07,
+      "loss": 0.9195,
+      "step": 21648
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2658016711736497e-07,
+      "loss": 0.9367,
+      "step": 21649
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2642460524389445e-07,
+      "loss": 0.9371,
+      "step": 21650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.262691377948422e-07,
+      "loss": 0.8764,
+      "step": 21651
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2611376477321914e-07,
+      "loss": 0.9741,
+      "step": 21652
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2595848618203676e-07,
+      "loss": 1.0355,
+      "step": 21653
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.258033020243027e-07,
+      "loss": 1.0047,
+      "step": 21654
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.256482123030245e-07,
+      "loss": 0.9268,
+      "step": 21655
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2549321702120588e-07,
+      "loss": 0.8584,
+      "step": 21656
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.253383161818511e-07,
+      "loss": 0.9724,
+      "step": 21657
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2518350978796056e-07,
+      "loss": 0.839,
+      "step": 21658
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2502879784253351e-07,
+      "loss": 0.9409,
+      "step": 21659
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2487418034856703e-07,
+      "loss": 0.883,
+      "step": 21660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2471965730905865e-07,
+      "loss": 0.922,
+      "step": 21661
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2456522872699995e-07,
+      "loss": 0.8167,
+      "step": 21662
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2441089460538514e-07,
+      "loss": 1.1161,
+      "step": 21663
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2425665494720242e-07,
+      "loss": 0.8886,
+      "step": 21664
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2410250975544158e-07,
+      "loss": 0.9422,
+      "step": 21665
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2394845903308806e-07,
+      "loss": 0.8497,
+      "step": 21666
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2379450278312778e-07,
+      "loss": 1.0064,
+      "step": 21667
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2364064100854223e-07,
+      "loss": 1.1261,
+      "step": 21668
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2348687371231348e-07,
+      "loss": 0.7792,
+      "step": 21669
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2333320089741973e-07,
+      "loss": 0.9781,
+      "step": 21670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2317962256683913e-07,
+      "loss": 0.8912,
+      "step": 21671
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2302613872354817e-07,
+      "loss": 0.811,
+      "step": 21672
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.228727493705184e-07,
+      "loss": 0.9211,
+      "step": 21673
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2271945451072353e-07,
+      "loss": 0.976,
+      "step": 21674
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.225662541471323e-07,
+      "loss": 0.853,
+      "step": 21675
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.22413148282714e-07,
+      "loss": 0.7648,
+      "step": 21676
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2226013692043404e-07,
+      "loss": 0.9501,
+      "step": 21677
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2210722006325782e-07,
+      "loss": 1.0369,
+      "step": 21678
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2195439771414742e-07,
+      "loss": 0.6818,
+      "step": 21679
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2180166987606435e-07,
+      "loss": 0.9813,
+      "step": 21680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.216490365519668e-07,
+      "loss": 0.8535,
+      "step": 21681
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2149649774481354e-07,
+      "loss": 0.975,
+      "step": 21682
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2134405345755773e-07,
+      "loss": 0.802,
+      "step": 21683
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.211917036931548e-07,
+      "loss": 0.962,
+      "step": 21684
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2103944845455627e-07,
+      "loss": 0.911,
+      "step": 21685
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2088728774471147e-07,
+      "loss": 0.8927,
+      "step": 21686
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.207352215665686e-07,
+      "loss": 0.9597,
+      "step": 21687
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2058324992307413e-07,
+      "loss": 0.843,
+      "step": 21688
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2043137281717244e-07,
+      "loss": 0.7719,
+      "step": 21689
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2027959025180612e-07,
+      "loss": 1.0252,
+      "step": 21690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.201279022299151e-07,
+      "loss": 0.8517,
+      "step": 21691
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1997630875444033e-07,
+      "loss": 1.0046,
+      "step": 21692
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.198248098283161e-07,
+      "loss": 0.8627,
+      "step": 21693
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.196734054544807e-07,
+      "loss": 0.9982,
+      "step": 21694
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.195220956358645e-07,
+      "loss": 0.8776,
+      "step": 21695
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1937088037540122e-07,
+      "loss": 0.8385,
+      "step": 21696
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1921975967602028e-07,
+      "loss": 0.9241,
+      "step": 21697
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.190687335406493e-07,
+      "loss": 0.9633,
+      "step": 21698
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1891780197221425e-07,
+      "loss": 0.9829,
+      "step": 21699
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1876696497364004e-07,
+      "loss": 0.8835,
+      "step": 21700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1861622254784821e-07,
+      "loss": 1.0004,
+      "step": 21701
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1846557469775976e-07,
+      "loss": 0.925,
+      "step": 21702
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1831502142629348e-07,
+      "loss": 1.0065,
+      "step": 21703
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1816456273636645e-07,
+      "loss": 0.8489,
+      "step": 21704
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1801419863089358e-07,
+      "loss": 0.9035,
+      "step": 21705
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.178639291127881e-07,
+      "loss": 0.838,
+      "step": 21706
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1771375418496155e-07,
+      "loss": 0.9975,
+      "step": 21707
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1756367385032385e-07,
+      "loss": 0.8254,
+      "step": 21708
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.174136881117821e-07,
+      "loss": 1.0214,
+      "step": 21709
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1726379697224288e-07,
+      "loss": 0.9431,
+      "step": 21710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1711400043460997e-07,
+      "loss": 0.8724,
+      "step": 21711
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1696429850178548e-07,
+      "loss": 0.8965,
+      "step": 21712
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1681469117666988e-07,
+      "loss": 1.1336,
+      "step": 21713
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1666517846216197e-07,
+      "loss": 1.0375,
+      "step": 21714
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1651576036115942e-07,
+      "loss": 0.9642,
+      "step": 21715
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1636643687655491e-07,
+      "loss": 0.9932,
+      "step": 21716
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1621720801124393e-07,
+      "loss": 0.893,
+      "step": 21717
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1606807376811579e-07,
+      "loss": 0.9039,
+      "step": 21718
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1591903415006156e-07,
+      "loss": 0.9334,
+      "step": 21719
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1577008915996779e-07,
+      "loss": 0.9392,
+      "step": 21720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1562123880072107e-07,
+      "loss": 0.9573,
+      "step": 21721
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1547248307520409e-07,
+      "loss": 1.0686,
+      "step": 21722
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1532382198630066e-07,
+      "loss": 0.9524,
+      "step": 21723
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1517525553688958e-07,
+      "loss": 0.8238,
+      "step": 21724
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1502678372985021e-07,
+      "loss": 0.8819,
+      "step": 21725
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.148784065680586e-07,
+      "loss": 0.8992,
+      "step": 21726
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1473012405438966e-07,
+      "loss": 0.9335,
+      "step": 21727
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1458193619171665e-07,
+      "loss": 0.9209,
+      "step": 21728
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.144338429829106e-07,
+      "loss": 1.1839,
+      "step": 21729
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1428584443083979e-07,
+      "loss": 0.8681,
+      "step": 21730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1413794053837301e-07,
+      "loss": 1.038,
+      "step": 21731
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1399013130837522e-07,
+      "loss": 0.8315,
+      "step": 21732
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1384241674371022e-07,
+      "loss": 0.7937,
+      "step": 21733
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1369479684724072e-07,
+      "loss": 0.9791,
+      "step": 21734
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1354727162182556e-07,
+      "loss": 1.1213,
+      "step": 21735
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1339984107032354e-07,
+      "loss": 0.9756,
+      "step": 21736
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1325250519559128e-07,
+      "loss": 0.9703,
+      "step": 21737
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1310526400048316e-07,
+      "loss": 0.7587,
+      "step": 21738
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1295811748785191e-07,
+      "loss": 0.8257,
+      "step": 21739
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1281106566054912e-07,
+      "loss": 0.9082,
+      "step": 21740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1266410852142196e-07,
+      "loss": 0.8925,
+      "step": 21741
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1251724607332037e-07,
+      "loss": 0.8663,
+      "step": 21742
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1237047831908765e-07,
+      "loss": 1.0071,
+      "step": 21743
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1222380526156929e-07,
+      "loss": 0.9624,
+      "step": 21744
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1207722690360522e-07,
+      "loss": 0.8405,
+      "step": 21745
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1193074324803654e-07,
+      "loss": 1.0274,
+      "step": 21746
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.117843542977004e-07,
+      "loss": 0.8784,
+      "step": 21747
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1163806005543398e-07,
+      "loss": 1.0297,
+      "step": 21748
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1149186052407057e-07,
+      "loss": 0.8697,
+      "step": 21749
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1134575570644457e-07,
+      "loss": 0.9067,
+      "step": 21750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1119974560538483e-07,
+      "loss": 1.0914,
+      "step": 21751
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.110538302237213e-07,
+      "loss": 0.8856,
+      "step": 21752
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1090800956428004e-07,
+      "loss": 0.9302,
+      "step": 21753
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1076228362988827e-07,
+      "loss": 0.9122,
+      "step": 21754
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1061665242336705e-07,
+      "loss": 0.8523,
+      "step": 21755
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1047111594754023e-07,
+      "loss": 1.0842,
+      "step": 21756
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1032567420522555e-07,
+      "loss": 1.1074,
+      "step": 21757
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1018032719924243e-07,
+      "loss": 0.9196,
+      "step": 21758
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1003507493240584e-07,
+      "loss": 0.9315,
+      "step": 21759
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0988991740753074e-07,
+      "loss": 0.8826,
+      "step": 21760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0974485462742878e-07,
+      "loss": 0.9408,
+      "step": 21761
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.095998865949116e-07,
+      "loss": 0.9366,
+      "step": 21762
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0945501331278696e-07,
+      "loss": 0.9116,
+      "step": 21763
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0931023478386204e-07,
+      "loss": 0.9396,
+      "step": 21764
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0916555101094184e-07,
+      "loss": 0.9948,
+      "step": 21765
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0902096199682965e-07,
+      "loss": 1.0004,
+      "step": 21766
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0887646774432715e-07,
+      "loss": 0.9493,
+      "step": 21767
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0873206825623316e-07,
+      "loss": 1.1299,
+      "step": 21768
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0858776353534661e-07,
+      "loss": 1.0724,
+      "step": 21769
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.084435535844619e-07,
+      "loss": 0.8568,
+      "step": 21770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0829943840637402e-07,
+      "loss": 0.8481,
+      "step": 21771
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0815541800387463e-07,
+      "loss": 1.0245,
+      "step": 21772
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0801149237975483e-07,
+      "loss": 0.9553,
+      "step": 21773
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0786766153680239e-07,
+      "loss": 0.9863,
+      "step": 21774
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0772392547780452e-07,
+      "loss": 0.8776,
+      "step": 21775
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0758028420554456e-07,
+      "loss": 0.97,
+      "step": 21776
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0743673772280804e-07,
+      "loss": 1.0454,
+      "step": 21777
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0729328603237499e-07,
+      "loss": 0.7962,
+      "step": 21778
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0714992913702427e-07,
+      "loss": 1.0122,
+      "step": 21779
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0700666703953366e-07,
+      "loss": 0.8967,
+      "step": 21780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0686349974267984e-07,
+      "loss": 0.9624,
+      "step": 21781
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0672042724923449e-07,
+      "loss": 1.0246,
+      "step": 21782
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0657744956197202e-07,
+      "loss": 0.9084,
+      "step": 21783
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.064345666836608e-07,
+      "loss": 0.8707,
+      "step": 21784
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0629177861707029e-07,
+      "loss": 0.9179,
+      "step": 21785
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0614908536496549e-07,
+      "loss": 0.8926,
+      "step": 21786
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0600648693011307e-07,
+      "loss": 0.9915,
+      "step": 21787
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0586398331527414e-07,
+      "loss": 0.9104,
+      "step": 21788
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0572157452321097e-07,
+      "loss": 0.8697,
+      "step": 21789
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0557926055668188e-07,
+      "loss": 0.9795,
+      "step": 21790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0543704141844413e-07,
+      "loss": 0.9665,
+      "step": 21791
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0529491711125384e-07,
+      "loss": 1.0009,
+      "step": 21792
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.051528876378638e-07,
+      "loss": 1.1971,
+      "step": 21793
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0501095300102626e-07,
+      "loss": 0.8605,
+      "step": 21794
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0486911320349125e-07,
+      "loss": 0.9681,
+      "step": 21795
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0472736824800656e-07,
+      "loss": 0.9152,
+      "step": 21796
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0458571813731944e-07,
+      "loss": 0.9809,
+      "step": 21797
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.044441628741727e-07,
+      "loss": 0.9172,
+      "step": 21798
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0430270246131025e-07,
+      "loss": 0.992,
+      "step": 21799
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0416133690147212e-07,
+      "loss": 0.9543,
+      "step": 21800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.040200661973978e-07,
+      "loss": 1.0694,
+      "step": 21801
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0387889035182397e-07,
+      "loss": 0.9659,
+      "step": 21802
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0373780936748623e-07,
+      "loss": 0.8936,
+      "step": 21803
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0359682324711795e-07,
+      "loss": 0.988,
+      "step": 21804
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0345593199344972e-07,
+      "loss": 0.8365,
+      "step": 21805
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.033151356092127e-07,
+      "loss": 0.9628,
+      "step": 21806
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0317443409713412e-07,
+      "loss": 0.8336,
+      "step": 21807
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0303382745994018e-07,
+      "loss": 0.9551,
+      "step": 21808
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0289331570035476e-07,
+      "loss": 0.9504,
+      "step": 21809
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0275289882110074e-07,
+      "loss": 0.9686,
+      "step": 21810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0261257682489756e-07,
+      "loss": 0.9355,
+      "step": 21811
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0247234971446585e-07,
+      "loss": 0.9995,
+      "step": 21812
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0233221749252065e-07,
+      "loss": 0.9606,
+      "step": 21813
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0219218016177867e-07,
+      "loss": 0.8842,
+      "step": 21814
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.020522377249511e-07,
+      "loss": 0.9102,
+      "step": 21815
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.019123901847513e-07,
+      "loss": 1.0164,
+      "step": 21816
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0177263754388766e-07,
+      "loss": 0.9212,
+      "step": 21817
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0163297980506803e-07,
+      "loss": 1.0116,
+      "step": 21818
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0149341697099801e-07,
+      "loss": 0.9418,
+      "step": 21819
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0135394904438211e-07,
+      "loss": 0.8899,
+      "step": 21820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0121457602792206e-07,
+      "loss": 1.0954,
+      "step": 21821
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0107529792431903e-07,
+      "loss": 1.0592,
+      "step": 21822
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0093611473626975e-07,
+      "loss": 0.8892,
+      "step": 21823
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.007970264664726e-07,
+      "loss": 1.0205,
+      "step": 21824
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0065803311762212e-07,
+      "loss": 0.9573,
+      "step": 21825
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0051913469241003e-07,
+      "loss": 0.7986,
+      "step": 21826
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0038033119352863e-07,
+      "loss": 0.8327,
+      "step": 21827
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0024162262366688e-07,
+      "loss": 1.0692,
+      "step": 21828
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0010300898551261e-07,
+      "loss": 1.1001,
+      "step": 21829
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.996449028175092e-08,
+      "loss": 0.9501,
+      "step": 21830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.982606651506576e-08,
+      "loss": 1.0007,
+      "step": 21831
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.968773768813833e-08,
+      "loss": 0.9074,
+      "step": 21832
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.954950380365037e-08,
+      "loss": 0.9752,
+      "step": 21833
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.941136486427805e-08,
+      "loss": 0.8879,
+      "step": 21834
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.927332087269981e-08,
+      "loss": 0.8154,
+      "step": 21835
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.913537183158905e-08,
+      "loss": 0.9447,
+      "step": 21836
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.899751774361866e-08,
+      "loss": 0.8287,
+      "step": 21837
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.88597586114598e-08,
+      "loss": 0.9029,
+      "step": 21838
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.872209443778147e-08,
+      "loss": 0.987,
+      "step": 21839
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.858452522525042e-08,
+      "loss": 1.0356,
+      "step": 21840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.844705097653284e-08,
+      "loss": 0.9681,
+      "step": 21841
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.830967169429107e-08,
+      "loss": 0.8555,
+      "step": 21842
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.81723873811885e-08,
+      "loss": 1.0045,
+      "step": 21843
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.803519803988304e-08,
+      "loss": 0.8753,
+      "step": 21844
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.78981036730342e-08,
+      "loss": 0.9556,
+      "step": 21845
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.776110428329711e-08,
+      "loss": 0.9296,
+      "step": 21846
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.762419987332683e-08,
+      "loss": 0.8561,
+      "step": 21847
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.748739044577571e-08,
+      "loss": 0.963,
+      "step": 21848
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.735067600329496e-08,
+      "loss": 0.9191,
+      "step": 21849
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.721405654853188e-08,
+      "loss": 1.0472,
+      "step": 21850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.707753208413495e-08,
+      "loss": 1.1662,
