diff --git "a/checkpoint-4800/trainer_state.json" "b/checkpoint-4800/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-4800/trainer_state.json"
@@ -0,0 +1,29056 @@
+{
+  "best_metric": 0.7037800550460815,
+  "best_model_checkpoint": "music_genres_classification/checkpoint-4800",
+  "epoch": 24.0,
+  "global_step": 4800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 2.2932,
+      "step": 1
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 2.2954,
+      "step": 2
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 2.3038,
+      "step": 3
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 2.3061,
+      "step": 4
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 2.3066,
+      "step": 5
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 2.3066,
+      "step": 6
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 2.309,
+      "step": 7
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 2.3315,
+      "step": 8
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 2.3248,
+      "step": 9
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 2.2967,
+      "step": 10
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.2e-06,
+      "loss": 2.3046,
+      "step": 11
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 2.2841,
+      "step": 12
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.6e-06,
+      "loss": 2.3072,
+      "step": 13
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 2.3001,
+      "step": 14
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 3e-06,
+      "loss": 2.2873,
+      "step": 15
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 2.2941,
+      "step": 16
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 2.2811,
+      "step": 17
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 2.3134,
+      "step": 18
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 2.3001,
+      "step": 19
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 2.3116,
+      "step": 20
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 2.3018,
+      "step": 21
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.4e-06,
+      "loss": 2.2695,
+      "step": 22
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 2.2833,
+      "step": 23
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 2.3257,
+      "step": 24
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5e-06,
+      "loss": 2.2971,
+      "step": 25
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.2e-06,
+      "loss": 2.2917,
+      "step": 26
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.400000000000001e-06,
+      "loss": 2.2933,
+      "step": 27
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 2.3002,
+      "step": 28
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.8e-06,
+      "loss": 2.3115,
+      "step": 29
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 6e-06,
+      "loss": 2.2865,
+      "step": 30
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 6.200000000000001e-06,
+      "loss": 2.3126,
+      "step": 31
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 2.2902,
+      "step": 32
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.600000000000001e-06,
+      "loss": 2.2887,
+      "step": 33
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.800000000000001e-06,
+      "loss": 2.3063,
+      "step": 34
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7e-06,
+      "loss": 2.3228,
+      "step": 35
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.2000000000000005e-06,
+      "loss": 2.3111,
+      "step": 36
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.4e-06,
+      "loss": 2.2944,
+      "step": 37
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 7.600000000000001e-06,
+      "loss": 2.2954,
+      "step": 38
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 7.800000000000002e-06,
+      "loss": 2.2872,
+      "step": 39
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 2.3117,
+      "step": 40
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.2e-06,
+      "loss": 2.2934,
+      "step": 41
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 2.3079,
+      "step": 42
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.6e-06,
+      "loss": 2.2983,
+      "step": 43
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.8e-06,
+      "loss": 2.2744,
+      "step": 44
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9e-06,
+      "loss": 2.2667,
+      "step": 45
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.200000000000002e-06,
+      "loss": 2.2932,
+      "step": 46
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.4e-06,
+      "loss": 2.309,
+      "step": 47
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 2.3159,
+      "step": 48
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.800000000000001e-06,
+      "loss": 2.3062,
+      "step": 49
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1e-05,
+      "loss": 2.2922,
+      "step": 50
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.9979797979798e-06,
+      "loss": 2.2964,
+      "step": 51
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.995959595959597e-06,
+      "loss": 2.2741,
+      "step": 52
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.993939393939395e-06,
+      "loss": 2.3057,
+      "step": 53
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.991919191919192e-06,
+      "loss": 2.2976,
+      "step": 54
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.989898989898991e-06,
+      "loss": 2.3251,
+      "step": 55
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.987878787878788e-06,
+      "loss": 2.271,
+      "step": 56
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.985858585858587e-06,
+      "loss": 2.2674,
+      "step": 57
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.983838383838384e-06,
+      "loss": 2.2882,
+      "step": 58
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.981818181818183e-06,
+      "loss": 2.2889,
+      "step": 59
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.97979797979798e-06,
+      "loss": 2.302,
+      "step": 60
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.977777777777778e-06,
+      "loss": 2.2741,
+      "step": 61
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.975757575757577e-06,
+      "loss": 2.2915,
+      "step": 62
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.973737373737374e-06,
+      "loss": 2.2719,
+      "step": 63
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.971717171717173e-06,
+      "loss": 2.3024,
+      "step": 64
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.96969696969697e-06,
+      "loss": 2.2989,
+      "step": 65
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.967676767676769e-06,
+      "loss": 2.2841,
+      "step": 66
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.965656565656566e-06,
+      "loss": 2.2746,
+      "step": 67
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.963636363636364e-06,
+      "loss": 2.2994,
+      "step": 68
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.961616161616162e-06,
+      "loss": 2.297,
+      "step": 69
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.95959595959596e-06,
+      "loss": 2.2823,
+      "step": 70
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.957575757575757e-06,
+      "loss": 2.2773,
+      "step": 71
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.955555555555556e-06,
+      "loss": 2.2754,
+      "step": 72
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.953535353535355e-06,
+      "loss": 2.279,
+      "step": 73
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.951515151515152e-06,
+      "loss": 2.2972,
+      "step": 74
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.94949494949495e-06,
+      "loss": 2.2766,
+      "step": 75
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.947474747474748e-06,
+      "loss": 2.3015,
+      "step": 76
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.945454545454546e-06,
+      "loss": 2.2929,
+      "step": 77
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.943434343434343e-06,
+      "loss": 2.2905,
+      "step": 78
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.941414141414142e-06,
+      "loss": 2.2789,
+      "step": 79
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.939393939393939e-06,
+      "loss": 2.2992,
+      "step": 80
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.937373737373738e-06,
+      "loss": 2.2996,
+      "step": 81
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.935353535353535e-06,
+      "loss": 2.2702,
+      "step": 82
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.933333333333334e-06,
+      "loss": 2.2906,
+      "step": 83
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.931313131313132e-06,
+      "loss": 2.3328,
+      "step": 84
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.92929292929293e-06,
+      "loss": 2.2985,
+      "step": 85
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.927272727272728e-06,
+      "loss": 2.2131,
+      "step": 86
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.925252525252525e-06,
+      "loss": 2.3006,
+      "step": 87
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.923232323232324e-06,
+      "loss": 2.3084,
+      "step": 88
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.921212121212121e-06,
+      "loss": 2.2748,
+      "step": 89
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.91919191919192e-06,
+      "loss": 2.2935,
+      "step": 90
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.917171717171717e-06,
+      "loss": 2.2855,
+      "step": 91
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.915151515151515e-06,
+      "loss": 2.3119,
+      "step": 92
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.913131313131314e-06,
+      "loss": 2.2845,
+      "step": 93
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.911111111111113e-06,
+      "loss": 2.3043,
+      "step": 94
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.90909090909091e-06,
+      "loss": 2.2593,
+      "step": 95
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.907070707070709e-06,
+      "loss": 2.2942,
+      "step": 96
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.905050505050506e-06,
+      "loss": 2.3092,
+      "step": 97
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.903030303030305e-06,
+      "loss": 2.2706,
+      "step": 98