+      "step": 21851
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.694110261274814e-08,
+      "loss": 0.8215,
+      "step": 21852
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.680476813701545e-08,
+      "loss": 1.084,
+      "step": 21853
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.66685286595781e-08,
+      "loss": 0.9089,
+      "step": 21854
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.65323841830762e-08,
+      "loss": 0.8614,
+      "step": 21855
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.639633471014654e-08,
+      "loss": 0.9098,
+      "step": 21856
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.626038024342532e-08,
+      "loss": 0.8002,
+      "step": 21857
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.612452078554769e-08,
+      "loss": 1.1439,
+      "step": 21858
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.598875633914428e-08,
+      "loss": 0.9271,
+      "step": 21859
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.585308690684692e-08,
+      "loss": 0.9145,
+      "step": 21860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.571751249128346e-08,
+      "loss": 1.0122,
+      "step": 21861
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.558203309508073e-08,
+      "loss": 0.8824,
+      "step": 21862
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.544664872086329e-08,
+      "loss": 0.9326,
+      "step": 21863
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.531135937125457e-08,
+      "loss": 0.9161,
+      "step": 21864
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.517616504887583e-08,
+      "loss": 0.9176,
+      "step": 21865
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.504106575634664e-08,
+      "loss": 1.1485,
+      "step": 21866
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.490606149628323e-08,
+      "loss": 0.9217,
+      "step": 21867
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.477115227130296e-08,
+      "loss": 0.9432,
+      "step": 21868
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.463633808401817e-08,
+      "loss": 0.9316,
+      "step": 21869
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.450161893704235e-08,
+      "loss": 0.9183,
+      "step": 21870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.43669948329845e-08,
+      "loss": 0.9449,
+      "step": 21871
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.423246577445366e-08,
+      "loss": 0.9374,
+      "step": 21872
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.409803176405607e-08,
+      "loss": 0.9682,
+      "step": 21873
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.396369280439633e-08,
+      "loss": 0.8985,
+      "step": 21874
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.382944889807676e-08,
+      "loss": 0.8944,
+      "step": 21875
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.369530004769923e-08,
+      "loss": 0.9382,
+      "step": 21876
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.356124625586161e-08,
+      "loss": 0.9154,
+      "step": 21877
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.342728752516239e-08,
+      "loss": 1.0837,
+      "step": 21878
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.329342385819618e-08,
+      "loss": 0.716,
+      "step": 21879
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.315965525755699e-08,
+      "loss": 0.8201,
+      "step": 21880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.302598172583665e-08,
+      "loss": 0.9177,
+      "step": 21881
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.289240326562421e-08,
+      "loss": 1.1335,
+      "step": 21882
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.27589198795087e-08,
+      "loss": 1.0194,
+      "step": 21883
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.262553157007581e-08,
+      "loss": 1.0822,
+      "step": 21884
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.249223833991072e-08,
+      "loss": 0.9903,
+      "step": 21885
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.23590401915947e-08,
+      "loss": 0.6787,
+      "step": 21886
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.222593712770956e-08,
+      "loss": 0.948,
+      "step": 21887
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.209292915083268e-08,
+      "loss": 0.8179,
+      "step": 21888
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.196001626354312e-08,
+      "loss": 0.8422,
+      "step": 21889
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.182719846841381e-08,
+      "loss": 0.796,
+      "step": 21890
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.169447576801993e-08,
+      "loss": 0.9725,
+      "step": 21891
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.156184816493164e-08,
+      "loss": 0.7301,
+      "step": 21892
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.142931566171964e-08,
+      "loss": 1.0452,
+      "step": 21893
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.12968782609508e-08,
+      "loss": 0.9478,
+      "step": 21894
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.116453596519193e-08,
+      "loss": 1.0891,
+      "step": 21895
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.103228877700598e-08,
+      "loss": 0.8811,
+      "step": 21896
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.090013669895703e-08,
+      "loss": 0.8652,
+      "step": 21897
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.076807973360302e-08,
+      "loss": 0.7413,
+      "step": 21898
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.063611788350524e-08,
+      "loss": 1.0282,
+      "step": 21899
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.050425115121775e-08,
+      "loss": 0.9544,
+      "step": 21900
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.03724795392974e-08,
+      "loss": 0.8855,
+      "step": 21901
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.024080305029659e-08,
+      "loss": 0.8888,
+      "step": 21902
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.010922168676661e-08,
+      "loss": 0.8423,
+      "step": 21903
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.997773545125599e-08,
+      "loss": 0.8634,
+      "step": 21904
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.984634434631379e-08,
+      "loss": 0.9721,
+      "step": 21905
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.971504837448409e-08,
+      "loss": 0.9388,
+      "step": 21906
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.958384753831206e-08,
+      "loss": 0.9262,
+      "step": 21907
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.945274184033903e-08,
+      "loss": 1.1397,
+      "step": 21908
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.932173128310517e-08,
+      "loss": 0.9782,
+      "step": 21909
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.919081586914902e-08,
+      "loss": 0.8944,
+      "step": 21910
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.905999560100687e-08,
+      "loss": 1.0104,
+      "step": 21911
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.89292704812128e-08,
+      "loss": 0.998,
+      "step": 21912
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.879864051230035e-08,
+      "loss": 0.926,
+      "step": 21913
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.866810569680029e-08,
+      "loss": 0.9123,
+      "step": 21914
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.853766603724167e-08,
+      "loss": 0.8698,
+      "step": 21915
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.840732153615139e-08,
+      "loss": 0.9759,
+      "step": 21916
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.827707219605519e-08,
+      "loss": 1.0826,
+      "step": 21917
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.814691801947717e-08,
+      "loss": 0.8338,
+      "step": 21918
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.801685900893753e-08,
+      "loss": 0.8513,
+      "step": 21919
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.788689516695758e-08,
+      "loss": 0.8621,
+      "step": 21920
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.775702649605477e-08,
+      "loss": 0.9061,
+      "step": 21921
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.762725299874486e-08,
+      "loss": 0.9932,
+      "step": 21922
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.749757467754305e-08,
+      "loss": 0.933,
+      "step": 21923
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.736799153496123e-08,
+      "loss": 0.9027,
+      "step": 21924
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.723850357351015e-08,
+      "loss": 0.8335,
+      "step": 21925
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.710911079569894e-08,
+      "loss": 0.9938,
+      "step": 21926
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.697981320403337e-08,
+      "loss": 0.9311,
+      "step": 21927
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.685061080102031e-08,
+      "loss": 0.8456,
+      "step": 21928
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.672150358916221e-08,
+      "loss": 0.9934,
+      "step": 21929
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.659249157096039e-08,
+      "loss": 0.9447,
+      "step": 21930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.646357474891398e-08,
+      "loss": 0.9153,
+      "step": 21931
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.633475312552208e-08,
+      "loss": 0.9528,
+      "step": 21932
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.620602670327882e-08,
+      "loss": 0.8643,
+      "step": 21933
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.607739548467997e-08,
+      "loss": 0.971,
+      "step": 21934
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.594885947221632e-08,
+      "loss": 1.1042,
+      "step": 21935
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.58204186683792e-08,
+      "loss": 0.8304,
+      "step": 21936
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.569207307565664e-08,
+      "loss": 1.1276,
+      "step": 21937
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.556382269653551e-08,
+      "loss": 0.7071,
+      "step": 21938
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.543566753349996e-08,
+      "loss": 0.8577,
+      "step": 21939
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.530760758903411e-08,
+      "loss": 0.8921,
+      "step": 21940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.51796428656182e-08,
+      "loss": 0.9881,
+      "step": 21941
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.50517733657319e-08,
+      "loss": 1.0128,
+      "step": 21942
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.492399909185267e-08,
+      "loss": 0.8058,
+      "step": 21943
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.479632004645633e-08,
+      "loss": 0.8633,
+      "step": 21944
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.466873623201587e-08,
+      "loss": 0.7804,
+      "step": 21945
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.454124765100435e-08,
+      "loss": 0.7565,
+      "step": 21946
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.441385430589034e-08,
+      "loss": 0.8266,
+      "step": 21947
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.42865561991435e-08,
+      "loss": 0.9969,
+      "step": 21948
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.41593533332291e-08,
+      "loss": 0.9592,
+      "step": 21949
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.403224571061241e-08,
+      "loss": 0.9653,
+      "step": 21950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.390523333375588e-08,
+      "loss": 1.0499,
+      "step": 21951
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.377831620511978e-08,
+      "loss": 1.081,
+      "step": 21952
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.365149432716435e-08,
+      "loss": 0.9209,
+      "step": 21953
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.352476770234542e-08,
+      "loss": 0.8549,
+      "step": 21954
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.339813633311932e-08,
+      "loss": 1.0077,
+      "step": 21955
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.327160022193859e-08,
+      "loss": 0.8449,
+      "step": 21956
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.314515937125567e-08,
+      "loss": 1.0039,
+      "step": 21957
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.301881378351972e-08,
+      "loss": 0.9041,
+      "step": 21958
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.289256346117936e-08,
+      "loss": 1.0499,
+      "step": 21959
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.276640840667927e-08,
+      "loss": 0.8603,
+      "step": 21960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.264034862246583e-08,
+      "loss": 0.8762,
+      "step": 21961
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.251438411097878e-08,
+      "loss": 0.9796,
+      "step": 21962
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.238851487466116e-08,
+      "loss": 0.9415,
+      "step": 21963
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.226274091594987e-08,
+      "loss": 0.9755,
+      "step": 21964
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.213706223728302e-08,
+      "loss": 0.7546,
+      "step": 21965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.201147884109472e-08,
+      "loss": 1.013,
+      "step": 21966
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.188599072981862e-08,
+      "loss": 0.8001,
+      "step": 21967
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.176059790588553e-08,
+      "loss": 0.8797,
+      "step": 21968
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.163530037172574e-08,
+      "loss": 0.8964,
+      "step": 21969
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.151009812976562e-08,
+      "loss": 0.9323,
+      "step": 21970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.13849911824327e-08,
+      "loss": 0.9511,
+      "step": 21971
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.125997953214893e-08,
+      "loss": 0.9621,
+      "step": 21972
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.113506318133735e-08,
+      "loss": 0.9068,
+      "step": 21973
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.101024213241826e-08,
+      "loss": 0.991,
+      "step": 21974
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.088551638780973e-08,
+      "loss": 0.8985,
+      "step": 21975
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.076088594992926e-08,
+      "loss": 0.7812,
+      "step": 21976
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.063635082118993e-08,
+      "loss": 0.8927,
+      "step": 21977
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.05119110040059e-08,
+      "loss": 0.9169,
+      "step": 21978
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.03875665007875e-08,
+      "loss": 1.0756,
+      "step": 21979
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.026331731394443e-08,
+      "loss": 0.9884,
+      "step": 21980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.013916344588257e-08,
+      "loss": 0.9662,
+      "step": 21981
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.001510489900943e-08,
+      "loss": 0.8963,
+      "step": 21982
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.989114167572698e-08,
+      "loss": 0.9552,
+      "step": 21983
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.976727377843829e-08,
+      "loss": 1.1131,
+      "step": 21984
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.964350120954146e-08,
+      "loss": 0.841,
+      "step": 21985
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.95198239714362e-08,
+      "loss": 0.9502,
+      "step": 21986
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.939624206651842e-08,
+      "loss": 1.0303,
+      "step": 21987
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.927275549718228e-08,
+      "loss": 1.1468,
+      "step": 21988
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.914936426581976e-08,
+      "loss": 0.7824,
+      "step": 21989
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.902606837482286e-08,
+      "loss": 0.9423,
+      "step": 21990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.890286782657908e-08,
+      "loss": 0.9083,
+      "step": 21991
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.877976262347653e-08,
+      "loss": 1.1321,
+      "step": 21992
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.865675276789886e-08,
+      "loss": 1.0116,
+      "step": 21993
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.853383826223137e-08,
+      "loss": 0.895,
+      "step": 21994
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.841101910885329e-08,
+      "loss": 0.9832,
+      "step": 21995
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.828829531014603e-08,
+      "loss": 1.0909,
+      "step": 21996
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.816566686848603e-08,
+      "loss": 0.9933,
+      "step": 21997
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.804313378625029e-08,
+      "loss": 0.9346,
+      "step": 21998
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.792069606581188e-08,
+      "loss": 0.8992,
+      "step": 21999
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.779835370954391e-08,
+      "loss": 0.847,
+      "step": 22000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.767610671981618e-08,
+      "loss": 1.0487,
+      "step": 22001
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.755395509899732e-08,
+      "loss": 0.9535,
+      "step": 22002
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.74318988494538e-08,
+      "loss": 0.8574,
+      "step": 22003
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.73099379735509e-08,
+      "loss": 0.9435,
+      "step": 22004
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.718807247365123e-08,
+      "loss": 0.9976,
+      "step": 22005
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.706630235211565e-08,
+      "loss": 0.9525,
+      "step": 22006
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.694462761130394e-08,
+      "loss": 0.9681,
+      "step": 22007
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.682304825357367e-08,
+      "loss": 0.8783,
+      "step": 22008
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.670156428127962e-08,
+      "loss": 0.8966,
+      "step": 22009
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.6580175696776e-08,
+      "loss": 0.8469,
+      "step": 22010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.645888250241485e-08,
+      "loss": 0.9885,
+      "step": 22011
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.633768470054593e-08,
+      "loss": 0.9172,
+      "step": 22012
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.621658229351791e-08,
+      "loss": 0.9056,
+      "step": 22013
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.609557528367617e-08,
+      "loss": 0.9739,
+      "step": 22014
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.597466367336603e-08,
+      "loss": 0.9342,
+      "step": 22015
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.585384746492952e-08,
+      "loss": 1.0353,
+      "step": 22016
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.573312666070809e-08,
+      "loss": 1.0162,
+      "step": 22017
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.561250126303987e-08,
+      "loss": 0.8874,
+      "step": 22018
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.5491971274263e-08,
+      "loss": 0.8619,
+      "step": 22019
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.537153669671115e-08,
+      "loss": 1.0262,
+      "step": 22020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.52511975327197e-08,
+      "loss": 0.8901,
+      "step": 22021
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.513095378461843e-08,
+      "loss": 1.117,
+      "step": 22022
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.501080545473827e-08,
+      "loss": 0.9774,
+      "step": 22023
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.489075254540678e-08,
+      "loss": 1.073,
+      "step": 22024
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.477079505894935e-08,
+      "loss": 0.94,
+      "step": 22025
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.465093299769021e-08,
+      "loss": 1.0133,
+      "step": 22026
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.453116636395308e-08,
+      "loss": 0.9103,
+      "step": 22027
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.441149516005609e-08,
+      "loss": 0.8996,
+      "step": 22028
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.429191938832015e-08,
+      "loss": 0.8035,
+      "step": 22029
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.417243905106009e-08,
+      "loss": 0.921,
+      "step": 22030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.405305415059238e-08,
+      "loss": 0.9196,
+      "step": 22031
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.393376468922797e-08,
+      "loss": 1.0256,
+      "step": 22032
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.38145706692811e-08,
+      "loss": 0.9226,
+      "step": 22033
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.369547209305827e-08,
+      "loss": 0.8161,
+      "step": 22034
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.35764689628693e-08,
+      "loss": 0.8173,
+      "step": 22035
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.345756128101789e-08,
+      "loss": 0.8478,
+      "step": 22036
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.333874904980887e-08,
+      "loss": 0.9864,
+      "step": 22037
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.322003227154428e-08,
+      "loss": 0.8883,
+      "step": 22038
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.310141094852396e-08,
+      "loss": 0.8886,
+      "step": 22039
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.298288508304607e-08,
+      "loss": 0.8434,
+      "step": 22040
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.28644546774071e-08,
+      "loss": 0.8164,
+      "step": 22041
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.274611973390189e-08,
+      "loss": 0.9765,
+      "step": 22042
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.262788025482249e-08,
+      "loss": 1.1409,
+      "step": 22043
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.250973624246038e-08,
+      "loss": 1.0313,
+      "step": 22044
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.239168769910431e-08,
+      "loss": 1.0819,
+      "step": 22045
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.22737346270419e-08,
+      "loss": 0.8335,
+      "step": 22046
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.215587702855797e-08,
+      "loss": 0.9888,
+      "step": 22047
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.203811490593626e-08,
+      "loss": 0.9125,
+      "step": 22048
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.192044826145772e-08,
+      "loss": 0.9283,
+      "step": 22049
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.18028770974033e-08,
+      "loss": 0.9364,
+      "step": 22050
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.168540141604952e-08,
+      "loss": 0.9519,
+      "step": 22051
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.156802121967343e-08,
+      "loss": 0.9122,
+      "step": 22052
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.14507365105488e-08,
+      "loss": 1.1039,
+      "step": 22053
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.133354729094877e-08,
+      "loss": 0.8662,
+      "step": 22054
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.121645356314211e-08,
+      "loss": 0.9991,
+      "step": 22055
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.109945532939977e-08,
+      "loss": 0.7881,
+      "step": 22056
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.098255259198661e-08,
+      "loss": 0.844,
+      "step": 22057
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.086574535316915e-08,
+      "loss": 0.9106,
+      "step": 22058
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.074903361520946e-08,
+      "loss": 0.8916,
+      "step": 22059
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.063241738036908e-08,
+      "loss": 1.0136,
+      "step": 22060
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.05158966509073e-08,
+      "loss": 0.979,
+      "step": 22061
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.039947142908232e-08,
+      "loss": 0.9291,
+      "step": 22062
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.028314171714846e-08,
+      "loss": 0.9845,
+      "step": 22063
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.016690751736111e-08,
+      "loss": 1.0116,
+      "step": 22064
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.005076883197182e-08,
+      "loss": 0.8578,
+      "step": 22065
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.993472566323046e-08,
+      "loss": 0.9944,
+      "step": 22066
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.981877801338521e-08,
+      "loss": 0.9775,
+      "step": 22067
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.970292588468264e-08,
+      "loss": 0.8773,
+      "step": 22068
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.958716927936815e-08,
+      "loss": 0.9997,
+      "step": 22069
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.94715081996833e-08,
+      "loss": 1.0285,
+      "step": 22070
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.93559426478696e-08,
+      "loss": 1.0851,
+      "step": 22071
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.924047262616585e-08,
+      "loss": 0.9708,
+      "step": 22072
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.912509813681024e-08,
+      "loss": 0.9238,
+      "step": 22073
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.900981918203652e-08,
+      "loss": 1.0022,
+      "step": 22074
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.889463576407962e-08,
+      "loss": 0.9455,
+      "step": 22075
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.87795478851705e-08,
+      "loss": 0.9095,
+      "step": 22076
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.866455554753904e-08,
+      "loss": 0.826,
+      "step": 22077
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.854965875341235e-08,
+      "loss": 0.7423,
+      "step": 22078
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.84348575050181e-08,
+      "loss": 1.0656,
+      "step": 22079
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.832015180458008e-08,
+      "loss": 0.7068,
+      "step": 22080
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.820554165432036e-08,
+      "loss": 0.8369,
+      "step": 22081
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.809102705645886e-08,
+      "loss": 0.8855,
+      "step": 22082
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.797660801321548e-08,
+      "loss": 0.9301,
+      "step": 22083
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.786228452680677e-08,
+      "loss": 0.9818,
+      "step": 22084
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.774805659944761e-08,
+      "loss": 0.9646,
+      "step": 22085
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.763392423335014e-08,
+      "loss": 0.8399,
+      "step": 22086
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.751988743072757e-08,
+      "loss": 0.8186,
+      "step": 22087
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.74059461937876e-08,
+      "loss": 0.8832,
+      "step": 22088
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.7292100524739e-08,
+      "loss": 0.8861,
+      "step": 22089
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.717835042578613e-08,
+      "loss": 0.9492,
+      "step": 22090
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.706469589913445e-08,
+      "loss": 1.0138,
+      "step": 22091
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.695113694698496e-08,
+      "loss": 1.0078,
+      "step": 22092
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.68376735715387e-08,
+      "loss": 1.0,
+      "step": 22093
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.672430577499334e-08,
+      "loss": 0.906,
+      "step": 22094
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.661103355954545e-08,
+      "loss": 1.0038,
+      "step": 22095
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.649785692738997e-08,
+      "loss": 1.074,
+      "step": 22096
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.638477588071901e-08,
+      "loss": 1.0857,
+      "step": 22097
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.627179042172416e-08,
+      "loss": 0.8434,
+      "step": 22098
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.615890055259422e-08,
+      "loss": 0.8161,
+      "step": 22099
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.604610627551578e-08,
+      "loss": 0.9112,
+      "step": 22100
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.593340759267542e-08,
+      "loss": 0.9913,
+      "step": 22101
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.582080450625582e-08,
+      "loss": 0.8594,
+      "step": 22102
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.570829701843917e-08,
+      "loss": 0.9452,
+      "step": 22103
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.559588513140481e-08,
+      "loss": 0.9246,
+      "step": 22104
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.548356884733043e-08,
+      "loss": 0.8709,
+      "step": 22105
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.53713481683932e-08,
+      "loss": 1.1506,
+      "step": 22106
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.525922309676636e-08,
+      "loss": 0.8738,
+      "step": 22107
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.514719363462319e-08,
+      "loss": 0.896,
+      "step": 22108
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.503525978413305e-08,
+      "loss": 0.9995,
+      "step": 22109
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.49234215474659e-08,
+      "loss": 0.9725,
+      "step": 22110
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.481167892678775e-08,
+      "loss": 0.8686,
+      "step": 22111
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.470003192426411e-08,
+      "loss": 0.9548,
+      "step": 22112
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.458848054205714e-08,
+      "loss": 1.0513,
+      "step": 22113
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.447702478232954e-08,
+      "loss": 0.8824,
+      "step": 22114
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.436566464723959e-08,
+      "loss": 0.8093,
+      "step": 22115
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.425440013894557e-08,
+      "loss": 0.757,
+      "step": 22116
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.414323125960243e-08,
+      "loss": 0.9186,
+      "step": 22117
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.403215801136508e-08,
+      "loss": 0.7139,
+      "step": 22118
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.392118039638462e-08,
+      "loss": 0.913,
+      "step": 22119
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.381029841681207e-08,
+      "loss": 0.8061,
+      "step": 22120
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.369951207479464e-08,
+      "loss": 0.9471,
+      "step": 22121
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.358882137248001e-08,
+      "loss": 1.0104,
+      "step": 22122
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.347822631201206e-08,
+      "loss": 1.0401,
+      "step": 22123
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.336772689553405e-08,
+      "loss": 0.9427,
+      "step": 22124
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.32573231251854e-08,
+      "loss": 0.9646,
+      "step": 22125
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.314701500310716e-08,
+      "loss": 0.8937,
+      "step": 22126
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.30368025314354e-08,
+      "loss": 1.0074,
+      "step": 22127
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.292668571230565e-08,
+      "loss": 1.0122,
+      "step": 22128
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.281666454785174e-08,
+      "loss": 0.6978,
+      "step": 22129
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.270673904020475e-08,
+      "loss": 0.9842,
+      "step": 22130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.259690919149464e-08,
+      "loss": 0.8912,
+      "step": 22131
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.248717500384971e-08,
+      "loss": 0.8296,
+      "step": 22132
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.237753647939548e-08,
+      "loss": 0.8703,
+      "step": 22133
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.226799362025692e-08,
+      "loss": 0.9844,
+      "step": 22134
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.215854642855512e-08,
+      "loss": 1.0446,
+      "step": 22135
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.204919490641171e-08,
+      "loss": 0.8343,
+      "step": 22136
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.193993905594553e-08,
+      "loss": 0.9718,
+      "step": 22137
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.183077887927214e-08,
+      "loss": 0.8737,
+      "step": 22138
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.172171437850816e-08,
+      "loss": 1.0337,
+      "step": 22139
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.161274555576468e-08,
+      "loss": 0.9027,
+      "step": 22140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.150387241315503e-08,
+      "loss": 1.0484,
+      "step": 22141
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.139509495278695e-08,
+      "loss": 1.011,
+      "step": 22142
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.12864131767693e-08,
+      "loss": 0.9794,
+      "step": 22143
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.117782708720654e-08,
+      "loss": 0.9707,
+      "step": 22144
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.106933668620307e-08,
+      "loss": 0.9525,
+      "step": 22145
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.096094197586111e-08,
+      "loss": 0.9871,
+      "step": 22146
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.085264295828063e-08,
+      "loss": 1.0457,
+      "step": 22147
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.074443963555887e-08,
+      "loss": 0.9145,
+      "step": 22148
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.063633200979412e-08,
+      "loss": 1.1401,
+      "step": 22149
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.052832008307974e-08,
+      "loss": 1.036,
+      "step": 22150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.042040385750902e-08,
+      "loss": 0.8922,
+      "step": 22151
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.031258333517198e-08,
+      "loss": 0.9137,
+      "step": 22152
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.02048585181586e-08,
+      "loss": 0.9013,
+      "step": 22153
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.009722940855555e-08,
+      "loss": 0.9336,
+      "step": 22154
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.998969600844784e-08,
+      "loss": 0.8285,
+      "step": 22155
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.988225831991934e-08,
+      "loss": 0.9025,
+      "step": 22156
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.977491634505117e-08,
+      "loss": 1.0532,
+      "step": 22157
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.966767008592389e-08,
+      "loss": 1.0789,
+      "step": 22158
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9560519544614725e-08,
+      "loss": 0.9512,
+      "step": 22159
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.945346472319924e-08,
+      "loss": 0.7839,
+      "step": 22160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9346505623752435e-08,
+      "loss": 0.9241,
+      "step": 22161
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.923964224834655e-08,
+      "loss": 0.8316,
+      "step": 22162
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9132874599051595e-08,
+      "loss": 0.9493,
+      "step": 22163
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9026202677937014e-08,
+      "loss": 0.7552,
+      "step": 22164
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.891962648706839e-08,
+      "loss": 1.0199,
+      "step": 22165
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.881314602851185e-08,
+      "loss": 0.9614,
+      "step": 22166
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.870676130432906e-08,
+      "loss": 0.8924,
+      "step": 22167
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8600472316582277e-08,
+      "loss": 0.9025,
+      "step": 22168
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.849427906732985e-08,
+      "loss": 0.9582,
+      "step": 22169
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8388181558630684e-08,
+      "loss": 0.8793,
+      "step": 22170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.82821797925387e-08,
+      "loss": 1.0726,
+      "step": 22171
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8176273771108906e-08,
+      "loss": 0.9599,
+      "step": 22172
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.807046349639245e-08,
+      "loss": 0.8854,
+      "step": 22173
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.796474897044047e-08,
+      "loss": 0.8924,
+      "step": 22174
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.7859130195300205e-08,
+      "loss": 0.8593,
+      "step": 22175
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.7753607173018364e-08,
+      "loss": 1.035,
+      "step": 22176
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.764817990563887e-08,
+      "loss": 1.0304,
+      "step": 22177
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.754284839520508e-08,
+      "loss": 0.99,
+      "step": 22178
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.743761264375702e-08,
+      "loss": 1.0267,
+      "step": 22179
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.733247265333475e-08,
+      "loss": 1.032,
+      "step": 22180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.7227428425974395e-08,
+      "loss": 0.8636,
+      "step": 22181
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.7122479963711005e-08,
+      "loss": 0.8634,
+      "step": 22182
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.7017627268578515e-08,
+      "loss": 0.8612,
+      "step": 22183
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6912870342608614e-08,
+      "loss": 0.8856,
+      "step": 22184
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.68082091878297e-08,
+      "loss": 0.783,
+      "step": 22185
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6703643806271245e-08,
+      "loss": 0.9212,
+      "step": 22186
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6599174199958306e-08,
+      "loss": 0.8014,
+      "step": 22187
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6494800370914814e-08,
+      "loss": 0.7672,
+      "step": 22188
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.639052232116304e-08,
+      "loss": 0.8746,
+      "step": 22189
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6286340052724156e-08,
+      "loss": 0.9906,
+      "step": 22190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.618225356761542e-08,
+      "loss": 0.7635,
+      "step": 22191
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6078262867854114e-08,
+      "loss": 1.0596,
+      "step": 22192
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.59743679554553e-08,
+      "loss": 0.8077,
+      "step": 22193
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.587056883243125e-08,
+      "loss": 0.9316,
+      "step": 22194
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.576686550079369e-08,
+      "loss": 0.9287,
+      "step": 22195
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.566325796255101e-08,
+      "loss": 0.9652,
+      "step": 22196
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.555974621971216e-08,
+      "loss": 0.9362,
+      "step": 22197
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.545633027428055e-08,
+      "loss": 0.9571,
+      "step": 22198
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.5353010128261775e-08,
+      "loss": 0.857,
+      "step": 22199
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.5249785783656476e-08,
+      "loss": 0.9292,
+      "step": 22200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.514665724246471e-08,
+      "loss": 1.051,
+      "step": 22201
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.504362450668488e-08,
+      "loss": 1.0137,
+      "step": 22202
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.494068757831372e-08,
+      "loss": 0.9922,
+      "step": 22203
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.483784645934409e-08,
+      "loss": 0.9545,
+      "step": 22204
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.47351011517705e-08,
+      "loss": 0.8093,
+      "step": 22205
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.4632451657581353e-08,
+      "loss": 0.7735,
+      "step": 22206
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.4529897978767286e-08,
+      "loss": 1.0795,
+      "step": 22207
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.442744011731449e-08,
+      "loss": 0.7617,
+      "step": 22208
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.432507807520804e-08,
+      "loss": 0.9011,
+      "step": 22209
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.4222811854431904e-08,
+      "loss": 0.9613,
+      "step": 22210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.412064145696672e-08,
+      "loss": 0.7605,
+      "step": 22211
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.4018566884791455e-08,
+      "loss": 1.0358,
+      "step": 22212
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.391658813988565e-08,
+      "loss": 0.9088,
+      "step": 22213
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.381470522422327e-08,
+      "loss": 0.7921,
+      "step": 22214
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3712918139779415e-08,
+      "loss": 0.9664,
+      "step": 22215
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.361122688852527e-08,
+      "loss": 1.1214,
+      "step": 22216
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.350963147243204e-08,
+      "loss": 0.8822,
+      "step": 22217
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.340813189346705e-08,
+      "loss": 1.016,
+      "step": 22218
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.330672815359761e-08,
+      "loss": 0.9223,
+      "step": 22219
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.320542025478825e-08,
+      "loss": 0.9294,
+      "step": 22220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3104208199002414e-08,
+      "loss": 0.9984,
+      "step": 22221
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.300309198819964e-08,
+      "loss": 0.8055,
+      "step": 22222
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2902071624340024e-08,
+      "loss": 0.8347,
+      "step": 22223
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.280114710938034e-08,
+      "loss": 0.9179,
+      "step": 22224
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.270031844527679e-08,
+      "loss": 0.8874,
+      "step": 22225
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.259958563398226e-08,
+      "loss": 0.8504,
+      "step": 22226
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.249894867744798e-08,
+      "loss": 0.9859,
+      "step": 22227
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.23984075776246e-08,
+      "loss": 0.7194,
+      "step": 22228
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.229796233645945e-08,
+      "loss": 1.0204,
+      "step": 22229
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2197612955899866e-08,
+      "loss": 0.9034,
+      "step": 22230
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2097359437888186e-08,
+      "loss": 0.7425,
+      "step": 22231
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.19972017843684e-08,
+      "loss": 0.8863,
+      "step": 22232
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1897139997280074e-08,
+      "loss": 0.756,
+      "step": 22233
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1797174078562775e-08,
+      "loss": 1.0988,
+      "step": 22234
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.169730403015216e-08,
+      "loss": 0.9881,
+      "step": 22235
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.159752985398392e-08,
+      "loss": 0.8141,
+      "step": 22236
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.149785155199094e-08,
+      "loss": 0.845,
+      "step": 22237
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.139826912610502e-08,
+      "loss": 1.0181,
+      "step": 22238
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.129878257825516e-08,
+      "loss": 0.8809,
+      "step": 22239
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1199391910368714e-08,
+      "loss": 0.9928,
+      "step": 22240
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.110009712437081e-08,
+      "loss": 0.8295,
+      "step": 22241
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.100089822218657e-08,
+      "loss": 1.0192,
+      "step": 22242
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.0901795205737234e-08,
+      "loss": 0.915,
+      "step": 22243
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.0802788076942945e-08,
+      "loss": 0.9941,
+      "step": 22244
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.07038768377216e-08,
+      "loss": 0.9089,
+      "step": 22245
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.060506148999056e-08,
+      "loss": 0.9919,
+      "step": 22246
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.050634203566329e-08,
+      "loss": 1.0107,
+      "step": 22247
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.0407718476653266e-08,
+      "loss": 1.0111,
+      "step": 22248
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.030919081487118e-08,
+      "loss": 0.791,
+      "step": 22249
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.02107590522255e-08,
+      "loss": 0.8519,
+      "step": 22250
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.0112423190623036e-08,
+      "loss": 1.0206,
+      "step": 22251
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.001418323197005e-08,
+      "loss": 1.1537,
+      "step": 22252
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9916039178169454e-08,
+      "loss": 0.9803,
+      "step": 22253
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9817991031121946e-08,
+      "loss": 0.757,
+      "step": 22254
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9720038792728795e-08,
+      "loss": 0.7022,
+      "step": 22255
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.962218246488681e-08,
+      "loss": 0.924,
+      "step": 22256
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.952442204949226e-08,
+      "loss": 0.8011,
+      "step": 22257