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.901010101010102e-06,
+      "loss": 2.2573,
+      "step": 99
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.8989898989899e-06,
+      "loss": 2.3258,
+      "step": 100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.896969696969699e-06,
+      "loss": 2.1982,
+      "step": 101
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.894949494949496e-06,
+      "loss": 2.2975,
+      "step": 102
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.892929292929295e-06,
+      "loss": 2.2455,
+      "step": 103
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.890909090909092e-06,
+      "loss": 2.2551,
+      "step": 104
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.88888888888889e-06,
+      "loss": 2.2441,
+      "step": 105
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.886868686868688e-06,
+      "loss": 2.2986,
+      "step": 106
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.884848484848486e-06,
+      "loss": 2.2845,
+      "step": 107
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.882828282828283e-06,
+      "loss": 2.322,
+      "step": 108
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.880808080808082e-06,
+      "loss": 2.2616,
+      "step": 109
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.87878787878788e-06,
+      "loss": 2.255,
+      "step": 110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.876767676767678e-06,
+      "loss": 2.3257,
+      "step": 111
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.874747474747477e-06,
+      "loss": 2.2894,
+      "step": 112
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.872727272727274e-06,
+      "loss": 2.2701,
+      "step": 113
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.870707070707072e-06,
+      "loss": 2.2946,
+      "step": 114
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.86868686868687e-06,
+      "loss": 2.2279,
+      "step": 115
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.866666666666668e-06,
+      "loss": 2.3107,
+      "step": 116
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.864646464646465e-06,
+      "loss": 2.2936,
+      "step": 117
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.862626262626264e-06,
+      "loss": 2.2687,
+      "step": 118
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.860606060606061e-06,
+      "loss": 2.2705,
+      "step": 119
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.85858585858586e-06,
+      "loss": 2.2552,
+      "step": 120
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.856565656565657e-06,
+      "loss": 2.3076,
+      "step": 121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.854545454545456e-06,
+      "loss": 2.2395,
+      "step": 122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.852525252525254e-06,
+      "loss": 2.2629,
+      "step": 123
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.850505050505051e-06,
+      "loss": 2.3,
+      "step": 124
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.84848484848485e-06,
+      "loss": 2.2742,
+      "step": 125
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.846464646464647e-06,
+      "loss": 2.2078,
+      "step": 126
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.844444444444446e-06,
+      "loss": 2.2945,
+      "step": 127
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.842424242424243e-06,
+      "loss": 2.2198,
+      "step": 128
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.840404040404042e-06,
+      "loss": 2.3014,
+      "step": 129
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.838383838383839e-06,
+      "loss": 2.2546,
+      "step": 130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.836363636363637e-06,
+      "loss": 2.2572,
+      "step": 131
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.834343434343434e-06,
+      "loss": 2.2433,
+      "step": 132
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.832323232323233e-06,
+      "loss": 2.3109,
+      "step": 133
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.830303030303032e-06,
+      "loss": 2.176,
+      "step": 134
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.828282828282829e-06,
+      "loss": 2.3006,
+      "step": 135
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.826262626262628e-06,
+      "loss": 2.2698,
+      "step": 136
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.824242424242425e-06,
+      "loss": 2.1046,
+      "step": 137
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.822222222222223e-06,
+      "loss": 2.2115,
+      "step": 138
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.82020202020202e-06,
+      "loss": 2.1585,
+      "step": 139
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.81818181818182e-06,
+      "loss": 2.2031,
+      "step": 140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.816161616161616e-06,
+      "loss": 2.2044,
+      "step": 141
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.814141414141415e-06,
+      "loss": 2.2848,
+      "step": 142
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.812121212121212e-06,
+      "loss": 2.2555,
+      "step": 143
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.81010101010101e-06,
+      "loss": 2.1314,
+      "step": 144
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.80808080808081e-06,
+      "loss": 2.2202,
+      "step": 145
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.806060606060607e-06,
+      "loss": 2.3462,
+      "step": 146
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.804040404040405e-06,
+      "loss": 2.2707,
+      "step": 147
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.802020202020202e-06,
+      "loss": 2.2974,
+      "step": 148
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.800000000000001e-06,
+      "loss": 2.2357,
+      "step": 149
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.797979797979798e-06,
+      "loss": 2.2728,
+      "step": 150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.795959595959597e-06,
+      "loss": 2.2798,
+      "step": 151
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.793939393939394e-06,
+      "loss": 2.2807,
+      "step": 152
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.791919191919193e-06,
+      "loss": 2.2654,
+      "step": 153
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.78989898989899e-06,
+      "loss": 2.2074,
+      "step": 154
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.787878787878788e-06,
+      "loss": 2.3228,
+      "step": 155
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.785858585858587e-06,
+      "loss": 2.2816,
+      "step": 156
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.783838383838384e-06,
+      "loss": 2.2866,
+      "step": 157
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.781818181818183e-06,
+      "loss": 2.2448,
+      "step": 158
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.77979797979798e-06,
+      "loss": 2.1804,
+      "step": 159
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.777777777777779e-06,
+      "loss": 2.2455,
+      "step": 160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.775757575757576e-06,
+      "loss": 2.2181,
+      "step": 161
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.773737373737374e-06,
+      "loss": 2.2438,
+      "step": 162
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.771717171717171e-06,
+      "loss": 2.2139,
+      "step": 163
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.76969696969697e-06,
+      "loss": 2.2163,
+      "step": 164
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.767676767676767e-06,
+      "loss": 2.3125,
+      "step": 165
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.765656565656566e-06,
+      "loss": 2.2794,
+      "step": 166
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.763636363636365e-06,
+      "loss": 2.3172,
+      "step": 167
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.761616161616162e-06,
+      "loss": 2.2236,
+      "step": 168
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.75959595959596e-06,
+      "loss": 2.1736,
+      "step": 169
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.757575757575758e-06,
+      "loss": 2.1504,
+      "step": 170
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.755555555555556e-06,
+      "loss": 2.1313,
+      "step": 171
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.753535353535353e-06,
+      "loss": 2.2934,
+      "step": 172
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.751515151515152e-06,
+      "loss": 2.3056,
+      "step": 173
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.749494949494949e-06,
+      "loss": 2.2612,
+      "step": 174
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.747474747474748e-06,
+      "loss": 2.0828,
+      "step": 175
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.745454545454547e-06,
+      "loss": 2.1816,
+      "step": 176
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.743434343434344e-06,
+      "loss": 2.1183,
+      "step": 177
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.741414141414142e-06,
+      "loss": 2.1275,
+      "step": 178
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.739393939393941e-06,
+      "loss": 2.1788,
+      "step": 179
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.737373737373738e-06,
+      "loss": 2.2433,
+      "step": 180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.735353535353537e-06,
+      "loss": 2.1245,
+      "step": 181
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.733333333333334e-06,
+      "loss": 2.2787,
+      "step": 182
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.731313131313133e-06,
+      "loss": 2.1969,
+      "step": 183
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.729292929292931e-06,
+      "loss": 2.0397,
+      "step": 184
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.727272727272728e-06,
+      "loss": 2.2858,
+      "step": 185
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.725252525252527e-06,
+      "loss": 2.146,
+      "step": 186
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.723232323232324e-06,
+      "loss": 2.0199,
+      "step": 187
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.721212121212123e-06,
+      "loss": 2.2831,
+      "step": 188
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.71919191919192e-06,
+      "loss": 2.2683,
+      "step": 189
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.717171717171719e-06,
+      "loss": 2.2371,
+      "step": 190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.715151515151516e-06,
+      "loss": 2.2635,
+      "step": 191
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.713131313131314e-06,
+      "loss": 2.2297,
+      "step": 192
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.711111111111111e-06,
+      "loss": 2.2301,
+      "step": 193
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.70909090909091e-06,
+      "loss": 2.2674,
+      "step": 194
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.707070707070709e-06,
+      "loss": 2.2585,
+      "step": 195
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.705050505050506e-06,
+      "loss": 2.1779,
+      "step": 196
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.703030303030305e-06,
+      "loss": 2.2938,
+      "step": 197
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.701010101010102e-06,
+      "loss": 2.2777,
+      "step": 198
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.6989898989899e-06,
+      "loss": 1.9855,
+      "step": 199
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.696969696969698e-06,
+      "loss": 2.1218,
+      "step": 200
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.39,
+      "eval_loss": 2.152782917022705,
+      "eval_roc_auc": 0.8530103123916681,
+      "eval_runtime": 95.2064,
+      "eval_samples_per_second": 2.101,
+      "eval_steps_per_second": 0.525,
+      "step": 200
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.694949494949496e-06,
+      "loss": 2.1429,
+      "step": 201
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.692929292929293e-06,
+      "loss": 2.2593,
+      "step": 202
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.690909090909092e-06,
+      "loss": 2.2602,
+      "step": 203
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.688888888888889e-06,
+      "loss": 2.0947,
+      "step": 204
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.686868686868688e-06,
+      "loss": 2.2526,
+      "step": 205
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.684848484848487e-06,
+      "loss": 2.0549,
+      "step": 206
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.682828282828284e-06,
+      "loss": 2.2288,
+      "step": 207
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.680808080808082e-06,
+      "loss": 2.2026,
+      "step": 208
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.67878787878788e-06,
+      "loss": 1.8229,
+      "step": 209
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.676767676767678e-06,
+      "loss": 2.2867,
+      "step": 210
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.674747474747475e-06,
+      "loss": 2.2737,
+      "step": 211
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.672727272727274e-06,
+      "loss": 2.2522,
+      "step": 212
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.670707070707071e-06,
+      "loss": 2.1348,
+      "step": 213
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.66868686868687e-06,
+      "loss": 2.2203,
+      "step": 214
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.666666666666667e-06,
+      "loss": 2.2929,
+      "step": 215
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.664646464646465e-06,
+      "loss": 2.2685,
+      "step": 216
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.662626262626264e-06,
+      "loss": 2.0404,
+      "step": 217
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.660606060606061e-06,
+      "loss": 2.1636,
+      "step": 218
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.65858585858586e-06,
+      "loss": 2.173,
+      "step": 219
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.656565656565657e-06,
+      "loss": 2.104,
+      "step": 220
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.654545454545456e-06,
+      "loss": 2.0499,
+      "step": 221
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.652525252525253e-06,
+      "loss": 2.2325,
+      "step": 222
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.650505050505052e-06,
+      "loss": 2.1291,
+      "step": 223
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.648484848484849e-06,
+      "loss": 2.2237,
+      "step": 224
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.646464646464647e-06,
+      "loss": 2.2387,
+      "step": 225
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.644444444444444e-06,
+      "loss": 2.0822,
+      "step": 226
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 9.642424242424243e-06,
+      "loss": 2.0834,
+      "step": 227
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 9.640404040404042e-06,
+      "loss": 2.293,
+      "step": 228
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 9.638383838383839e-06,
+      "loss": 2.2375,
+      "step": 229
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 9.636363636363638e-06,
+      "loss": 2.1509,
+      "step": 230
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 9.634343434343435e-06,
+      "loss": 2.3852,
+      "step": 231
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 9.632323232323233e-06,
+      "loss": 2.1294,
+      "step": 232
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 9.63030303030303e-06,
+      "loss": 2.2615,
+      "step": 233
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 9.628282828282829e-06,
+      "loss": 2.2,
+      "step": 234
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 9.626262626262626e-06,
+      "loss": 2.2859,
+      "step": 235
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 9.624242424242425e-06,
+      "loss": 2.0985,
+      "step": 236
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 9.622222222222222e-06,
+      "loss": 2.251,
+      "step": 237
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 9.62020202020202e-06,
+      "loss": 2.2482,
+      "step": 238
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 9.61818181818182e-06,
+      "loss": 2.2097,
+      "step": 239
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 9.616161616161616e-06,
+      "loss": 2.0572,
+      "step": 240
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.614141414141415e-06,
+      "loss": 2.3019,
+      "step": 241
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.612121212121212e-06,
+      "loss": 2.2236,
+      "step": 242
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.610101010101011e-06,
+      "loss": 2.1363,
+      "step": 243
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.608080808080808e-06,
+      "loss": 2.0712,
+      "step": 244
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.606060606060607e-06,
+      "loss": 2.0515,
+      "step": 245
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.604040404040404e-06,
+      "loss": 2.1827,
+      "step": 246
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.602020202020203e-06,
+      "loss": 2.2268,
+      "step": 247
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 2.0737,
+      "step": 248
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 9.597979797979798e-06,
+      "loss": 2.224,
+      "step": 249
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 9.595959595959597e-06,
+      "loss": 2.1697,
+      "step": 250
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 9.593939393939394e-06,
+      "loss": 2.2166,
+      "step": 251
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 9.591919191919193e-06,
+      "loss": 2.2622,
+      "step": 252
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 9.58989898989899e-06,
+      "loss": 1.9912,
+      "step": 253
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 9.587878787878789e-06,
+      "loss": 2.2339,
+      "step": 254
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 9.585858585858586e-06,
+      "loss": 2.3089,
+      "step": 255
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 9.583838383838384e-06,
+      "loss": 2.0928,
+      "step": 256
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 9.581818181818181e-06,
+      "loss": 2.2706,
+      "step": 257
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 9.57979797979798e-06,
+      "loss": 2.2972,
+      "step": 258
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 9.577777777777779e-06,
+      "loss": 1.9961,
+      "step": 259
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 9.575757575757576e-06,
+      "loss": 2.2101,
+      "step": 260
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 9.573737373737375e-06,
+      "loss": 1.9884,
+      "step": 261
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 9.571717171717172e-06,
+      "loss": 2.1481,
+      "step": 262
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 9.56969696969697e-06,
+      "loss": 2.1575,
+      "step": 263
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 9.56767676767677e-06,
+      "loss": 2.1933,
+      "step": 264
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 9.565656565656566e-06,
+      "loss": 2.2463,
+      "step": 265
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 9.563636363636365e-06,
+      "loss": 2.0857,
+      "step": 266
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 9.561616161616164e-06,
+      "loss": 2.0031,
+      "step": 267
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.55959595959596e-06,
+      "loss": 2.1913,
+      "step": 268
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.55757575757576e-06,
+      "loss": 2.2415,
+      "step": 269
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.555555555555556e-06,
+      "loss": 2.2165,
+      "step": 270
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.553535353535355e-06,
+      "loss": 2.2508,
+      "step": 271
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.551515151515152e-06,
+      "loss": 2.0715,
+      "step": 272
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.549494949494951e-06,
+      "loss": 2.2275,
+      "step": 273
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.547474747474748e-06,
+      "loss": 2.1235,
+      "step": 274
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.545454545454547e-06,
+      "loss": 2.2547,
+      "step": 275
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.543434343434344e-06,
+      "loss": 2.0889,
+      "step": 276
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.541414141414143e-06,
+      "loss": 2.0548,