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.942675754843862e-08,
+      "loss": 0.8107,
+      "step": 22258
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.932918896361938e-08,
+      "loss": 1.1081,
+      "step": 22259
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9231716296923584e-08,
+      "loss": 1.0912,
+      "step": 22260
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.913433955024027e-08,
+      "loss": 1.0649,
+      "step": 22261
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9037058725455724e-08,
+      "loss": 1.025,
+      "step": 22262
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.8939873824455644e-08,
+      "loss": 1.0017,
+      "step": 22263
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.884278484912186e-08,
+      "loss": 0.9162,
+      "step": 22264
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.874579180133676e-08,
+      "loss": 0.8595,
+      "step": 22265
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.864889468297829e-08,
+      "loss": 0.8917,
+      "step": 22266
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.855209349592438e-08,
+      "loss": 0.9225,
+      "step": 22267
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.845538824205076e-08,
+      "loss": 0.9276,
+      "step": 22268
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.8358778923230375e-08,
+      "loss": 0.7639,
+      "step": 22269
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.826226554133562e-08,
+      "loss": 0.8845,
+      "step": 22270
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.816584809823666e-08,
+      "loss": 0.8384,
+      "step": 22271
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.806952659580033e-08,
+      "loss": 1.1504,
+      "step": 22272
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.797330103589459e-08,
+      "loss": 0.9527,
+      "step": 22273
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.787717142038184e-08,
+      "loss": 1.0024,
+      "step": 22274
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7781137751126136e-08,
+      "loss": 0.85,
+      "step": 22275
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.768520002998711e-08,
+      "loss": 0.7444,
+      "step": 22276
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.758935825882438e-08,
+      "loss": 0.8732,
+      "step": 22277
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.749361243949424e-08,
+      "loss": 1.0211,
+      "step": 22278
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.739796257385243e-08,
+      "loss": 0.8695,
+      "step": 22279
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.73024086637508e-08,
+      "loss": 0.8606,
+      "step": 22280
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.720695071104231e-08,
+      "loss": 0.8651,
+      "step": 22281
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.711158871757548e-08,
+      "loss": 0.8491,
+      "step": 22282
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.701632268519773e-08,
+      "loss": 1.0547,
+      "step": 22283
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.692115261575536e-08,
+      "loss": 0.8581,
+      "step": 22284
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6826078511092444e-08,
+      "loss": 0.7598,
+      "step": 22285
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6731100373050286e-08,
+      "loss": 0.9384,
+      "step": 22286
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.663621820346964e-08,
+      "loss": 0.9262,
+      "step": 22287
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.654143200418903e-08,
+      "loss": 0.9736,
+      "step": 22288
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6446741777043666e-08,
+      "loss": 1.0441,
+      "step": 22289
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.635214752386985e-08,
+      "loss": 1.0073,
+      "step": 22290
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.625764924649889e-08,
+      "loss": 0.9854,
+      "step": 22291
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6163246946763215e-08,
+      "loss": 0.8195,
+      "step": 22292
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.60689406264897e-08,
+      "loss": 1.0481,
+      "step": 22293
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.597473028750743e-08,
+      "loss": 0.9492,
+      "step": 22294
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.588061593164106e-08,
+      "loss": 0.9593,
+      "step": 22295
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5786597560713575e-08,
+      "loss": 0.934,
+      "step": 22296
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5692675176547405e-08,
+      "loss": 0.7849,
+      "step": 22297
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.559884878096166e-08,
+      "loss": 0.968,
+      "step": 22298
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.550511837577376e-08,
+      "loss": 0.9224,
+      "step": 22299
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.541148396280115e-08,
+      "loss": 0.788,
+      "step": 22300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.531794554385682e-08,
+      "loss": 0.9218,
+      "step": 22301
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5224503120753216e-08,
+      "loss": 0.8534,
+      "step": 22302
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.51311566953011e-08,
+      "loss": 1.0315,
+      "step": 22303
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.503790626930904e-08,
+      "loss": 1.005,
+      "step": 22304
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4944751844583354e-08,
+      "loss": 0.7508,
+      "step": 22305
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.485169342292983e-08,
+      "loss": 0.9616,
+      "step": 22306
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4758731006149804e-08,
+      "loss": 0.916,
+      "step": 22307
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.466586459604572e-08,
+      "loss": 1.0115,
+      "step": 22308
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.457309419441613e-08,
+      "loss": 0.8515,
+      "step": 22309
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.448041980305906e-08,
+      "loss": 1.0367,
+      "step": 22310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.438784142376973e-08,
+      "loss": 1.0802,
+      "step": 22311
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.42953590583417e-08,
+      "loss": 0.9853,
+      "step": 22312
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.420297270856744e-08,
+      "loss": 1.0118,
+      "step": 22313
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.41106823762355e-08,
+      "loss": 0.8823,
+      "step": 22314
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.401848806313558e-08,
+      "loss": 0.8451,
+      "step": 22315
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.39263897710529e-08,
+      "loss": 1.1802,
+      "step": 22316
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.383438750177271e-08,
+      "loss": 1.0224,
+      "step": 22317
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.374248125707692e-08,
+      "loss": 1.1631,
+      "step": 22318
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3650671038745765e-08,
+      "loss": 1.0486,
+      "step": 22319
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.355895684855893e-08,
+      "loss": 0.967,
+      "step": 22320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.346733868829278e-08,
+      "loss": 0.9432,
+      "step": 22321
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.337581655972256e-08,
+      "loss": 0.9584,
+      "step": 22322
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3284390464622404e-08,
+      "loss": 0.9442,
+      "step": 22323
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3193060404762014e-08,
+      "loss": 0.962,
+      "step": 22324
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3101826381911627e-08,
+      "loss": 0.8871,
+      "step": 22325
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3010688397839286e-08,
+      "loss": 1.0207,
+      "step": 22326
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.291964645431024e-08,
+      "loss": 0.7269,
+      "step": 22327
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.282870055308863e-08,
+      "loss": 1.0021,
+      "step": 22328
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.273785069593695e-08,
+      "loss": 0.8311,
+      "step": 22329
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2647096884614326e-08,
+      "loss": 0.8137,
+      "step": 22330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2556439120879925e-08,
+      "loss": 0.9107,
+      "step": 22331
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.246587740648955e-08,
+      "loss": 0.9126,
+      "step": 22332
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.237541174319848e-08,
+      "loss": 1.0601,
+      "step": 22333
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.228504213275919e-08,
+      "loss": 0.9275,
+      "step": 22334
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.219476857692306e-08,
+      "loss": 0.8581,
+      "step": 22335
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.210459107743814e-08,
+      "loss": 1.0912,
+      "step": 22336
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.201450963605247e-08,
+      "loss": 0.9773,
+      "step": 22337
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.192452425451077e-08,
+      "loss": 0.9144,
+      "step": 22338
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1834634934556636e-08,
+      "loss": 0.796,
+      "step": 22339
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1744841677932024e-08,
+      "loss": 0.9565,
+      "step": 22340
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.165514448637664e-08,
+      "loss": 0.9352,
+      "step": 22341
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1565543361627435e-08,
+      "loss": 0.7645,
+      "step": 22342
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.147603830542191e-08,
+      "loss": 0.9102,
+      "step": 22343
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.138662931949255e-08,
+      "loss": 0.911,
+      "step": 22344
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1297316405573e-08,
+      "loss": 1.045,
+      "step": 22345
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1208099565392404e-08,
+      "loss": 0.7424,
+      "step": 22346
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.11189788006805e-08,
+      "loss": 0.8822,
+      "step": 22347
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1029954113163684e-08,
+      "loss": 1.1231,
+      "step": 22348
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.094102550456669e-08,
+      "loss": 0.9222,
+      "step": 22349
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0852192976612023e-08,
+      "loss": 0.8389,
+      "step": 22350
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0763456531021093e-08,
+      "loss": 0.9648,
+      "step": 22351
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.067481616951419e-08,
+      "loss": 0.9106,
+      "step": 22352
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.05862718938066e-08,
+      "loss": 0.8907,
+      "step": 22353
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0497823705615836e-08,
+      "loss": 0.9451,
+      "step": 22354
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.040947160665443e-08,
+      "loss": 0.8771,
+      "step": 22355
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0321215598634335e-08,
+      "loss": 0.9447,
+      "step": 22356
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.0233055683265296e-08,
+      "loss": 0.8999,
+      "step": 22357
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.014499186225651e-08,
+      "loss": 1.0005,
+      "step": 22358
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.005702413731327e-08,
+      "loss": 1.0493,
+      "step": 22359
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9969152510139775e-08,
+      "loss": 0.9589,
+      "step": 22360
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9881376982439104e-08,
+      "loss": 0.9665,
+      "step": 22361
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.979369755591156e-08,
+      "loss": 1.0087,
+      "step": 22362
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.970611423225579e-08,
+      "loss": 0.9528,
+      "step": 22363
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.961862701316932e-08,
+      "loss": 0.8913,
+      "step": 22364
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9531235900346356e-08,
+      "loss": 0.8769,
+      "step": 22365
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.944394089548054e-08,
+      "loss": 1.0843,
+      "step": 22366
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9356742000262736e-08,
+      "loss": 0.8076,
+      "step": 22367
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9269639216383825e-08,
+      "loss": 1.066,
+      "step": 22368
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.918263254552912e-08,
+      "loss": 1.0244,
+      "step": 22369
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.909572198938616e-08,
+      "loss": 0.7318,
+      "step": 22370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9008907549638596e-08,
+      "loss": 0.8607,
+      "step": 22371
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.892218922796787e-08,
+      "loss": 0.89,
+      "step": 22372
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8835567026054286e-08,
+      "loss": 0.8895,
+      "step": 22373
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.874904094557652e-08,
+      "loss": 0.93,
+      "step": 22374
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.866261098820989e-08,
+      "loss": 0.8889,
+      "step": 22375
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8576277155630836e-08,
+      "loss": 0.987,
+      "step": 22376
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.849003944951024e-08,
+      "loss": 0.8998,
+      "step": 22377
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.84038978715201e-08,
+      "loss": 0.7897,
+      "step": 22378
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.831785242332853e-08,
+      "loss": 0.8577,
+      "step": 22379
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8231903106603074e-08,
+      "loss": 1.0176,
+      "step": 22380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.814604992300908e-08,
+      "loss": 0.8307,
+      "step": 22381
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.806029287420965e-08,
+      "loss": 0.8614,
+      "step": 22382
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.797463196186735e-08,
+      "loss": 1.077,
+      "step": 22383
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.788906718764029e-08,
+      "loss": 0.9836,
+      "step": 22384
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.780359855318716e-08,
+      "loss": 0.9228,
+      "step": 22385
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.771822606016329e-08,
+      "loss": 1.0087,
+      "step": 22386
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7632949710223465e-08,
+      "loss": 0.8659,
+      "step": 22387
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7547769505019707e-08,
+      "loss": 0.9707,
+      "step": 22388
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.746268544620235e-08,
+      "loss": 0.9448,
+      "step": 22389
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.737769753541898e-08,
+      "loss": 0.981,
+      "step": 22390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7292805774317706e-08,
+      "loss": 0.8517,
+      "step": 22391
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.720801016454223e-08,
+      "loss": 0.7915,
+      "step": 22392
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.712331070773623e-08,
+      "loss": 1.007,
+      "step": 22393
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.703870740554005e-08,
+      "loss": 0.9177,
+      "step": 22394
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.695420025959351e-08,
+      "loss": 0.9019,
+      "step": 22395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.686978927153306e-08,
+      "loss": 0.9555,
+      "step": 22396
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.678547444299463e-08,
+      "loss": 0.9244,
+      "step": 22397
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.67012557756119e-08,
+      "loss": 0.7519,
+      "step": 22398
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.66171332710169e-08,
+      "loss": 0.9442,
+      "step": 22399
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6533106930838335e-08,
+      "loss": 0.928,
+      "step": 22400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6449176756705454e-08,
+      "loss": 0.929,
+      "step": 22401
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6365342750243616e-08,
+      "loss": 0.814,
+      "step": 22402
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6281604913077085e-08,
+      "loss": 1.0288,
+      "step": 22403
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6197963246829e-08,
+      "loss": 0.9943,
+      "step": 22404
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.611441775311864e-08,
+      "loss": 1.0196,
+      "step": 22405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.603096843356635e-08,
+      "loss": 0.9155,
+      "step": 22406
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.594761528978752e-08,
+      "loss": 0.8784,
+      "step": 22407
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.586435832339808e-08,
+      "loss": 0.8092,
+      "step": 22408
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.578119753601006e-08,
+      "loss": 0.8303,
+      "step": 22409
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.569813292923552e-08,
+      "loss": 0.9287,
+      "step": 22410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.561516450468372e-08,
+      "loss": 1.0358,
+      "step": 22411
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.553229226396171e-08,
+      "loss": 0.7807,
+      "step": 22412
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.544951620867543e-08,
+      "loss": 0.8409,
+      "step": 22413
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.536683634042859e-08,
+      "loss": 1.1045,
+      "step": 22414
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.528425266082325e-08,
+      "loss": 0.8736,
+      "step": 22415
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5201765171459236e-08,
+      "loss": 0.9164,
+      "step": 22416
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5119373873934715e-08,
+      "loss": 0.9927,
+      "step": 22417
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.503707876984674e-08,
+      "loss": 0.7084,
+      "step": 22418
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.495487986078849e-08,
+      "loss": 0.9271,
+      "step": 22419
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.487277714835369e-08,
+      "loss": 0.7742,
+      "step": 22420
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.479077063413161e-08,
+      "loss": 0.9902,
+      "step": 22421
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.470886031971321e-08,
+      "loss": 1.0334,
+      "step": 22422
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4627046206683336e-08,
+      "loss": 0.9109,
+      "step": 22423
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.454532829662905e-08,
+      "loss": 0.9856,
+      "step": 22424
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.446370659113185e-08,
+      "loss": 0.9112,
+      "step": 22425
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.438218109177438e-08,
+      "loss": 1.0357,
+      "step": 22426
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.430075180013537e-08,
+      "loss": 0.9549,
+      "step": 22427
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.421941871779355e-08,
+      "loss": 0.9198,
+      "step": 22428
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4138181846323225e-08,
+      "loss": 0.9097,
+      "step": 22429
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.40570411872998e-08,
+      "loss": 0.8822,
+      "step": 22430
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3975996742294793e-08,
+      "loss": 0.854,
+      "step": 22431
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.389504851287806e-08,
+      "loss": 0.9368,
+      "step": 22432
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3814196500618346e-08,
+      "loss": 1.0075,
+      "step": 22433
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3733440707082176e-08,
+      "loss": 1.0017,
+      "step": 22434
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3652781133833856e-08,
+      "loss": 1.071,
+      "step": 22435
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.357221778243658e-08,
+      "loss": 0.9576,
+      "step": 22436
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.349175065445076e-08,
+      "loss": 0.9534,
+      "step": 22437
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3411379751436265e-08,
+      "loss": 0.9266,
+      "step": 22438
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3331105074949075e-08,
+      "loss": 0.9589,
+      "step": 22439
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.325092662654516e-08,
+      "loss": 0.8116,
+      "step": 22440
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.317084440777774e-08,
+      "loss": 0.772,
+      "step": 22441
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.309085842019888e-08,
+      "loss": 0.9942,
+      "step": 22442
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.301096866535791e-08,
+      "loss": 0.8262,
+      "step": 22443
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2931175144802484e-08,
+      "loss": 1.018,
+      "step": 22444
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.285147786007914e-08,
+      "loss": 0.9119,
+      "step": 22445
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.277187681273164e-08,
+      "loss": 1.0481,
+      "step": 22446
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2692372004302086e-08,
+      "loss": 0.8803,
+      "step": 22447
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.261296343633091e-08,
+      "loss": 0.9283,
+      "step": 22448
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2533651110357444e-08,
+      "loss": 0.9583,
+      "step": 22449
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.245443502791656e-08,
+      "loss": 1.0165,
+      "step": 22450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.237531519054538e-08,
+      "loss": 0.9746,
+      "step": 22451
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2296291599774875e-08,
+      "loss": 1.0336,
+      "step": 22452
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2217364257137173e-08,
+      "loss": 0.9307,
+      "step": 22453
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.213853316416105e-08,
+      "loss": 0.8169,
+      "step": 22454
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.205979832237416e-08,
+      "loss": 0.9244,
+      "step": 22455
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1981159733301405e-08,
+      "loss": 0.8382,
+      "step": 22456
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1902617398466564e-08,
+      "loss": 0.9643,
+      "step": 22457
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.182417131939175e-08,
+      "loss": 0.8595,
+      "step": 22458
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.174582149759742e-08,
+      "loss": 1.0445,
+      "step": 22459
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.166756793459958e-08,
+      "loss": 0.9376,
+      "step": 22460
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1589410631916454e-08,
+      "loss": 0.6891,
+      "step": 22461
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.151134959106073e-08,
+      "loss": 0.8983,
+      "step": 22462
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1433384813546186e-08,
+      "loss": 0.7821,
+      "step": 22463
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.135551630088274e-08,
+      "loss": 1.0453,
+      "step": 22464
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.127774405457917e-08,
+      "loss": 0.9536,
+      "step": 22465
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.120006807614151e-08,
+      "loss": 0.9305,
+      "step": 22466
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.112248836707632e-08,
+      "loss": 0.8174,
+      "step": 22467
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.104500492888518e-08,
+      "loss": 1.0086,
+      "step": 22468
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.096761776307022e-08,
+      "loss": 0.9,
+      "step": 22469
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.089032687113025e-08,
+      "loss": 0.7503,
+      "step": 22470
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0813132254563505e-08,
+      "loss": 1.091,
+      "step": 22471
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0736033914864906e-08,
+      "loss": 0.9245,
+      "step": 22472
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.06590318535277e-08,
+      "loss": 0.8942,
+      "step": 22473
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0582126072045694e-08,
+      "loss": 0.9367,
+      "step": 22474
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.050531657190714e-08,
+      "loss": 0.9192,
+      "step": 22475
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.042860335460085e-08,
+      "loss": 0.786,
+      "step": 22476
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0351986421612855e-08,
+      "loss": 0.9692,
+      "step": 22477
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.027546577442808e-08,
+      "loss": 0.7992,
+      "step": 22478
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0199041414528674e-08,
+      "loss": 0.9343,
+      "step": 22479
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.012271334339567e-08,
+      "loss": 0.9611,
+      "step": 22480
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.004648156250789e-08,
+      "loss": 0.9037,
+      "step": 22481
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.997034607334193e-08,
+      "loss": 0.9184,
+      "step": 22482
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.989430687737327e-08,
+      "loss": 0.8376,
+      "step": 22483
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.981836397607463e-08,
+      "loss": 0.9306,
+      "step": 22484
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9742517370917602e-08,
+      "loss": 0.8477,
+      "step": 22485
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.966676706337268e-08,
+      "loss": 0.9772,
+      "step": 22486
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.959111305490592e-08,
+      "loss": 0.9431,
+      "step": 22487
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.951555534698447e-08,
+      "loss": 0.8619,
+      "step": 22488
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.944009394107161e-08,
+      "loss": 0.9325,
+      "step": 22489
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9364728838628952e-08,
+      "loss": 0.9013,
+      "step": 22490
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9289460041117546e-08,
+      "loss": 0.8744,
+      "step": 22491
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9214287549995114e-08,
+      "loss": 0.8306,
+      "step": 22492
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9139211366718267e-08,
+      "loss": 0.9696,
+      "step": 22493
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9064231492741957e-08,
+      "loss": 0.8486,
+      "step": 22494
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8989347929518353e-08,
+      "loss": 0.9857,
+      "step": 22495
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8914560678498517e-08,
+      "loss": 0.7561,
+      "step": 22496
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8839869741131287e-08,
+      "loss": 0.8384,
+      "step": 22497
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8765275118863844e-08,
+      "loss": 1.0371,
+      "step": 22498
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8690776813141697e-08,
+      "loss": 0.9192,
+      "step": 22499
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8616374825408133e-08,
+      "loss": 0.8883,
+      "step": 22500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.854206915710478e-08,
+      "loss": 0.9824,
+      "step": 22501
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8467859809670485e-08,
+      "loss": 0.9334,
+      "step": 22502
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.839374678454465e-08,
+      "loss": 0.9165,
+      "step": 22503
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8319730083161134e-08,
+      "loss": 0.7034,
+      "step": 22504
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8245809706956007e-08,
+      "loss": 0.9605,
+      "step": 22505
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8171985657359792e-08,
+      "loss": 0.7317,
+      "step": 22506
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8098257935804118e-08,
+      "loss": 1.0005,
+      "step": 22507
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8024626543716184e-08,
+      "loss": 1.1943,
+      "step": 22508
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7951091482524285e-08,
+      "loss": 0.9936,
+      "step": 22509
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7877652753651173e-08,
+      "loss": 1.0611,
+      "step": 22510
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7804310358521268e-08,
+      "loss": 0.8979,
+      "step": 22511
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.77310642985551e-08,
+      "loss": 0.955,
+      "step": 22512
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.765791457517153e-08,
+      "loss": 0.9345,
+      "step": 22513
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7584861189787758e-08,
+      "loss": 0.8343,
+      "step": 22514
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7511904143819323e-08,
+      "loss": 0.9465,
+      "step": 22515
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7439043438679537e-08,
+      "loss": 0.9427,
+      "step": 22516
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7366279075781154e-08,
+      "loss": 1.171,
+      "step": 22517
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.72936110565325e-08,
+      "loss": 0.9222,
+      "step": 22518
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.722103938234244e-08,
+      "loss": 0.9598,
+      "step": 22519
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7148564054615967e-08,
+      "loss": 1.0575,
+      "step": 22520
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.707618507475862e-08,
+      "loss": 1.0239,
+      "step": 22521
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.700390244417206e-08,
+      "loss": 0.9055,
+      "step": 22522
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.693171616425738e-08,
+      "loss": 0.9471,
+      "step": 22523
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6859626236411806e-08,
+      "loss": 1.1066,
+      "step": 22524
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.678763266203366e-08,
+      "loss": 0.8944,
+      "step": 22525
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6715735442516267e-08,
+      "loss": 0.9551,
+      "step": 22526
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.664393457925407e-08,
+      "loss": 0.9068,
+      "step": 22527
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6572230073637072e-08,
+      "loss": 0.9534,
+      "step": 22528
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6500621927054716e-08,
+      "loss": 0.9013,
+      "step": 22529
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6429110140894775e-08,
+      "loss": 0.8202,
+      "step": 22530
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6357694716543368e-08,
+      "loss": 0.9275,
+      "step": 22531
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6286375655382722e-08,
+      "loss": 0.9298,
+      "step": 22532
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6215152958795066e-08,
+      "loss": 1.0775,
+      "step": 22533
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6144026628160958e-08,
+      "loss": 1.0259,
+      "step": 22534
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6072996664858186e-08,
+      "loss": 0.8521,
+      "step": 22535
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6002063070262872e-08,
+      "loss": 0.9161,
+      "step": 22536
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.593122584574892e-08,
+      "loss": 0.8994,
+      "step": 22537
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5860484992689117e-08,
+      "loss": 0.8435,
+      "step": 22538
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.578984051245459e-08,
+      "loss": 0.8762,
+      "step": 22539
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5719292406413132e-08,
+      "loss": 1.0422,
+      "step": 22540
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5648840675931986e-08,
+      "loss": 1.0037,
+      "step": 22541
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.557848532237672e-08,
+      "loss": 1.1484,
+      "step": 22542
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5508226347109587e-08,
+      "loss": 0.9216,
+      "step": 22543
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.543806375149227e-08,
+      "loss": 1.0821,
+      "step": 22544
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5367997536883683e-08,
+      "loss": 0.965,
+      "step": 22545
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.529802770464218e-08,
+      "loss": 0.827,
+      "step": 22546
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5228154256122238e-08,
+      "loss": 1.0022,
+      "step": 22547
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.515837719267944e-08,
+      "loss": 0.8779,
+      "step": 22548
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5088696515663814e-08,
+      "loss": 1.1752,
+      "step": 22549
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.50191122264265e-08,
+      "loss": 0.9575,
+      "step": 22550
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.494962432631476e-08,
+      "loss": 0.9249,
+      "step": 22551
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4880232816676398e-08,
+      "loss": 1.1216,
+      "step": 22552
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.481093769885423e-08,
+      "loss": 1.0215,
+      "step": 22553
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4741738974192187e-08,
+      "loss": 0.9354,
+      "step": 22554
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4672636644029747e-08,
+      "loss": 1.1409,
+      "step": 22555
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.460363070970695e-08,
+      "loss": 0.9595,
+      "step": 22556
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.453472117255995e-08,
+      "loss": 0.9795,
+      "step": 22557
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4465908033924347e-08,
+      "loss": 0.8307,
+      "step": 22558
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4397191295132962e-08,
+      "loss": 0.9986,
+      "step": 22559
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4328570957517507e-08,
+      "loss": 0.9615,
+      "step": 22560
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.426004702240692e-08,
+      "loss": 0.9227,
+      "step": 22561
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4191619491129582e-08,
+      "loss": 0.9146,
+      "step": 22562
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4123288365010544e-08,
+      "loss": 0.9477,
+      "step": 22563
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4055053645374304e-08,
+      "loss": 0.8336,
+      "step": 22564
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3986915333542584e-08,
+      "loss": 0.7697,
+      "step": 22565
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3918873430835986e-08,
+      "loss": 1.1153,
+      "step": 22566
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.38509279385718e-08,
+      "loss": 1.0819,
+      "step": 22567
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3783078858067855e-08,
+      "loss": 0.8563,
+      "step": 22568
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3715326190637545e-08,
+      "loss": 0.7848,
+      "step": 22569
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3647669937593708e-08,
+      "loss": 1.0711,
+      "step": 22570
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.358011010024808e-08,
+      "loss": 0.8933,
+      "step": 22571
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.351264667990849e-08,
+      "loss": 0.8513,
+      "step": 22572
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3445279677882792e-08,
+      "loss": 0.9544,
+      "step": 22573
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.337800909547605e-08,
+      "loss": 1.016,
+      "step": 22574
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3310834933991113e-08,
+      "loss": 0.9715,
+      "step": 22575
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3243757194730265e-08,
+      "loss": 0.911,
+      "step": 22576
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3176775878992474e-08,
+      "loss": 0.8693,
+      "step": 22577
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3109890988075035e-08,
+      "loss": 0.8386,
+      "step": 22578
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.304310252327524e-08,
+      "loss": 1.0855,
+      "step": 22579
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2976410485886502e-08,
+      "loss": 0.8533,
+      "step": 22580
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2909814877200563e-08,
+      "loss": 0.9744,
+      "step": 22581
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2843315698508062e-08,
+      "loss": 0.9026,
+      "step": 22582
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2776912951097408e-08,
+      "loss": 0.9143,
+      "step": 22583
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2710606636254795e-08,
+      "loss": 0.9086,
+      "step": 22584
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2644396755265306e-08,
+      "loss": 1.0567,
+      "step": 22585
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2578283309411808e-08,
+      "loss": 0.8598,
+      "step": 22586
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2512266299974384e-08,
+      "loss": 1.0113,
+      "step": 22587
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.244634572823312e-08,
+      "loss": 0.9125,
+      "step": 22588
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2380521595464777e-08,
+      "loss": 0.9691,
+      "step": 22589
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2314793902945e-08,
+      "loss": 1.0416,
+      "step": 22590
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.22491626519461e-08,
+      "loss": 0.9662,
+      "step": 22591
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2183627843741505e-08,
+      "loss": 0.9358,
+      "step": 22592
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.211818947959965e-08,
+      "loss": 0.9009,
+      "step": 22593
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2052847560788403e-08,
+      "loss": 0.8807,
+      "step": 22594
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1987602088573977e-08,
+      "loss": 0.8719,
+      "step": 22595
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1922453064220918e-08,
+      "loss": 0.9128,
+      "step": 22596
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1857400488990986e-08,
+      "loss": 1.025,
+      "step": 22597
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.179244436414485e-08,
+      "loss": 0.8023,
+      "step": 22598
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.172758469094094e-08,
+      "loss": 1.0255,
+      "step": 22599
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1662821470635477e-08,
+      "loss": 0.9452,
+      "step": 22600
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1598154704483564e-08,
+      "loss": 1.0951,
+      "step": 22601
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1533584393738093e-08,
+      "loss": 0.8091,
+      "step": 22602
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1469110539650283e-08,
+      "loss": 0.9212,
+      "step": 22603
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1404733143469137e-08,
+      "loss": 0.9755,
+      "step": 22604
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1340452206441986e-08,
+      "loss": 0.9044,
+      "step": 22605
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1276267729814503e-08,
+      "loss": 0.8501,
+      "step": 22606
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1212179714829583e-08,
+      "loss": 1.0013,
+      "step": 22607
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1148188162729567e-08,
+      "loss": 0.9728,
+      "step": 22608
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.108429307475346e-08,
+      "loss": 0.905,
+      "step": 22609
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1020494452140272e-08,
+      "loss": 0.9109,
+      "step": 22610
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0956792296125127e-08,
+      "loss": 0.8455,
+      "step": 22611
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.089318660794315e-08,
+      "loss": 0.9262,
+      "step": 22612
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.082967738882613e-08,
+      "loss": 0.9312,
+      "step": 22613
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.076626464000475e-08,
+      "loss": 0.9074,
+      "step": 22614
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.070294836270692e-08,
+      "loss": 0.8586,
+      "step": 22615
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0639728558160544e-08,
+      "loss": 0.9139,
+      "step": 22616
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0576605227589643e-08,
+      "loss": 1.0053,
+      "step": 22617
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0513578372218235e-08,
+      "loss": 1.0051,
+      "step": 22618
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0450647993265905e-08,
+      "loss": 0.8294,
+      "step": 22619
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0387814091952786e-08,
+      "loss": 1.0343,
+      "step": 22620
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0325076669496234e-08,
+      "loss": 0.8044,
+      "step": 22621
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.026243572711195e-08,
+      "loss": 0.9829,
+      "step": 22622
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0199891266013404e-08,
+      "loss": 0.9603,
+      "step": 22623
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0137443287412406e-08,
+      "loss": 0.8898,
+      "step": 22624
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.007509179251854e-08,
+      "loss": 0.9801,
+      "step": 22625
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.001283678254029e-08,
+      "loss": 1.0924,
+      "step": 22626
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9950678258683354e-08,
+      "loss": 0.9087,
+      "step": 22627
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.988861622215288e-08,
+      "loss": 1.059,
+      "step": 22628
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.982665067415068e-08,
+      "loss": 0.8541,
+      "step": 22629
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9764781615876916e-08,
+      "loss": 0.8859,
+      "step": 22630
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.970300904853173e-08,
+      "loss": 1.0451,
+      "step": 22631
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.964133297331028e-08,
+      "loss": 0.731,
+      "step": 22632
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9579753391408275e-08,
+      "loss": 1.1064,
+      "step": 22633
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9518270304019206e-08,
+      "loss": 0.9939,
+      "step": 22634
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9456883712333784e-08,
+      "loss": 0.934,
+      "step": 22635
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9395593617541063e-08,
+      "loss": 0.8197,
+      "step": 22636
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9334400020829534e-08,
+      "loss": 0.9479,
+      "step": 22637
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.927330292338381e-08,
+      "loss": 0.8998,
+      "step": 22638
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9212302326387934e-08,
+      "loss": 0.8593,
+      "step": 22639
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9151398231024297e-08,
+      "loss": 0.9937,
+      "step": 22640
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9090590638471962e-08,
+      "loss": 0.9515,
+      "step": 22641
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.902987954990998e-08,
+      "loss": 0.8753,
+      "step": 22642
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.896926496651408e-08,
+      "loss": 0.7473,
+      "step": 22643
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8908746889458875e-08,
+      "loss": 0.8802,
+      "step": 22644
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.884832531991676e-08,
+      "loss": 0.8543,
+      "step": 22645
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8788000259057916e-08,
+      "loss": 0.7783,
+      "step": 22646
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.872777170805251e-08,
+      "loss": 0.7416,
+      "step": 22647
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8667639668065728e-08,
+      "loss": 0.9668,
+      "step": 22648
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8607604140263858e-08,
+      "loss": 0.9256,
+      "step": 22649
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8547665125809854e-08,
+      "loss": 0.9229,
+      "step": 22650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8487822625865014e-08,
+      "loss": 0.9095,
+      "step": 22651
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8428076641587855e-08,
+      "loss": 0.6876,
+      "step": 22652
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.836842717413745e-08,
+      "loss": 1.027,
+      "step": 22653
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8308874224668427e-08,
+      "loss": 0.7904,
+      "step": 22654
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8249417794334866e-08,
+      "loss": 0.9518,
+      "step": 22655
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.819005788428918e-08,
+      "loss": 0.835,
+      "step": 22656
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8130794495680447e-08,
+      "loss": 0.8984,
+      "step": 22657
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8071627629657752e-08,
+      "loss": 1.0543,
+      "step": 22658
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8012557287367394e-08,
+      "loss": 0.725,
+      "step": 22659
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7953583469953463e-08,
+      "loss": 1.0387,
+      "step": 22660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.789470617855893e-08,
+      "loss": 1.0711,
+      "step": 22661
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7835925414323997e-08,
+      "loss": 0.9203,
+      "step": 22662
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7777241178388304e-08,
+      "loss": 0.9994,
+      "step": 22663
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.771865347188817e-08,