+      "step": 277
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.539393939393941e-06,
+      "loss": 2.2075,
+      "step": 278
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.537373737373738e-06,
+      "loss": 2.1354,
+      "step": 279
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.535353535353537e-06,
+      "loss": 2.2058,
+      "step": 280
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 9.533333333333334e-06,
+      "loss": 1.9105,
+      "step": 281
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 9.531313131313133e-06,
+      "loss": 2.2032,
+      "step": 282
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 9.52929292929293e-06,
+      "loss": 1.9071,
+      "step": 283
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 9.527272727272729e-06,
+      "loss": 2.2162,
+      "step": 284
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 9.525252525252526e-06,
+      "loss": 2.2957,
+      "step": 285
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 9.523232323232324e-06,
+      "loss": 2.1415,
+      "step": 286
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 9.521212121212121e-06,
+      "loss": 1.9986,
+      "step": 287
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 9.51919191919192e-06,
+      "loss": 2.0168,
+      "step": 288
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 9.517171717171719e-06,
+      "loss": 2.0034,
+      "step": 289
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 9.515151515151516e-06,
+      "loss": 2.245,
+      "step": 290
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 9.513131313131315e-06,
+      "loss": 2.1407,
+      "step": 291
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 9.511111111111112e-06,
+      "loss": 2.0898,
+      "step": 292
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.50909090909091e-06,
+      "loss": 1.7919,
+      "step": 293
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.507070707070707e-06,
+      "loss": 2.2812,
+      "step": 294
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.505050505050506e-06,
+      "loss": 1.9966,
+      "step": 295
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.503030303030303e-06,
+      "loss": 2.2843,
+      "step": 296
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.501010101010102e-06,
+      "loss": 2.0761,
+      "step": 297
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 9.498989898989899e-06,
+      "loss": 2.1573,
+      "step": 298
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.496969696969698e-06,
+      "loss": 2.0133,
+      "step": 299
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.494949494949497e-06,
+      "loss": 2.2282,
+      "step": 300
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.492929292929294e-06,
+      "loss": 2.0954,
+      "step": 301
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.490909090909092e-06,
+      "loss": 2.1722,
+      "step": 302
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.48888888888889e-06,
+      "loss": 1.9792,
+      "step": 303
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.486868686868688e-06,
+      "loss": 2.0275,
+      "step": 304
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.484848484848485e-06,
+      "loss": 2.1401,
+      "step": 305
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.482828282828284e-06,
+      "loss": 2.173,
+      "step": 306
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.480808080808081e-06,
+      "loss": 2.2548,
+      "step": 307
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.47878787878788e-06,
+      "loss": 2.1648,
+      "step": 308
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.476767676767677e-06,
+      "loss": 2.017,
+      "step": 309
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.474747474747475e-06,
+      "loss": 2.0506,
+      "step": 310
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.472727272727274e-06,
+      "loss": 2.22,
+      "step": 311
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.470707070707071e-06,
+      "loss": 2.0549,
+      "step": 312
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.46868686868687e-06,
+      "loss": 2.1448,
+      "step": 313
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.466666666666667e-06,
+      "loss": 1.9624,
+      "step": 314
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.464646464646466e-06,
+      "loss": 2.011,
+      "step": 315
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.462626262626263e-06,
+      "loss": 2.0256,
+      "step": 316
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.460606060606061e-06,
+      "loss": 2.1084,
+      "step": 317
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.458585858585858e-06,
+      "loss": 1.8642,
+      "step": 318
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.456565656565657e-06,
+      "loss": 2.042,
+      "step": 319
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.454545454545456e-06,
+      "loss": 2.1082,
+      "step": 320
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.452525252525253e-06,
+      "loss": 2.168,
+      "step": 321
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.450505050505052e-06,
+      "loss": 2.1859,
+      "step": 322
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.448484848484849e-06,
+      "loss": 2.2166,
+      "step": 323
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.446464646464648e-06,
+      "loss": 2.1043,
+      "step": 324
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.444444444444445e-06,
+      "loss": 2.1258,
+      "step": 325
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.442424242424243e-06,
+      "loss": 2.2085,
+      "step": 326
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.44040404040404e-06,
+      "loss": 2.2362,
+      "step": 327
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.438383838383839e-06,
+      "loss": 2.1057,
+      "step": 328
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.436363636363636e-06,
+      "loss": 2.0491,
+      "step": 329
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.434343434343435e-06,
+      "loss": 2.1843,
+      "step": 330
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 9.432323232323234e-06,
+      "loss": 2.1145,
+      "step": 331
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 9.43030303030303e-06,
+      "loss": 1.9742,
+      "step": 332
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 9.42828282828283e-06,
+      "loss": 2.09,
+      "step": 333
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 9.426262626262626e-06,
+      "loss": 2.1787,
+      "step": 334
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 9.424242424242425e-06,
+      "loss": 1.8769,
+      "step": 335
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 9.422222222222222e-06,
+      "loss": 2.2219,
+      "step": 336
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 9.420202020202021e-06,
+      "loss": 1.9286,
+      "step": 337
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 9.418181818181818e-06,
+      "loss": 2.1465,
+      "step": 338
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 9.416161616161617e-06,
+      "loss": 2.1204,
+      "step": 339
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 9.414141414141414e-06,
+      "loss": 1.7688,
+      "step": 340
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 9.412121212121212e-06,
+      "loss": 1.9369,
+      "step": 341
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 9.410101010101011e-06,
+      "loss": 2.0248,
+      "step": 342
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 9.408080808080808e-06,
+      "loss": 1.9912,
+      "step": 343
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.406060606060607e-06,
+      "loss": 1.8951,
+      "step": 344
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.404040404040404e-06,
+      "loss": 2.1753,
+      "step": 345
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.402020202020203e-06,
+      "loss": 2.1642,
+      "step": 346
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.4e-06,
+      "loss": 2.0389,
+      "step": 347
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.397979797979799e-06,
+      "loss": 1.9538,
+      "step": 348
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 9.395959595959597e-06,
+      "loss": 2.0918,
+      "step": 349
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 9.393939393939396e-06,
+      "loss": 2.0183,
+      "step": 350
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 9.391919191919193e-06,
+      "loss": 2.2232,
+      "step": 351
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 9.389898989898992e-06,
+      "loss": 2.081,
+      "step": 352
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 9.387878787878789e-06,
+      "loss": 2.0184,
+      "step": 353
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 9.385858585858588e-06,
+      "loss": 2.1095,
+      "step": 354
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 9.383838383838385e-06,
+      "loss": 2.0399,
+      "step": 355
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 9.381818181818183e-06,
+      "loss": 2.1704,
+      "step": 356
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 9.37979797979798e-06,
+      "loss": 2.0027,
+      "step": 357
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 9.377777777777779e-06,
+      "loss": 1.9848,
+      "step": 358
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 9.375757575757576e-06,
+      "loss": 1.9639,
+      "step": 359
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.373737373737375e-06,
+      "loss": 1.6844,
+      "step": 360
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.371717171717174e-06,
+      "loss": 2.0172,
+      "step": 361
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.36969696969697e-06,
+      "loss": 2.1155,
+      "step": 362
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.36767676767677e-06,
+      "loss": 2.0813,
+      "step": 363
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.365656565656566e-06,
+      "loss": 2.1317,
+      "step": 364
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.363636363636365e-06,
+      "loss": 2.2794,
+      "step": 365
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 9.361616161616162e-06,
+      "loss": 2.0128,
+      "step": 366
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 9.359595959595961e-06,
+      "loss": 1.8473,
+      "step": 367