+      "loss": 1.0692,
+      "step": 22664
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7660162295958794e-08,
+      "loss": 0.9683,
+      "step": 22665
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7601767651733715e-08,
+      "loss": 0.921,
+      "step": 22666
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7543469540344248e-08,
+      "loss": 0.8751,
+      "step": 22667
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.748526796292005e-08,
+      "loss": 0.9478,
+      "step": 22668
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.742716292058799e-08,
+      "loss": 0.9086,
+      "step": 22669
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7369154414474954e-08,
+      "loss": 0.8788,
+      "step": 22670
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7311242445703926e-08,
+      "loss": 0.9587,
+      "step": 22671
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7253427015397338e-08,
+      "loss": 0.8943,
+      "step": 22672
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7195708124675415e-08,
+      "loss": 1.0183,
+      "step": 22673
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.71380857746567e-08,
+      "loss": 1.0031,
+      "step": 22674
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7080559966456967e-08,
+      "loss": 0.841,
+      "step": 22675
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7023130701190882e-08,
+      "loss": 0.9376,
+      "step": 22676
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6965797979971442e-08,
+      "loss": 0.9063,
+      "step": 22677
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.690856180390943e-08,
+      "loss": 0.9893,
+      "step": 22678
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6851422174113398e-08,
+      "loss": 0.8722,
+      "step": 22679
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.679437909169135e-08,
+      "loss": 0.9503,
+      "step": 22680
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.67374325577474e-08,
+      "loss": 0.9573,
+      "step": 22681
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6680582573385118e-08,
+      "loss": 0.8697,
+      "step": 22682
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6623829139706394e-08,
+      "loss": 0.9358,
+      "step": 22683
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.656717225781035e-08,
+      "loss": 1.0347,
+      "step": 22684
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6510611928794996e-08,
+      "loss": 0.9594,
+      "step": 22685
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.645414815375612e-08,
+      "loss": 0.8882,
+      "step": 22686
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6397780933787853e-08,
+      "loss": 0.9466,
+      "step": 22687
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6341510269982098e-08,
+      "loss": 0.8771,
+      "step": 22688
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.628533616342909e-08,
+      "loss": 0.7917,
+      "step": 22689
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6229258615216847e-08,
+      "loss": 1.0496,
+      "step": 22690
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.617327762643284e-08,
+      "loss": 1.1054,
+      "step": 22691
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6117393198160636e-08,
+      "loss": 0.9123,
+      "step": 22692
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6061605331483265e-08,
+      "loss": 0.9045,
+      "step": 22693
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6005914027481528e-08,
+      "loss": 1.0039,
+      "step": 22694
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5950319287235116e-08,
+      "loss": 0.7601,
+      "step": 22695
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.589482111182039e-08,
+      "loss": 0.9656,
+      "step": 22696
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5839419502313714e-08,
+      "loss": 0.9448,
+      "step": 22697
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5784114459787004e-08,
+      "loss": 0.5466,
+      "step": 22698
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5728905985312182e-08,
+      "loss": 0.9757,
+      "step": 22699
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5673794079959502e-08,
+      "loss": 0.8742,
+      "step": 22700
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5618778744797e-08,
+      "loss": 0.9941,
+      "step": 22701
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5563859980889383e-08,
+      "loss": 0.7268,
+      "step": 22702
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.550903778930135e-08,
+      "loss": 0.9169,
+      "step": 22703
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5454312171095385e-08,
+      "loss": 0.9272,
+      "step": 22704
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.53996831273312e-08,
+      "loss": 0.9994,
+      "step": 22705
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.534515065906683e-08,
+      "loss": 0.8877,
+      "step": 22706
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5290714767360326e-08,
+      "loss": 0.8355,
+      "step": 22707
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5236375453265285e-08,
+      "loss": 0.9717,
+      "step": 22708
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5182132717834753e-08,
+      "loss": 1.0342,
+      "step": 22709
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5127986562119556e-08,
+      "loss": 1.0939,
+      "step": 22710
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5073936987168857e-08,
+      "loss": 0.9456,
+      "step": 22711
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.501998399403015e-08,
+      "loss": 0.9037,
+      "step": 22712
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4966127583748158e-08,
+      "loss": 1.0347,
+      "step": 22713
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4912367757366485e-08,
+      "loss": 0.982,
+      "step": 22714
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4858704515927635e-08,
+      "loss": 0.8674,
+      "step": 22715
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4805137860469666e-08,
+      "loss": 1.061,
+      "step": 22716
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4751667792031743e-08,
+      "loss": 0.8017,
+      "step": 22717
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4698294311649153e-08,
+      "loss": 0.7547,
+      "step": 22718
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4645017420356622e-08,
+      "loss": 1.0443,
+      "step": 22719
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4591837119186104e-08,
+      "loss": 0.894,
+      "step": 22720
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4538753409167328e-08,
+      "loss": 0.9267,
+      "step": 22721
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4485766291330029e-08,
+      "loss": 1.103,
+      "step": 22722
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4432875766699495e-08,
+      "loss": 1.0257,
+      "step": 22723
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4380081836301018e-08,
+      "loss": 0.9504,
+      "step": 22724
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4327384501158227e-08,
+      "loss": 0.8982,
+      "step": 22725
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4274783762290855e-08,
+      "loss": 1.0049,
+      "step": 22726
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4222279620718649e-08,
+      "loss": 0.861,
+      "step": 22727
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4169872077459124e-08,
+      "loss": 0.9091,
+      "step": 22728
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4117561133526469e-08,
+      "loss": 0.8694,
+      "step": 22729
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4065346789935985e-08,
+      "loss": 0.9242,
+      "step": 22730
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4013229047697973e-08,
+      "loss": 0.9278,
+      "step": 22731
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3961207907822738e-08,
+      "loss": 0.9982,
+      "step": 22732
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3909283371318361e-08,
+      "loss": 0.9858,
+      "step": 22733
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3857455439190149e-08,
+      "loss": 1.0012,
+      "step": 22734
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3805724112442854e-08,
+      "loss": 0.9574,
+      "step": 22735
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3754089392079007e-08,
+      "loss": 0.9426,
+      "step": 22736
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3702551279098363e-08,
+      "loss": 0.7623,
+      "step": 22737
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3651109774499571e-08,
+      "loss": 0.8068,
+      "step": 22738
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3599764879279053e-08,
+      "loss": 0.8605,
+      "step": 22739
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3548516594432682e-08,
+      "loss": 0.8808,
+      "step": 22740
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3497364920952437e-08,
+      "loss": 1.0517,
+      "step": 22741
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3446309859829199e-08,
+      "loss": 0.9593,
+      "step": 22742
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3395351412053282e-08,
+      "loss": 0.9939,
+      "step": 22743
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3344489578610564e-08,
+      "loss": 0.8473,
+      "step": 22744
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3293724360488037e-08,
+      "loss": 0.6807,
+      "step": 22745
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3243055758667688e-08,
+      "loss": 1.1012,
+      "step": 22746
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.319248377413207e-08,
+      "loss": 0.8584,
+      "step": 22747
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3142008407861507e-08,
+      "loss": 0.8687,
+      "step": 22748
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3091629660832439e-08,
+      "loss": 0.6427,
+      "step": 22749
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3041347534022419e-08,
+      "loss": 0.9004,
+      "step": 22750
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2991162028405113e-08,
+      "loss": 0.7894,
+      "step": 22751
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.294107314495252e-08,
+      "loss": 1.0293,
+      "step": 22752
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2891080884634977e-08,
+      "loss": 0.8771,
+      "step": 22753
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2841185248422261e-08,
+      "loss": 0.9192,
+      "step": 22754
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2791386237279713e-08,
+      "loss": 0.8921,
+      "step": 22755
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2741683852173225e-08,
+      "loss": 1.046,
+      "step": 22756
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2692078094065363e-08,
+      "loss": 0.8807,
+      "step": 22757
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2642568963916469e-08,
+      "loss": 0.9162,
+      "step": 22758
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2593156462687439e-08,
+      "loss": 0.7621,
+      "step": 22759
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2543840591333622e-08,
+      "loss": 0.8714,
+      "step": 22760
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2494621350812585e-08,
+      "loss": 0.7412,
+      "step": 22761
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2445498742076345e-08,
+      "loss": 0.8489,
+      "step": 22762
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2396472766077473e-08,
+      "loss": 0.9579,
+      "step": 22763
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2347543423764652e-08,
+      "loss": 1.0599,
+      "step": 22764
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.229871071608768e-08,
+      "loss": 0.8758,
+      "step": 22765
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2249974643991912e-08,
+      "loss": 1.0613,
+      "step": 22766
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2201335208421038e-08,
+      "loss": 1.0887,
+      "step": 22767
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2152792410317637e-08,
+      "loss": 1.0111,
+      "step": 22768
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2104346250623177e-08,
+      "loss": 0.8904,
+      "step": 22769
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2055996730275243e-08,
+      "loss": 0.8587,
+      "step": 22770
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2007743850210863e-08,
+      "loss": 1.0786,
+      "step": 22771
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1959587611364843e-08,
+      "loss": 0.8255,
+      "step": 22772
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1911528014670881e-08,
+      "loss": 1.0314,
+      "step": 22773
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.186356506105879e-08,
+      "loss": 0.8722,
+      "step": 22774
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1815698751458937e-08,
+      "loss": 0.8396,
+      "step": 22775
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1767929086798913e-08,
+      "loss": 0.9628,
+      "step": 22776
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1720256068002977e-08,
+      "loss": 1.002,
+      "step": 22777
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1672679695995948e-08,
+      "loss": 0.8488,
+      "step": 22778
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1625199971699308e-08,
+      "loss": 0.8557,
+      "step": 22779
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1577816896032879e-08,
+      "loss": 0.9298,
+      "step": 22780
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1530530469914258e-08,
+      "loss": 0.9836,
+      "step": 22781
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1483340694260492e-08,
+      "loss": 1.0093,
+      "step": 22782
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1436247569985293e-08,
+      "loss": 0.8832,
+      "step": 22783
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1389251098000709e-08,
+      "loss": 1.0352,
+      "step": 22784
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1342351279218233e-08,
+      "loss": 0.839,
+      "step": 22785
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.129554811454603e-08,
+      "loss": 0.8246,
+      "step": 22786
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1248841604891148e-08,
+      "loss": 0.9331,
+      "step": 22787
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1202231751157866e-08,
+      "loss": 0.9613,
+      "step": 22788
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1155718554249906e-08,
+      "loss": 0.8212,
+      "step": 22789
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1109302015068768e-08,
+      "loss": 0.8512,
+      "step": 22790
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1062982134512623e-08,
+      "loss": 0.9869,
+      "step": 22791
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1016758913479642e-08,
+      "loss": 0.9674,
+      "step": 22792
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0970632352865219e-08,
+      "loss": 0.8469,
+      "step": 22793
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0924602453563638e-08,
+      "loss": 0.9145,
+      "step": 22794
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0878669216465298e-08,
+      "loss": 0.8372,
+      "step": 22795
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.083283264246171e-08,
+      "loss": 0.8774,
+      "step": 22796
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0787092732439941e-08,
+      "loss": 0.935,
+      "step": 22797
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0741449487286503e-08,
+      "loss": 1.0348,
+      "step": 22798
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.069590290788569e-08,
+      "loss": 0.8198,
+      "step": 22799
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.065045299512013e-08,
+      "loss": 0.9442,
+      "step": 22800
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.060509974987023e-08,
+      "loss": 0.9299,
+      "step": 22801
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0559843173014728e-08,
+      "loss": 0.7411,
+      "step": 22802
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0514683265430147e-08,
+      "loss": 0.9672,
+      "step": 22803
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0469620027992455e-08,
+      "loss": 0.8654,
+      "step": 22804
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0424653461573731e-08,
+      "loss": 1.0356,
+      "step": 22805
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0379783567045498e-08,
+      "loss": 0.8678,
+      "step": 22806
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0335010345277064e-08,
+      "loss": 1.0408,
+      "step": 22807
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0290333797136065e-08,
+      "loss": 0.9288,
+      "step": 22808
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0245753923487921e-08,
+      "loss": 0.9257,
+      "step": 22809
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0201270725196388e-08,
+      "loss": 1.0526,
+      "step": 22810
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0156884203122996e-08,
+      "loss": 0.8743,
+      "step": 22811
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0112594358128725e-08,
+      "loss": 0.8185,
+      "step": 22812
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0068401191070665e-08,
+      "loss": 1.012,
+      "step": 22813
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.002430470280591e-08,
+      "loss": 0.798,
+      "step": 22814
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.980304894187665e-09,
+      "loss": 0.8026,
+      "step": 22815
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.936401766069693e-09,
+      "loss": 0.933,
+      "step": 22816
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.892595319301867e-09,
+      "loss": 1.0232,
+      "step": 22817
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.848885554732957e-09,
+      "loss": 0.9115,
+      "step": 22818
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.805272473210058e-09,
+      "loss": 0.9085,
+      "step": 22819
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.761756075577499e-09,
+      "loss": 0.9127,
+      "step": 22820
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.718336362679604e-09,
+      "loss": 0.9612,
+      "step": 22821
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.675013335356254e-09,
+      "loss": 0.9954,
+      "step": 22822
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.631786994448445e-09,
+      "loss": 0.826,
+      "step": 22823
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.588657340792173e-09,
+      "loss": 0.7295,
+      "step": 22824
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.545624375223439e-09,
+      "loss": 0.9542,
+      "step": 22825
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.502688098576573e-09,
+      "loss": 1.0186,
+      "step": 22826
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.459848511683133e-09,
+      "loss": 1.0307,
+      "step": 22827
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.417105615373013e-09,
+      "loss": 1.082,
+      "step": 22828
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.374459410474435e-09,
+      "loss": 0.9155,
+      "step": 22829
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.331909897813407e-09,
+      "loss": 0.9929,
+      "step": 22830
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.289457078214825e-09,
+      "loss": 0.8839,
+      "step": 22831
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.24710095250081e-09,
+      "loss": 0.9816,
+      "step": 22832
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.204841521492369e-09,
+      "loss": 0.9545,
+      "step": 22833
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.16267878600774e-09,
+      "loss": 0.6884,
+      "step": 22834
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.120612746864043e-09,
+      "loss": 1.0484,
+      "step": 22835
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.078643404876741e-09,
+      "loss": 0.8205,
+      "step": 22836
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.03677076085796e-09,
+      "loss": 0.961,
+      "step": 22837
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.994994815620383e-09,
+      "loss": 0.9253,
+      "step": 22838
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.953315569972254e-09,
+      "loss": 0.8886,
+      "step": 22839
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.911733024721259e-09,
+      "loss": 0.9849,
+      "step": 22840
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.870247180673418e-09,
+      "loss": 0.9014,
+      "step": 22841
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.828858038632538e-09,
+      "loss": 0.9351,
+      "step": 22842
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.787565599400194e-09,
+      "loss": 0.9306,
+      "step": 22843
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.746369863776305e-09,
+      "loss": 0.8568,
+      "step": 22844
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.705270832559676e-09,
+      "loss": 0.8562,
+      "step": 22845
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.664268506546336e-09,
+      "loss": 0.9005,
+      "step": 22846
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.623362886530095e-09,
+      "loss": 0.9888,
+      "step": 22847
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.582553973304208e-09,
+      "loss": 1.0504,
+      "step": 22848
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.541841767659154e-09,
+      "loss": 1.0116,
+      "step": 22849
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.501226270383745e-09,
+      "loss": 0.8641,
+      "step": 22850
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.460707482264574e-09,
+      "loss": 1.0823,
+      "step": 22851
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.420285404087124e-09,
+      "loss": 0.9324,
+      "step": 22852
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.379960036634106e-09,
+      "loss": 0.9915,
+      "step": 22853
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.339731380687666e-09,
+      "loss": 0.8144,
+      "step": 22854
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.299599437026073e-09,
+      "loss": 0.892,
+      "step": 22855
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.259564206427595e-09,
+      "loss": 0.9526,
+      "step": 22856
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.21962568966772e-09,
+      "loss": 0.9868,
+      "step": 22857
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.179783887520277e-09,
+      "loss": 0.9598,
+      "step": 22858
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.140038800756867e-09,
+      "loss": 0.9863,
+      "step": 22859
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.100390430148541e-09,
+      "loss": 0.8411,
+      "step": 22860
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.060838776462465e-09,
+      "loss": 0.9049,
+      "step": 22861
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.021383840465247e-09,
+      "loss": 0.8864,
+      "step": 22862
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.982025622921274e-09,
+      "loss": 0.9735,
+      "step": 22863
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.942764124593272e-09,
+      "loss": 1.1117,
+      "step": 22864
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.903599346241741e-09,
+      "loss": 0.8028,
+      "step": 22865
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.864531288625521e-09,
+      "loss": 0.7665,
+      "step": 22866
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.825559952501783e-09,
+      "loss": 1.1021,
+      "step": 22867
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.786685338625478e-09,
+      "loss": 1.1072,
+      "step": 22868
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.747907447749336e-09,
+      "loss": 0.9013,
+      "step": 22869
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.709226280625537e-09,
+      "loss": 0.8366,
+      "step": 22870
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.670641838002924e-09,
+      "loss": 0.99,
+      "step": 22871
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.632154120628676e-09,
+      "loss": 1.0236,
+      "step": 22872
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.593763129249421e-09,
+      "loss": 0.9355,
+      "step": 22873
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.55546886460845e-09,
+      "loss": 0.8217,
+      "step": 22874
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.51727132744795e-09,
+      "loss": 1.0559,
+      "step": 22875
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.479170518507883e-09,
+      "loss": 0.8405,
+      "step": 22876
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.441166438525993e-09,
+      "loss": 0.9089,
+      "step": 22877
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.403259088239467e-09,
+      "loss": 1.0074,
+      "step": 22878
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.365448468382163e-09,
+      "loss": 0.7859,
+      "step": 22879
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.327734579686274e-09,
+      "loss": 0.8342,
+      "step": 22880
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.290117422883436e-09,
+      "loss": 1.0668,
+      "step": 22881
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.252596998701955e-09,
+      "loss": 1.015,
+      "step": 22882
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.215173307869028e-09,
+      "loss": 0.746,
+      "step": 22883
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.177846351109629e-09,
+      "loss": 0.9423,
+      "step": 22884
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.140616129146516e-09,
+      "loss": 0.8908,
+      "step": 22885
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.103482642701331e-09,
+      "loss": 0.9112,
+      "step": 22886
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.066445892494056e-09,
+      "loss": 1.0004,
+      "step": 22887
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.029505879241894e-09,
+      "loss": 1.006,
+      "step": 22888
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.992662603659828e-09,
+      "loss": 1.0211,
+      "step": 22889
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.955916066462842e-09,
+      "loss": 0.9982,
+      "step": 22890
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.919266268362035e-09,
+      "loss": 0.929,
+      "step": 22891
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.8827132100679485e-09,
+      "loss": 0.8848,
+      "step": 22892
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.846256892288905e-09,
+      "loss": 0.8329,
+      "step": 22893
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.809897315730451e-09,
+      "loss": 0.9322,
+      "step": 22894
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.773634481098135e-09,
+      "loss": 0.9254,
+      "step": 22895
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.737468389093615e-09,
+      "loss": 0.9061,
+      "step": 22896
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.701399040418555e-09,
+      "loss": 0.9091,
+      "step": 22897
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.665426435770727e-09,
+      "loss": 0.8964,
+      "step": 22898
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.629550575847355e-09,
+      "loss": 0.8715,
+      "step": 22899
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.59377146134399e-09,
+      "loss": 0.9264,
+      "step": 22900
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.558089092953967e-09,
+      "loss": 0.8238,
+      "step": 22901
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.522503471367847e-09,
+      "loss": 0.9111,
+      "step": 22902
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.487014597275631e-09,
+      "loss": 0.9394,
+      "step": 22903
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.451622471365104e-09,
+      "loss": 0.8936,
+      "step": 22904
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.416327094321273e-09,
+      "loss": 0.861,
+      "step": 22905
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.38112846682859e-09,
+      "loss": 1.0111,
+      "step": 22906
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.346026589568733e-09,
+      "loss": 1.0082,
+      "step": 22907
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.311021463221712e-09,
+      "loss": 0.9259,
+      "step": 22908
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.276113088465874e-09,
+      "loss": 0.8292,
+      "step": 22909
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.2413014659779005e-09,
+      "loss": 0.9263,
+      "step": 22910
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.206586596431696e-09,
+      "loss": 0.9658,
+      "step": 22911
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.171968480500057e-09,
+      "loss": 0.9602,
+      "step": 22912
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.137447118853556e-09,
+      "loss": 0.9411,
+      "step": 22913
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.10302251216166e-09,
+      "loss": 0.8138,
+      "step": 22914
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.068694661090502e-09,
+      "loss": 1.0218,
+      "step": 22915
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.034463566306215e-09,
+      "loss": 1.1413,
+      "step": 22916
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.000329228470492e-09,
+      "loss": 0.753,
+      "step": 22917
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.966291648246136e-09,
+      "loss": 0.9215,
+      "step": 22918
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.9323508262920656e-09,
+      "loss": 0.8766,
+      "step": 22919
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.898506763265532e-09,
+      "loss": 0.9547,
+      "step": 22920
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.864759459822677e-09,
+      "loss": 0.954,
+      "step": 22921
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.831108916617423e-09,
+      "loss": 0.9278,
+      "step": 22922
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.797555134301469e-09,
+      "loss": 0.7951,
+      "step": 22923
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.764098113525407e-09,
+      "loss": 0.9689,
+      "step": 22924
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.7307378549364966e-09,
+      "loss": 0.8628,
+      "step": 22925
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.697474359181998e-09,
+      "loss": 0.7951,
+      "step": 22926
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.66430762690584e-09,
+      "loss": 0.9105,
+      "step": 22927
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.6312376587502885e-09,
+      "loss": 0.8234,
+      "step": 22928
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.598264455357605e-09,
+      "loss": 0.8677,
+      "step": 22929
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.565388017365059e-09,
+      "loss": 0.9723,
+      "step": 22930
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.532608345410473e-09,
+      "loss": 1.0452,
+      "step": 22931
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.4999254401283396e-09,
+      "loss": 0.8088,
+      "step": 22932
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.467339302152597e-09,
+      "loss": 0.856,
+      "step": 22933
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.43484993211385e-09,
+      "loss": 1.0021,
+      "step": 22934
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.4024573306421526e-09,
+      "loss": 0.833,
+      "step": 22935
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.3701614983647786e-09,
+      "loss": 1.0736,
+      "step": 22936
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.3379624359073404e-09,
+      "loss": 0.958,
+      "step": 22937
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.3058601438937814e-09,
+      "loss": 1.086,
+      "step": 22938
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.273854622946939e-09,
+      "loss": 0.8182,
+      "step": 22939
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.241945873685761e-09,
+      "loss": 1.0183,
+      "step": 22940
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.210133896728642e-09,
+      "loss": 0.9251,
+      "step": 22941
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.1784186926923105e-09,
+      "loss": 0.9975,
+      "step": 22942
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.146800262191276e-09,
+      "loss": 0.9823,
+      "step": 22943
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.115278605837826e-09,
+      "loss": 0.7611,
+      "step": 22944
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0838537242431375e-09,
+      "loss": 1.1052,
+      "step": 22945
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.052525618015613e-09,
+      "loss": 0.8431,
+      "step": 22946
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0212942877619905e-09,
+      "loss": 0.8913,
+      "step": 22947
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.990159734087896e-09,
+      "loss": 1.0378,
+      "step": 22948
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.959121957596736e-09,
+      "loss": 0.8225,
+      "step": 22949
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.9281809588891394e-09,
+      "loss": 0.8647,
+      "step": 22950
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.897336738565184e-09,
+      "loss": 1.0275,
+      "step": 22951
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.866589297222168e-09,
+      "loss": 0.9118,
+      "step": 22952
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.835938635455728e-09,
+      "loss": 0.8237,
+      "step": 22953
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.805384753860387e-09,
+      "loss": 0.8113,
+      "step": 22954
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7749276530273394e-09,
+      "loss": 0.8746,
+      "step": 22955
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.744567333546668e-09,
+      "loss": 1.0581,
+      "step": 22956
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7143037960073465e-09,
+      "loss": 0.9146,
+      "step": 22957
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.684137040995018e-09,
+      "loss": 0.7895,
+      "step": 22958
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.654067069094214e-09,
+      "loss": 0.8354,
+      "step": 22959
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.624093880888358e-09,
+      "loss": 1.0138,
+      "step": 22960
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.594217476956986e-09,
+      "loss": 0.9112,
+      "step": 22961
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.564437857879633e-09,
+      "loss": 0.8746,
+      "step": 22962
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.534755024233062e-09,
+      "loss": 0.8893,
+      "step": 22963
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.505168976592922e-09,
+      "loss": 1.0178,
+      "step": 22964
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.475679715531533e-09,
+      "loss": 0.9982,
+      "step": 22965
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4462872416206615e-09,
+      "loss": 0.948,
+      "step": 22966
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.416991555429851e-09,
+      "loss": 0.8611,
+      "step": 22967
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.387792657526424e-09,
+      "loss": 1.1409,
+      "step": 22968
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3586905484765965e-09,
+      "loss": 0.9948,
+      "step": 22969
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.329685228844361e-09,
+      "loss": 0.8238,
+      "step": 22970
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.300776699190379e-09,
+      "loss": 0.9147,
+      "step": 22971
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2719649600764244e-09,
+      "loss": 0.9745,
+      "step": 22972
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.243250012059275e-09,
+      "loss": 0.8646,
+      "step": 22973
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.21463185569626e-09,
+      "loss": 0.8306,
+      "step": 22974
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.186110491541384e-09,
+      "loss": 0.7379,
+      "step": 22975
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.157685920147536e-09,
+      "loss": 0.8559,
+      "step": 22976
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1293581420653876e-09,
+      "loss": 0.8779,
+      "step": 22977
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1011271578433876e-09,
+      "loss": 1.0807,
+      "step": 22978
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.072992968028877e-09,
+      "loss": 0.8693,
+      "step": 22979
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0449555731669755e-09,
+      "loss": 0.9453,
+      "step": 22980
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.017014973800581e-09,
+      "loss": 0.8585,
+      "step": 22981
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9891711704714845e-09,
+      "loss": 0.7742,
+      "step": 22982
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.961424163718696e-09,
+      "loss": 1.0312,
+      "step": 22983
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.933773954080122e-09,
+      "loss": 0.8552,
+      "step": 22984
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.906220542091443e-09,
+      "loss": 0.8813,
+      "step": 22985
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.878763928286677e-09,
+      "loss": 0.8769,
+      "step": 22986
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.851404113197066e-09,
+      "loss": 0.9255,
+      "step": 22987
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.824141097353295e-09,
+      "loss": 0.9011,
+      "step": 22988
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.796974881283833e-09,
+      "loss": 0.9449,
+      "step": 22989
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.769905465514368e-09,
+      "loss": 1.0045,
+      "step": 22990
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.742932850569481e-09,
+      "loss": 0.8383,
+      "step": 22991
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7160570369720873e-09,
+      "loss": 1.0781,
+      "step": 22992
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.689278025242882e-09,
+      "loss": 1.2448,
+      "step": 22993
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6625958159003383e-09,
+      "loss": 0.8891,
+      "step": 22994
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.636010409461821e-09,
+      "loss": 0.9232,
+      "step": 22995
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.609521806441918e-09,
+      "loss": 0.9656,
+      "step": 22996
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.583130007354663e-09,
+      "loss": 0.767,
+      "step": 22997
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5568350127102025e-09,
+      "loss": 1.0399,
+      "step": 22998
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.53063682301924e-09,
+      "loss": 0.9197,
+      "step": 22999
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5045354387885923e-09,
+      "loss": 0.7704,
+      "step": 23000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4785308605245207e-09,
+      "loss": 0.8931,
+      "step": 23001
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.452623088729956e-09,
+      "loss": 0.7271,
+      "step": 23002
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.42681212390783e-09,
+      "loss": 0.938,
+      "step": 23003
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4010979665577425e-09,
+      "loss": 0.9666,
+      "step": 23004
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.375480617177629e-09,
+      "loss": 0.9651,
+      "step": 23005
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3499600762643136e-09,
+      "loss": 0.9338,
+      "step": 23006
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.324536344312401e-09,
+      "loss": 0.7695,
+      "step": 23007
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2992094218137206e-09,
+      "loss": 0.994,
+      "step": 23008
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2739793092601e-09,
+      "loss": 0.8075,
+      "step": 23009
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2488460071389285e-09,
+      "loss": 0.8438,
+      "step": 23010
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2238095159381476e-09,
+      "loss": 1.0633,
+      "step": 23011
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.198869836142926e-09,
+      "loss": 0.8285,
+      "step": 23012
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.174026968235655e-09,
+      "loss": 0.9641,
+      "step": 23013
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1492809126987266e-09,
+      "loss": 0.8207,
+      "step": 23014
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.124631670010647e-09,
+      "loss": 0.8426,
+      "step": 23015
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1000792406493672e-09,
+      "loss": 1.1049,
+      "step": 23016
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0756236250900627e-09,
+      "loss": 1.0244,
+      "step": 23017
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.051264823807909e-09,
+      "loss": 0.9514,
+      "step": 23018
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0270028372736402e-09,
+      "loss": 0.8931,
+      "step": 23019
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0028376659574365e-09,
+      "loss": 0.7922,
+      "step": 23020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9787693103278113e-09,
+      "loss": 1.0644,
+      "step": 23021
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.954797770851614e-09,
+      "loss": 1.0271,
+      "step": 23022
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.930923047992362e-09,
+      "loss": 0.951,
+      "step": 23023
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9071451422124642e-09,
+      "loss": 0.8331,
+      "step": 23024
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8834640539737723e-09,
+      "loss": 0.8598,
+      "step": 23025
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.859879783734254e-09,
+      "loss": 0.9857,
+      "step": 23026
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.836392331950766e-09,
+      "loss": 0.744,
+      "step": 23027
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.813001699078499e-09,
+      "loss": 0.8905,
+      "step": 23028
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7897078855715353e-09,
+      "loss": 0.9195,
+      "step": 23029
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7665108918795146e-09,
+      "loss": 0.8851,
+      "step": 23030
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7434107184531876e-09,
+      "loss": 0.8775,
+      "step": 23031
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.720407365739974e-09,
+      "loss": 1.0434,
+      "step": 23032
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.697500834185074e-09,
+      "loss": 0.979,
+      "step": 23033
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6746911242325757e-09,
+      "loss": 0.9961,
+      "step": 23034
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6519782363243483e-09,
+      "loss": 0.8967,
+      "step": 23035
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6293621709000404e-09,
+      "loss": 0.9217,
+      "step": 23036
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.606842928398745e-09,
+      "loss": 0.914,
+      "step": 23037
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5844205092562245e-09,
+      "loss": 1.06,
+      "step": 23038
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5620949139065767e-09,
+      "loss": 0.944,
+      "step": 23039
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.539866142783343e-09,
+      "loss": 0.9457,
+      "step": 23040
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5177341963161795e-09,
+      "loss": 1.0177,
+      "step": 23041
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.495699074934188e-09,
+      "loss": 0.9433,
+      "step": 23042
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.473760779064804e-09,
+      "loss": 0.7142,
+      "step": 23043
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4519193091321337e-09,
+      "loss": 1.0232,
+      "step": 23044
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.430174665560281e-09,
+      "loss": 0.9048,
+      "step": 23045
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.408526848770021e-09,
+      "loss": 0.905,
+      "step": 23046
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3869758591810177e-09,
+      "loss": 0.8304,
+      "step": 23047
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.365521697210715e-09,
+      "loss": 0.8351,
+      "step": 23048
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3441643632748923e-09,
+      "loss": 0.8886,
+      "step": 23049
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3229038577871067e-09,
+      "loss": 0.9476,
+      "step": 23050
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.301740181159251e-09,
+      "loss": 0.8371,
+      "step": 23051
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.280673333801553e-09,
+      "loss": 0.8436,
+      "step": 23052
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2597033161220196e-09,
+      "loss": 1.0631,
+      "step": 23053
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.238830128527547e-09,
+      "loss": 0.9501,
+      "step": 23054
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.218053771421702e-09,
+      "loss": 0.8303,
+      "step": 23055
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1973742452074954e-09,
+      "loss": 0.9783,
+      "step": 23056
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1767915502851622e-09,
+      "loss": 0.9702,
+      "step": 23057
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.156305687054383e-09,
+      "loss": 0.8469,
+      "step": 23058
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1359166559109524e-09,
+      "loss": 1.0584,
+      "step": 23059
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1156244572506644e-09,
+      "loss": 0.7369,
+      "step": 23060
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0954290914665386e-09,
+      "loss": 0.9691,
+      "step": 23061
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0753305589493732e-09,
+      "loss": 0.852,
+      "step": 23062
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0553288600894117e-09,
+      "loss": 0.9546,
+      "step": 23063
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0354239952735665e-09,
+      "loss": 0.9764,
+      "step": 23064
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0156159648876406e-09,
+      "loss": 0.9649,
+      "step": 23065
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9959047693152157e-09,
+      "loss": 0.902,
+      "step": 23066
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9762904089382086e-09,
+      "loss": 0.9865,
+      "step": 23067
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9567728841374256e-09,
+      "loss": 0.9952,
+      "step": 23068
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.937352195290343e-09,
+      "loss": 0.876,
+      "step": 23069
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.918028342772771e-09,