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 9.357575757575758e-06,
+      "loss": 2.3378,
+      "step": 368
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 9.355555555555557e-06,
+      "loss": 2.3487,
+      "step": 369
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 9.353535353535354e-06,
+      "loss": 2.0717,
+      "step": 370
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 9.351515151515152e-06,
+      "loss": 2.099,
+      "step": 371
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 9.349494949494951e-06,
+      "loss": 1.8947,
+      "step": 372
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 9.347474747474748e-06,
+      "loss": 2.0128,
+      "step": 373
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 9.345454545454547e-06,
+      "loss": 2.2158,
+      "step": 374
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 9.343434343434344e-06,
+      "loss": 2.0152,
+      "step": 375
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 9.341414141414143e-06,
+      "loss": 2.2708,
+      "step": 376
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 9.33939393939394e-06,
+      "loss": 2.0469,
+      "step": 377
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 9.337373737373739e-06,
+      "loss": 2.1349,
+      "step": 378
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.335353535353536e-06,
+      "loss": 1.7899,
+      "step": 379
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.333333333333334e-06,
+      "loss": 2.169,
+      "step": 380
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.331313131313133e-06,
+      "loss": 1.9634,
+      "step": 381
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.32929292929293e-06,
+      "loss": 2.2009,
+      "step": 382
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.327272727272729e-06,
+      "loss": 1.9424,
+      "step": 383
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.325252525252526e-06,
+      "loss": 2.0749,
+      "step": 384
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 9.323232323232325e-06,
+      "loss": 2.0405,
+      "step": 385
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 9.321212121212122e-06,
+      "loss": 2.0156,
+      "step": 386
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.31919191919192e-06,
+      "loss": 2.1461,
+      "step": 387
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.317171717171717e-06,
+      "loss": 2.1603,
+      "step": 388
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.315151515151516e-06,
+      "loss": 2.1265,
+      "step": 389
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 9.313131313131313e-06,
+      "loss": 2.2353,
+      "step": 390
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 9.311111111111112e-06,
+      "loss": 2.0105,
+      "step": 391
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 9.30909090909091e-06,
+      "loss": 1.7324,
+      "step": 392
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 9.307070707070708e-06,
+      "loss": 2.2114,
+      "step": 393
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.305050505050506e-06,
+      "loss": 2.2254,
+      "step": 394
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 9.303030303030303e-06,
+      "loss": 1.9696,
+      "step": 395
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 9.301010101010102e-06,
+      "loss": 2.039,
+      "step": 396
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 9.2989898989899e-06,
+      "loss": 2.0867,
+      "step": 397
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 9.296969696969698e-06,
+      "loss": 2.2206,
+      "step": 398
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 9.294949494949495e-06,
+      "loss": 2.1514,
+      "step": 399
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 9.292929292929294e-06,
+      "loss": 2.0262,
+      "step": 400
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.4,
+      "eval_loss": 1.9272809028625488,
+      "eval_roc_auc": 0.8615201204558526,
+      "eval_runtime": 94.6661,
+      "eval_samples_per_second": 2.113,
+      "eval_steps_per_second": 0.528,
+      "step": 400
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 9.29090909090909e-06,
+      "loss": 2.0294,
+      "step": 401
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 9.28888888888889e-06,
+      "loss": 2.1193,
+      "step": 402
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.286868686868688e-06,
+      "loss": 2.2414,
+      "step": 403
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.284848484848485e-06,
+      "loss": 1.8922,
+      "step": 404
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 9.282828282828284e-06,
+      "loss": 1.7514,
+      "step": 405
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.280808080808081e-06,
+      "loss": 2.206,
+      "step": 406
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.27878787878788e-06,
+      "loss": 2.1376,
+      "step": 407
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.276767676767677e-06,
+      "loss": 2.1602,
+      "step": 408
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.274747474747476e-06,
+      "loss": 2.1349,
+      "step": 409
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 9.272727272727273e-06,
+      "loss": 2.0569,
+      "step": 410
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.270707070707071e-06,
+      "loss": 1.9973,
+      "step": 411
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.268686868686868e-06,
+      "loss": 2.1153,
+      "step": 412
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.266666666666667e-06,
+      "loss": 2.0438,
+      "step": 413
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 9.264646464646466e-06,
+      "loss": 2.055,
+      "step": 414
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.262626262626263e-06,
+      "loss": 2.0734,
+      "step": 415
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.260606060606062e-06,
+      "loss": 2.0387,
+      "step": 416
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.258585858585859e-06,
+      "loss": 1.7832,
+      "step": 417
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 9.256565656565657e-06,
+      "loss": 1.8364,
+      "step": 418
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.254545454545454e-06,
+      "loss": 1.9508,
+      "step": 419
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.252525252525253e-06,
+      "loss": 2.1001,
+      "step": 420
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.25050505050505e-06,
+      "loss": 2.2306,
+      "step": 421
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.248484848484849e-06,
+      "loss": 1.9892,
+      "step": 422
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.246464646464646e-06,
+      "loss": 2.1173,
+      "step": 423
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.244444444444445e-06,
+      "loss": 2.0736,
+      "step": 424
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.242424242424244e-06,
+      "loss": 2.0678,
+      "step": 425
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.24040404040404e-06,
+      "loss": 1.9815,
+      "step": 426
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.23838383838384e-06,
+      "loss": 1.6816,
+      "step": 427
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.236363636363636e-06,
+      "loss": 1.885,
+      "step": 428
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.234343434343435e-06,
+      "loss": 2.045,
+      "step": 429
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.232323232323232e-06,
+      "loss": 1.7785,
+      "step": 430
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.23030303030303e-06,
+      "loss": 1.9455,
+      "step": 431
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 9.228282828282828e-06,
+      "loss": 1.9833,
+      "step": 432
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.226262626262627e-06,
+      "loss": 1.9038,
+      "step": 433
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.224242424242424e-06,
+      "loss": 1.8939,
+      "step": 434
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.222222222222224e-06,
+      "loss": 2.0951,
+      "step": 435
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9.220202020202021e-06,
+      "loss": 1.9811,
+      "step": 436
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.21818181818182e-06,
+      "loss": 2.0743,
+      "step": 437
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.216161616161617e-06,
+      "loss": 1.9986,
+      "step": 438
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.214141414141416e-06,
+      "loss": 2.0857,
+      "step": 439
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 9.212121212121213e-06,
+      "loss": 1.7992,
+      "step": 440
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 9.210101010101011e-06,
+      "loss": 1.8867,
+      "step": 441
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 9.208080808080808e-06,
+      "loss": 1.9496,
+      "step": 442
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 9.206060606060607e-06,
+      "loss": 2.2744,
+      "step": 443
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 9.204040404040406e-06,
+      "loss": 2.016,
+      "step": 444
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 9.202020202020203e-06,
+      "loss": 2.2733,
+      "step": 445
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 9.200000000000002e-06,
+      "loss": 2.0235,
+      "step": 446
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 9.197979797979799e-06,
+      "loss": 1.6186,
+      "step": 447
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 9.195959595959597e-06,
+      "loss": 1.943,
+      "step": 448
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 9.193939393939395e-06,
+      "loss": 2.0729,
+      "step": 449
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 9.191919191919193e-06,
+      "loss": 1.8487,
+      "step": 450
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 9.18989898989899e-06,
+      "loss": 1.7939,
+      "step": 451
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 9.187878787878789e-06,
+      "loss": 2.0209,
+      "step": 452
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 9.185858585858588e-06,
+      "loss": 2.0803,
+      "step": 453
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 9.183838383838385e-06,
+      "loss": 1.966,
+      "step": 454
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 9.181818181818184e-06,
+      "loss": 1.9908,
+      "step": 455
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 9.17979797979798e-06,
+      "loss": 1.86,
+      "step": 456
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 9.17777777777778e-06,
+      "loss": 2.3029,
+      "step": 457
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 9.175757575757576e-06,
+      "loss": 2.2278,
+      "step": 458
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 9.173737373737375e-06,
+      "loss": 1.8542,
+      "step": 459
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 9.171717171717172e-06,
+      "loss": 2.0356,
+      "step": 460
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 9.169696969696971e-06,
+      "loss": 1.8754,
+      "step": 461
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 9.167676767676768e-06,
+      "loss": 1.9455,
+      "step": 462
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 9.165656565656567e-06,
+      "loss": 1.9996,
+      "step": 463
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 9.163636363636365e-06,
+      "loss": 1.8018,
+      "step": 464
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 9.161616161616162e-06,
+      "loss": 2.1249,
+      "step": 465
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 9.159595959595961e-06,
+      "loss": 2.1546,
+      "step": 466
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 9.157575757575758e-06,
+      "loss": 1.9317,
+      "step": 467
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 9.155555555555557e-06,
+      "loss": 1.6425,
+      "step": 468
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 9.153535353535354e-06,
+      "loss": 2.1707,
+      "step": 469
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 9.151515151515153e-06,
+      "loss": 2.0209,
+      "step": 470
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 9.14949494949495e-06,
+      "loss": 1.565,
+      "step": 471
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 9.147474747474748e-06,
+      "loss": 2.1292,
+      "step": 472
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 9.145454545454546e-06,
+      "loss": 1.9247,
+      "step": 473
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 9.143434343434344e-06,
+      "loss": 1.5309,
+      "step": 474
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 9.141414141414143e-06,
+      "loss": 2.1277,
+      "step": 475
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 9.13939393939394e-06,
+      "loss": 2.0475,
+      "step": 476
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 9.137373737373739e-06,
+      "loss": 1.9636,
+      "step": 477
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 9.135353535353536e-06,
+      "loss": 1.5715,
+      "step": 478
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 9.133333333333335e-06,
+      "loss": 1.8619,
+      "step": 479
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 9.131313131313132e-06,
+      "loss": 2.0989,
+      "step": 480
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 9.12929292929293e-06,
+      "loss": 2.1222,
+      "step": 481
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 9.127272727272727e-06,
+      "loss": 1.9005,
+      "step": 482
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 9.125252525252526e-06,
+      "loss": 1.901,
+      "step": 483
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 9.123232323232323e-06,
+      "loss": 2.0177,
+      "step": 484
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 9.121212121212122e-06,
+      "loss": 1.888,
+      "step": 485
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 9.11919191919192e-06,
+      "loss": 1.7763,
+      "step": 486
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 9.117171717171718e-06,
+      "loss": 2.0391,
+      "step": 487
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 9.115151515151516e-06,
+      "loss": 1.7916,
+      "step": 488
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 9.113131313131313e-06,
+      "loss": 2.2296,
+      "step": 489
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 9.111111111111112e-06,
+      "loss": 1.8244,
+      "step": 490
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 9.10909090909091e-06,
+      "loss": 1.8734,
+      "step": 491
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 9.107070707070708e-06,
+      "loss": 2.1904,
+      "step": 492
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 9.105050505050505e-06,
+      "loss": 1.9907,
+      "step": 493
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 9.103030303030304e-06,
+      "loss": 2.0297,
+      "step": 494
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 9.1010101010101e-06,
+      "loss": 2.1472,
+      "step": 495
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 9.0989898989899e-06,
+      "loss": 1.791,
+      "step": 496
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 9.096969696969698e-06,
+      "loss": 2.0319,
+      "step": 497
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 9.094949494949495e-06,
+      "loss": 2.1021,
+      "step": 498
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 9.092929292929294e-06,
+      "loss": 1.8365,
+      "step": 499
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 9.090909090909091e-06,
+      "loss": 2.03,
+      "step": 500
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 9.08888888888889e-06,
+      "loss": 1.4563,
+      "step": 501
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 9.086868686868687e-06,
+      "loss": 2.0023,
+      "step": 502
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 9.084848484848486e-06,
+      "loss": 1.9615,
+      "step": 503
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 9.082828282828283e-06,
+      "loss": 1.9015,
+      "step": 504
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 9.080808080808081e-06,
+      "loss": 2.1222,
+      "step": 505
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 9.078787878787878e-06,
+      "loss": 2.042,
+      "step": 506
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 9.076767676767677e-06,
+      "loss": 1.7651,
+      "step": 507
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 9.074747474747476e-06,
+      "loss": 1.9154,
+      "step": 508
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 9.072727272727273e-06,
+      "loss": 2.1894,
+      "step": 509
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 9.070707070707072e-06,
+      "loss": 1.6939,
+      "step": 510
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 9.068686868686869e-06,
+      "loss": 2.1474,
+      "step": 511
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 9.066666666666667e-06,
+      "loss": 1.9356,
+      "step": 512
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 9.064646464646464e-06,
+      "loss": 1.7902,
+      "step": 513
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.062626262626263e-06,
+      "loss": 1.9876,
+      "step": 514
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.06060606060606e-06,
+      "loss": 2.1486,
+      "step": 515
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.058585858585859e-06,
+      "loss": 2.1834,
+      "step": 516
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.056565656565656e-06,
+      "loss": 1.9198,
+      "step": 517
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.054545454545455e-06,
+      "loss": 2.1156,
+      "step": 518
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.052525252525253e-06,
+      "loss": 1.8419,
+      "step": 519
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 9.050505050505052e-06,
+      "loss": 2.0569,
+      "step": 520
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 9.04848484848485e-06,
+      "loss": 2.1956,
+      "step": 521
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 9.046464646464648e-06,
+      "loss": 1.7464,
+      "step": 522
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 9.044444444444445e-06,
+      "loss": 1.8283,
+      "step": 523
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 9.042424242424244e-06,
+      "loss": 1.7445,
+      "step": 524
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 9.040404040404042e-06,
+      "loss": 1.9674,
+      "step": 525
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 9.03838383838384e-06,
+      "loss": 1.6975,
+      "step": 526
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 9.036363636363638e-06,
+      "loss": 1.8238,
+      "step": 527
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 9.034343434343435e-06,
+      "loss": 2.0458,
+      "step": 528
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 9.032323232323234e-06,
+      "loss": 1.8965,
+      "step": 529
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 9.030303030303031e-06,
+      "loss": 1.9193,
+      "step": 530
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 9.02828282828283e-06,
+      "loss": 2.0905,
+      "step": 531
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 9.026262626262627e-06,
+      "loss": 2.1747,
+      "step": 532
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 9.024242424242426e-06,
+      "loss": 2.0444,
+      "step": 533
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 9.022222222222223e-06,
+      "loss": 1.7067,
+      "step": 534
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 9.020202020202021e-06,
+      "loss": 2.0689,
+      "step": 535
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 9.01818181818182e-06,
+      "loss": 1.6681,
+      "step": 536
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 9.016161616161617e-06,
+      "loss": 1.9335,
+      "step": 537
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 9.014141414141416e-06,
+      "loss": 1.7857,
+      "step": 538
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 9.012121212121213e-06,
+      "loss": 2.2934,
+      "step": 539
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 9.010101010101012e-06,
+      "loss": 2.0515,
+      "step": 540
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 9.008080808080809e-06,
+      "loss": 1.6908,
+      "step": 541
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 9.006060606060607e-06,
+      "loss": 1.8589,
+      "step": 542
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 9.004040404040404e-06,
+      "loss": 1.5844,
+      "step": 543
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 9.002020202020203e-06,