+      "loss": 0.9342,
+      "step": 23070
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.89880132696052e-09,
+      "loss": 0.9045,
+      "step": 23071
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8796711482244046e-09,
+      "loss": 0.784,
+      "step": 23072
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8606378069363495e-09,
+      "loss": 0.9066,
+      "step": 23073
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8417013034643939e-09,
+      "loss": 1.0541,
+      "step": 23074
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8228616381760212e-09,
+      "loss": 0.9839,
+      "step": 23075
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8041188114359399e-09,
+      "loss": 0.84,
+      "step": 23076
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7854728236071928e-09,
+      "loss": 0.9761,
+      "step": 23077
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7669236750506024e-09,
+      "loss": 1.0349,
+      "step": 23078
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7484713661269914e-09,
+      "loss": 0.8268,
+      "step": 23079
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.730115897192186e-09,
+      "loss": 0.9853,
+      "step": 23080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.711857268603123e-09,
+      "loss": 0.8842,
+      "step": 23081
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6936954807128537e-09,
+      "loss": 0.8572,
+      "step": 23082
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6756305338733181e-09,
+      "loss": 1.0618,
+      "step": 23083
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.657662428434792e-09,
+      "loss": 1.0188,
+      "step": 23084
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.639791164744775e-09,
+      "loss": 1.012,
+      "step": 23085
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.622016743150212e-09,
+      "loss": 0.8796,
+      "step": 23086
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6043391639958273e-09,
+      "loss": 1.1105,
+      "step": 23087
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.586758427623014e-09,
+      "loss": 0.8538,
+      "step": 23088
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5692745343726112e-09,
+      "loss": 0.9104,
+      "step": 23089
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5518874845843469e-09,
+      "loss": 0.989,
+      "step": 23090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5345972785940632e-09,
+      "loss": 0.8501,
+      "step": 23091
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5174039167376032e-09,
+      "loss": 1.0009,
+      "step": 23092
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.500307399346923e-09,
+      "loss": 0.8343,
+      "step": 23093
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4833077267539796e-09,
+      "loss": 0.9136,
+      "step": 23094
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4664048992885094e-09,
+      "loss": 0.9211,
+      "step": 23095
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4495989172774727e-09,
+      "loss": 0.9809,
+      "step": 23096
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4328897810467202e-09,
+      "loss": 0.8066,
+      "step": 23097
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4162774909193267e-09,
+      "loss": 0.9999,
+      "step": 23098
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3997620472183671e-09,
+      "loss": 1.0151,
+      "step": 23099
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3833434502624753e-09,
+      "loss": 0.9791,
+      "step": 23100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3670217003708408e-09,
+      "loss": 0.9583,
+      "step": 23101
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3507967978593217e-09,
+      "loss": 0.813,
+      "step": 23102
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3346687430421113e-09,
+      "loss": 0.9788,
+      "step": 23103
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3186375362317373e-09,
+      "loss": 0.9232,
+      "step": 23104
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.302703177739062e-09,
+      "loss": 0.6767,
+      "step": 23105
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2868656678727277e-09,
+      "loss": 0.7455,
+      "step": 23106
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2711250069391557e-09,
+      "loss": 0.9642,
+      "step": 23107
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2554811952436574e-09,
+      "loss": 0.9402,
+      "step": 23108
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2399342330893239e-09,
+      "loss": 0.9122,
+      "step": 23109
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2244841207775804e-09,
+      "loss": 1.0488,
+      "step": 23110
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2091308586070771e-09,
+      "loss": 0.858,
+      "step": 23111
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.193874446876464e-09,
+      "loss": 0.8731,
+      "step": 23112
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1787148858799503e-09,
+      "loss": 1.0944,
+      "step": 23113
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1636521759122999e-09,
+      "loss": 0.8582,
+      "step": 23114
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1486863172649466e-09,
+      "loss": 0.878,
+      "step": 23115
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1338173102276584e-09,
+      "loss": 0.7856,
+      "step": 23116
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1190451550885383e-09,
+      "loss": 0.9758,
+      "step": 23117
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1043698521345792e-09,
+      "loss": 1.0274,
+      "step": 23118
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0897914016488876e-09,
+      "loss": 1.0351,
+      "step": 23119
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0753098039151255e-09,
+      "loss": 1.0751,
+      "step": 23120
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0609250592130693e-09,
+      "loss": 1.0555,
+      "step": 23121
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.046637167821385e-09,
+      "loss": 0.9672,
+      "step": 23122
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.032446130017628e-09,
+      "loss": 0.9376,
+      "step": 23123
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0183519460760238e-09,
+      "loss": 0.8854,
+      "step": 23124
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.004354616269687e-09,
+      "loss": 1.0787,
+      "step": 23125
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.904541408700675e-10,
+      "loss": 0.9807,
+      "step": 23126
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.76650520146949e-10,
+      "loss": 0.9666,
+      "step": 23127
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.629437543667852e-10,
+      "loss": 0.9743,
+      "step": 23128
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.493338437954746e-10,
+      "loss": 1.0748,
+      "step": 23129
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.358207886972503e-10,
+      "loss": 0.7189,
+      "step": 23130
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.224045893335698e-10,
+      "loss": 1.04,
+      "step": 23131
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.090852459636701e-10,
+      "loss": 0.9342,
+      "step": 23132
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.958627588467883e-10,
+      "loss": 0.8843,
+      "step": 23133
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.827371282388309e-10,
+      "loss": 1.0104,
+      "step": 23134
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.697083543934837e-10,
+      "loss": 0.9495,
+      "step": 23135
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.567764375633226e-10,
+      "loss": 0.6757,
+      "step": 23136
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.439413779992578e-10,
+      "loss": 0.8518,
+      "step": 23137
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.312031759499795e-10,
+      "loss": 0.9491,
+      "step": 23138
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.18561831661957e-10,
+      "loss": 0.9311,
+      "step": 23139
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.060173453805498e-10,
+      "loss": 0.9196,
+      "step": 23140
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.935697173483414e-10,
+      "loss": 0.9863,
+      "step": 23141
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.812189478068056e-10,
+      "loss": 1.0347,
+      "step": 23142
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.689650369951951e-10,
+      "loss": 1.0162,
+      "step": 23143
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.568079851510979e-10,
+      "loss": 0.9949,
+      "step": 23144
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.44747792509326e-10,
+      "loss": 1.1538,
+      "step": 23145
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.327844593041366e-10,
+      "loss": 0.9619,
+      "step": 23146
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.209179857675664e-10,
+      "loss": 0.9457,
+      "step": 23147
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.091483721288761e-10,
+      "loss": 0.8914,
+      "step": 23148
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.974756186167719e-10,
+      "loss": 0.9605,
+      "step": 23149
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.858997254566291e-10,
+      "loss": 0.834,
+      "step": 23150
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.744206928732677e-10,
+      "loss": 0.8064,
+      "step": 23151
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.630385210892876e-10,
+      "loss": 0.8354,
+      "step": 23152
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.517532103245128e-10,
+      "loss": 0.9428,
+      "step": 23153
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.405647607982124e-10,
+      "loss": 0.8831,
+      "step": 23154
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.294731727263248e-10,
+      "loss": 0.8846,
+      "step": 23155
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.184784463247884e-10,
+      "loss": 1.0015,
+      "step": 23156
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.075805818062108e-10,
+      "loss": 1.0351,
+      "step": 23157
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.967795793820896e-10,
+      "loss": 0.9495,
+      "step": 23158
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.860754392605916e-10,
+      "loss": 0.9366,
+      "step": 23159
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.754681616504387e-10,
+      "loss": 0.9271,
+      "step": 23160
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.64957746755912e-10,
+      "loss": 0.7955,
+      "step": 23161
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.545441947818476e-10,
+      "loss": 0.915,
+      "step": 23162
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.442275059291957e-10,
+      "loss": 1.0445,
+      "step": 23163
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.340076803977967e-10,
+      "loss": 0.9686,
+      "step": 23164
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.238847183863804e-10,
+      "loss": 0.9294,
+      "step": 23165
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.138586200903462e-10,
+      "loss": 1.1112,
+      "step": 23166
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.039293857045379e-10,
+      "loss": 1.0345,
+      "step": 23167
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.940970154204694e-10,
+      "loss": 1.0341,
+      "step": 23168
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.843615094296539e-10,
+      "loss": 0.8563,
+      "step": 23169
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.747228679202742e-10,
+      "loss": 0.9521,
+      "step": 23170
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.651810910788479e-10,
+      "loss": 0.9285,
+      "step": 23171
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.5573617909078216e-10,
+      "loss": 0.885,
+      "step": 23172
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.4638813213815356e-10,
+      "loss": 0.7725,
+      "step": 23173
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.371369504030387e-10,
+      "loss": 1.0205,
+      "step": 23174
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.279826340641835e-10,
+      "loss": 0.8785,
+      "step": 23175
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.1892518329922363e-10,
+      "loss": 0.9143,
+      "step": 23176
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.0996459828357424e-10,
+      "loss": 1.0414,
+      "step": 23177
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.011008791904303e-10,
+      "loss": 1.03,
+      "step": 23178
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.9233402619187624e-10,
+      "loss": 0.9388,
+      "step": 23179
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.836640394583313e-10,
+      "loss": 0.8806,
+      "step": 23180
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.75090919156329e-10,
+      "loss": 0.9375,
+      "step": 23181
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.66614665453513e-10,
+      "loss": 0.919,
+      "step": 23182
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.5823527851308604e-10,
+      "loss": 0.8621,
+      "step": 23183
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.4995275849769583e-10,
+      "loss": 0.8461,
+      "step": 23184
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.417671055683247e-10,
+      "loss": 0.976,
+      "step": 23185
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.3367831988262435e-10,
+      "loss": 0.7549,
+      "step": 23186
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2568640159824636e-10,
+      "loss": 0.804,
+      "step": 23187
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.1779135086895676e-10,
+      "loss": 1.0474,
+      "step": 23188
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.099931678485213e-10,
+      "loss": 0.8866,
+      "step": 23189
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.022918526884855e-10,
+      "loss": 1.0372,
+      "step": 23190
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.94687405536509e-10,
+      "loss": 0.9544,
+      "step": 23191
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8717982654136166e-10,
+      "loss": 1.0458,
+      "step": 23192
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7976911584792764e-10,
+      "loss": 0.9466,
+      "step": 23193
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.724552735999808e-10,
+      "loss": 0.8938,
+      "step": 23194
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6523829993907456e-10,
+      "loss": 0.8545,
+      "step": 23195
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5811819500454195e-10,
+      "loss": 1.0289,
+      "step": 23196
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5109495893516076e-10,
+      "loss": 0.8564,
+      "step": 23197
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4416859186693343e-10,
+      "loss": 0.8739,
+      "step": 23198
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3733909393364175e-10,
+      "loss": 1.1332,
+      "step": 23199
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3060646526740228e-10,
+      "loss": 1.118,
+      "step": 23200
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2397070599922132e-10,
+      "loss": 0.8608,
+      "step": 23201
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1743181625788478e-10,
+      "loss": 0.9841,
+      "step": 23202
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1098979616940297e-10,
+      "loss": 0.9269,
+      "step": 23203
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0464464585867594e-10,
+      "loss": 1.008,
+      "step": 23204
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9839636544893847e-10,
+      "loss": 1.0131,
+      "step": 23205
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9224495506120487e-10,
+      "loss": 0.9457,
+      "step": 23206
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8619041481426902e-10,
+      "loss": 0.8944,
+      "step": 23207
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8023274482636965e-10,
+      "loss": 0.9435,
+      "step": 23208
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7437194521185974e-10,
+      "loss": 0.9573,
+      "step": 23209
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6860801608453715e-10,
+      "loss": 1.031,
+      "step": 23210
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.629409575565344e-10,
+      "loss": 0.8381,
+      "step": 23211
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5737076973776354e-10,
+      "loss": 0.9793,
+      "step": 23212
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5189745273536115e-10,
+      "loss": 0.7038,
+      "step": 23213
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4652100665590863e-10,
+      "loss": 0.8244,
+      "step": 23214
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.412414316032118e-10,
+      "loss": 0.8701,
+      "step": 23215
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.360587276799663e-10,
+      "loss": 0.8323,
+      "step": 23216
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3097289498664733e-10,
+      "loss": 0.8233,
+      "step": 23217
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2598393362150963e-10,
+      "loss": 1.0447,
+      "step": 23218
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2109184368114258e-10,
+      "loss": 0.7191,
+      "step": 23219
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1629662526047026e-10,
+      "loss": 1.0026,
+      "step": 23220
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1159827845219628e-10,
+      "loss": 0.8829,
+      "step": 23221
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0699680334791407e-10,
+      "loss": 0.8778,
+      "step": 23222
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0249220003588633e-10,
+      "loss": 1.0142,
+      "step": 23223
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.808446860437581e-11,
+      "loss": 0.921,
+      "step": 23224
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.377360913831458e-11,
+      "loss": 0.9571,
+      "step": 23225
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.955962172096932e-11,
+      "loss": 0.8849,
+      "step": 23226
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.544250643449659e-11,
+      "loss": 0.7276,
+      "step": 23227
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.14222633582773e-11,
+      "loss": 0.8442,
+      "step": 23228
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.749889257002708e-11,
+      "loss": 1.0764,
+      "step": 23229
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.36723941463513e-11,
+      "loss": 0.9005,
+      "step": 23230
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.994276816107982e-11,
+      "loss": 0.9455,
+      "step": 23231
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.63100146863771e-11,
+      "loss": 0.8343,
+      "step": 23232
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.277413379274233e-11,
+      "loss": 0.9098,
+      "step": 23233
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.933512554900933e-11,
+      "loss": 0.8561,
+      "step": 23234
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.599299002123637e-11,
+      "loss": 0.8616,
+      "step": 23235
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.27477272743715e-11,
+      "loss": 0.9905,
+      "step": 23236
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.959933737169742e-11,
+      "loss": 0.9561,
+      "step": 23237
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.65478203737213e-11,
+      "loss": 1.0777,
+      "step": 23238
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.359317633928495e-11,
+      "loss": 0.918,
+      "step": 23239
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.073540532667508e-11,
+      "loss": 1.0245,
+      "step": 23240
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.7974507390292625e-11,
+      "loss": 0.9211,
+      "step": 23241
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.53104825839834e-11,
+      "loss": 0.967,
+      "step": 23242
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2743330959372764e-11,
+      "loss": 0.898,
+      "step": 23243
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0273052565865656e-11,
+      "loss": 0.9278,
+      "step": 23244
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7899647452311883e-11,
+      "loss": 0.9317,
+      "step": 23245
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.562311566312037e-11,
+      "loss": 0.9338,
+      "step": 23246
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3443457244365364e-11,
+      "loss": 0.972,
+      "step": 23247
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1360672236570012e-11,
+      "loss": 0.9419,
+      "step": 23248
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9374760680812565e-11,
+      "loss": 0.7992,
+      "step": 23249
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7485722615395717e-11,
+      "loss": 0.9217,
+      "step": 23250
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5693558076956828e-11,
+      "loss": 0.7999,
+      "step": 23251
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3998267100467922e-11,
+      "loss": 0.8743,
+      "step": 23252
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.239984971868058e-11,
+      "loss": 0.9783,
+      "step": 23253
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0898305962125932e-11,
+      "loss": 1.1066,
+      "step": 23254
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.493635860224892e-12,
+      "loss": 0.8783,
+      "step": 23255
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.185839440733034e-12,
+      "loss": 0.9131,
+      "step": 23256
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.974916728075265e-12,
+      "loss": 0.9944,
+      "step": 23257
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.860867746121379e-12,
+      "loss": 1.1631,
+      "step": 23258
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.843692516520726e-12,
+      "loss": 1.0899,
+      "step": 23259
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.923391058702209e-12,
+      "loss": 0.9706,
+      "step": 23260
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0999633904293947e-12,
+      "loss": 0.9163,
+      "step": 23261
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.37340952835563e-12,
+      "loss": 0.8853,
+      "step": 23262
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7437294858035913e-12,
+      "loss": 0.9524,
+      "step": 23263
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.210923275540843e-12,
+      "loss": 1.0115,
+      "step": 23264
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.74990907559392e-13,
+      "loss": 0.8022,
+      "step": 23265
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.359323907410229e-13,
+      "loss": 0.9323,
+      "step": 23266
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9374773063685072e-13,
+      "loss": 0.9564,
+      "step": 23267
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.8436932797990553e-14,
+      "loss": 1.0764,
+      "step": 23268
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 0.8761,
+      "step": 23269
+    },
+    {
+      "epoch": 1.0,
+      "step": 23269,
+      "total_flos": 1.3704558297061786e+19,
+      "train_loss": 0.9649609070239394,
+      "train_runtime": 251998.0019,
+      "train_samples_per_second": 2.955,
+      "train_steps_per_second": 0.092
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 23269,
+  "num_train_epochs": 1,
+  "save_steps": 1000,
+  "total_flos": 1.3704558297061786e+19,
+  "trial_name": null,
+  "trial_params": null
+}