+      "loss": 1.6067,
+      "step": 544
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 9e-06,
+      "loss": 1.6464,
+      "step": 545
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 8.997979797979799e-06,
+      "loss": 1.9629,
+      "step": 546
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 8.995959595959598e-06,
+      "loss": 1.774,
+      "step": 547
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 8.993939393939395e-06,
+      "loss": 1.8639,
+      "step": 548
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 8.991919191919193e-06,
+      "loss": 1.7871,
+      "step": 549
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 8.98989898989899e-06,
+      "loss": 2.0846,
+      "step": 550
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 8.98787878787879e-06,
+      "loss": 1.8979,
+      "step": 551
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 8.985858585858586e-06,
+      "loss": 2.0684,
+      "step": 552
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 8.983838383838385e-06,
+      "loss": 2.1108,
+      "step": 553
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 8.981818181818182e-06,
+      "loss": 1.9738,
+      "step": 554
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 8.97979797979798e-06,
+      "loss": 1.8163,
+      "step": 555
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 8.977777777777778e-06,
+      "loss": 1.612,
+      "step": 556
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 8.975757575757577e-06,
+      "loss": 2.2979,
+      "step": 557
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 8.973737373737375e-06,
+      "loss": 1.8524,
+      "step": 558
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 8.971717171717172e-06,
+      "loss": 2.1915,
+      "step": 559
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 8.969696969696971e-06,
+      "loss": 1.9974,
+      "step": 560
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.967676767676768e-06,
+      "loss": 1.8363,
+      "step": 561
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.965656565656567e-06,
+      "loss": 1.6714,
+      "step": 562
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 8.963636363636364e-06,
+      "loss": 1.7524,
+      "step": 563
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 8.961616161616163e-06,
+      "loss": 1.7228,
+      "step": 564
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.95959595959596e-06,
+      "loss": 2.0731,
+      "step": 565
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.957575757575758e-06,
+      "loss": 2.0048,
+      "step": 566
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 8.955555555555555e-06,
+      "loss": 2.1137,
+      "step": 567
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 8.953535353535354e-06,
+      "loss": 2.3907,
+      "step": 568
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 8.951515151515153e-06,
+      "loss": 2.2295,
+      "step": 569
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 8.94949494949495e-06,
+      "loss": 2.0533,
+      "step": 570
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 8.947474747474749e-06,
+      "loss": 1.8128,
+      "step": 571
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 8.945454545454546e-06,
+      "loss": 1.7129,
+      "step": 572
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.943434343434344e-06,
+      "loss": 1.9075,
+      "step": 573
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.941414141414142e-06,
+      "loss": 1.6914,
+      "step": 574
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.93939393939394e-06,
+      "loss": 1.7457,
+      "step": 575
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.937373737373737e-06,
+      "loss": 1.5131,
+      "step": 576
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.935353535353536e-06,
+      "loss": 2.1524,
+      "step": 577
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 8.933333333333333e-06,
+      "loss": 1.536,
+      "step": 578
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 8.931313131313132e-06,
+      "loss": 2.0683,
+      "step": 579
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 8.92929292929293e-06,
+      "loss": 2.0121,
+      "step": 580
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 8.927272727272728e-06,
+      "loss": 2.0444,
+      "step": 581
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 8.925252525252526e-06,
+      "loss": 1.6644,
+      "step": 582
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 8.923232323232323e-06,
+      "loss": 1.6146,
+      "step": 583
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 8.921212121212122e-06,
+      "loss": 2.0772,
+      "step": 584
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 8.919191919191919e-06,
+      "loss": 1.5158,
+      "step": 585
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 8.917171717171718e-06,
+      "loss": 2.2125,
+      "step": 586
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 8.915151515151515e-06,
+      "loss": 2.0298,
+      "step": 587
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 8.913131313131314e-06,
+      "loss": 2.0404,
+      "step": 588
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 8.91111111111111e-06,
+      "loss": 1.9077,
+      "step": 589
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 8.90909090909091e-06,
+      "loss": 1.8521,
+      "step": 590
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.907070707070708e-06,
+      "loss": 1.8307,
+      "step": 591
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.905050505050505e-06,
+      "loss": 1.817,
+      "step": 592
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.903030303030304e-06,
+      "loss": 1.7808,
+      "step": 593
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 8.901010101010101e-06,
+      "loss": 1.9523,
+      "step": 594
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.8989898989899e-06,
+      "loss": 2.0321,
+      "step": 595
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.896969696969697e-06,
+      "loss": 1.8821,
+      "step": 596
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 8.894949494949495e-06,
+      "loss": 1.7514,
+      "step": 597
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.892929292929293e-06,
+      "loss": 1.9981,
+      "step": 598
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 8.890909090909091e-06,
+      "loss": 1.89,
+      "step": 599
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 8.888888888888888e-06,
+      "loss": 1.7837,
+      "step": 600
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.465,
+      "eval_loss": 1.749820351600647,
+      "eval_roc_auc": 0.893670439955805,
+      "eval_runtime": 95.5238,
+      "eval_samples_per_second": 2.094,
+      "eval_steps_per_second": 0.523,
+      "step": 600
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 8.886868686868687e-06,
+      "loss": 1.6575,
+      "step": 601
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 8.884848484848486e-06,
+      "loss": 1.6596,
+      "step": 602
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 8.882828282828283e-06,
+      "loss": 1.6216,
+      "step": 603
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 8.880808080808082e-06,
+      "loss": 1.5891,
+      "step": 604
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 8.87878787878788e-06,
+      "loss": 1.7113,
+      "step": 605
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 8.876767676767677e-06,
+      "loss": 1.3989,
+      "step": 606
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 8.874747474747476e-06,
+      "loss": 1.7831,
+      "step": 607
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 8.872727272727275e-06,
+      "loss": 1.8975,
+      "step": 608
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 8.870707070707072e-06,
+      "loss": 1.876,
+      "step": 609
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 8.86868686868687e-06,
+      "loss": 1.8439,
+      "step": 610
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 8.866666666666668e-06,
+      "loss": 1.8982,
+      "step": 611
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 8.864646464646466e-06,
+      "loss": 1.5357,
+      "step": 612
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 8.862626262626263e-06,
+      "loss": 1.9224,
+      "step": 613
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 8.860606060606062e-06,
+      "loss": 1.037,
+      "step": 614
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 8.85858585858586e-06,
+      "loss": 1.7186,
+      "step": 615
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 8.856565656565658e-06,
+      "loss": 1.8692,
+      "step": 616
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 8.854545454545455e-06,
+      "loss": 2.1977,
+      "step": 617
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 8.852525252525254e-06,
+      "loss": 1.8494,
+      "step": 618
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 8.850505050505052e-06,
+      "loss": 2.1215,
+      "step": 619
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 8.84848484848485e-06,
+      "loss": 2.0956,
+      "step": 620
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 8.846464646464648e-06,
+      "loss": 2.0335,
+      "step": 621
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 8.844444444444445e-06,
+      "loss": 2.0326,
+      "step": 622
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 8.842424242424244e-06,
+      "loss": 2.118,
+      "step": 623
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 8.840404040404041e-06,
+      "loss": 1.4407,
+      "step": 624
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 8.83838383838384e-06,
+      "loss": 1.5994,
+      "step": 625
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 8.836363636363637e-06,
+      "loss": 1.5573,
+      "step": 626
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 8.834343434343436e-06,
+      "loss": 1.6486,
+      "step": 627
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 8.832323232323233e-06,
+      "loss": 1.803,
+      "step": 628
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 8.830303030303031e-06,
+      "loss": 1.9556,
+      "step": 629
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 8.82828282828283e-06,
+      "loss": 1.9729,
+      "step": 630
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 8.826262626262627e-06,
+      "loss": 2.1698,
+      "step": 631
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 8.824242424242426e-06,
+      "loss": 1.7565,
+      "step": 632
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 8.822222222222223e-06,
+      "loss": 1.915,
+      "step": 633
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 8.820202020202022e-06,
+      "loss": 1.917,
+      "step": 634
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 8.818181818181819e-06,
+      "loss": 2.052,
+      "step": 635
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 8.816161616161617e-06,
+      "loss": 1.8508,
+      "step": 636
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 8.814141414141414e-06,
+      "loss": 2.1709,
+      "step": 637
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 8.812121212121213e-06,
+      "loss": 1.8739,
+      "step": 638
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 8.81010101010101e-06,
+      "loss": 2.009,
+      "step": 639
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 8.808080808080809e-06,
+      "loss": 2.1573,
+      "step": 640
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 8.806060606060608e-06,
+      "loss": 1.8666,
+      "step": 641
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 8.804040404040405e-06,
+      "loss": 1.7116,
+      "step": 642
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 8.802020202020203e-06,
+      "loss": 1.8869,
+      "step": 643
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 8.8e-06,
+      "loss": 1.5702,
+      "step": 644
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 8.7979797979798e-06,
+      "loss": 2.0569,
+      "step": 645
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 8.795959595959596e-06,
+      "loss": 1.5342,
+      "step": 646
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 8.793939393939395e-06,
+      "loss": 1.8325,
+      "step": 647
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 8.791919191919192e-06,
+      "loss": 1.9863,
+      "step": 648
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 8.78989898989899e-06,
+      "loss": 1.8299,
+      "step": 649
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 8.787878787878788e-06,
+      "loss": 2.2175,
+      "step": 650
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 8.785858585858587e-06,
+      "loss": 1.9919,
+      "step": 651
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 8.783838383838385e-06,
+      "loss": 1.5427,
+      "step": 652
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 8.781818181818182e-06,
+      "loss": 2.1991,
+      "step": 653
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 8.779797979797981e-06,
+      "loss": 1.4521,
+      "step": 654
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 8.777777777777778e-06,
+      "loss": 2.0491,
+      "step": 655
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 8.775757575757577e-06,
+      "loss": 1.6083,
+      "step": 656
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 8.773737373737374e-06,
+      "loss": 1.628,
+      "step": 657
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 8.771717171717173e-06,
+      "loss": 2.022,
+      "step": 658
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 8.76969696969697e-06,
+      "loss": 1.6902,
+      "step": 659
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 8.767676767676768e-06,
+      "loss": 1.5954,
+      "step": 660
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 8.765656565656565e-06,
+      "loss": 1.8429,
+      "step": 661
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 8.763636363636364e-06,
+      "loss": 2.1641,
+      "step": 662
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 8.761616161616163e-06,
+      "loss": 1.8101,
+      "step": 663
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 8.75959595959596e-06,
+      "loss": 1.7301,
+      "step": 664
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 8.757575757575759e-06,
+      "loss": 1.8491,
+      "step": 665
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 8.755555555555556e-06,
+      "loss": 2.0537,
+      "step": 666
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 8.753535353535354e-06,
+      "loss": 2.024,
+      "step": 667
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 8.751515151515151e-06,
+      "loss": 2.2037,
+      "step": 668
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 8.74949494949495e-06,
+      "loss": 1.9218,
+      "step": 669
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 8.747474747474747e-06,
+      "loss": 1.8785,
+      "step": 670
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 8.745454545454546e-06,
+      "loss": 1.7207,
+      "step": 671
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 8.743434343434343e-06,
+      "loss": 2.1411,
+      "step": 672
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 8.741414141414142e-06,
+      "loss": 1.8279,
+      "step": 673
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 8.73939393939394e-06,
+      "loss": 1.7939,
+      "step": 674
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 8.737373737373738e-06,
+      "loss": 1.8367,
+      "step": 675
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 8.735353535353536e-06,
+      "loss": 1.8977,
+      "step": 676
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 8.733333333333333e-06,
+      "loss": 2.0187,
+      "step": 677
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 8.731313131313132e-06,
+      "loss": 2.4034,
+      "step": 678
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 8.729292929292929e-06,
+      "loss": 1.8761,
+      "step": 679
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 8.727272727272728e-06,
+      "loss": 1.9795,
+      "step": 680
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 8.725252525252525e-06,
+      "loss": 1.8917,
+      "step": 681
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 8.723232323232324e-06,
+      "loss": 1.9493,
+      "step": 682
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 8.72121212121212e-06,
+      "loss": 1.6774,
+      "step": 683
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 8.71919191919192e-06,
+      "loss": 1.4636,
+      "step": 684
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 8.717171717171718e-06,
+      "loss": 1.7278,
+      "step": 685
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 8.715151515151515e-06,
+      "loss": 1.8859,
+      "step": 686
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 8.713131313131314e-06,
+      "loss": 1.5925,
+      "step": 687
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 8.711111111111111e-06,
+      "loss": 2.1307,
+      "step": 688
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 8.70909090909091e-06,
+      "loss": 1.6724,
+      "step": 689
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 8.707070707070707e-06,
+      "loss": 2.1351,
+      "step": 690
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 8.705050505050507e-06,
+      "loss": 1.5232,
+      "step": 691
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 8.703030303030304e-06,
+      "loss": 1.9213,
+      "step": 692
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 8.701010101010103e-06,
+      "loss": 1.5634,
+      "step": 693
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 8.6989898989899e-06,
+      "loss": 1.7688,
+      "step": 694
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 8.696969696969699e-06,
+      "loss": 1.6219,
+      "step": 695
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 8.694949494949496e-06,
+      "loss": 1.7022,
+      "step": 696
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 8.692929292929294e-06,
+      "loss": 1.8848,
+      "step": 697
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 8.690909090909091e-06,
+      "loss": 1.8362,
+      "step": 698
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 8.68888888888889e-06,
+      "loss": 2.0544,
+      "step": 699
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 8.686868686868687e-06,
+      "loss": 1.9435,
+      "step": 700
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 8.684848484848486e-06,
+      "loss": 2.1384,
+      "step": 701
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 8.682828282828285e-06,
+      "loss": 1.2298,
+      "step": 702
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 8.680808080808082e-06,
+      "loss": 2.2155,
+      "step": 703
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 8.67878787878788e-06,
+      "loss": 1.7138,
+      "step": 704
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 8.676767676767678e-06,
+      "loss": 1.631,
+      "step": 705
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 8.674747474747476e-06,
+      "loss": 1.9203,
+      "step": 706
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 8.672727272727273e-06,
+      "loss": 2.0832,
+      "step": 707
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 8.670707070707072e-06,
+      "loss": 1.8483,
+      "step": 708
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 8.668686868686869e-06,
+      "loss": 2.0432,
+      "step": 709
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 8.666666666666668e-06,
+      "loss": 1.7995,
+      "step": 710
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 8.664646464646465e-06,
+      "loss": 2.0168,
+      "step": 711
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 8.662626262626264e-06,
+      "loss": 1.8457,
+      "step": 712
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 8.660606060606062e-06,
+      "loss": 2.0003,
+      "step": 713
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 8.65858585858586e-06,
+      "loss": 1.6838,
+      "step": 714
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 8.656565656565658e-06,
+      "loss": 1.3178,
+      "step": 715
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 8.654545454545455e-06,
+      "loss": 2.1532,
+      "step": 716
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 8.652525252525254e-06,
+      "loss": 1.7997,
+      "step": 717
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 8.650505050505051e-06,
+      "loss": 1.6028,
+      "step": 718
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 8.64848484848485e-06,
+      "loss": 2.022,
+      "step": 719
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 8.646464646464647e-06,
+      "loss": 1.8132,
+      "step": 720
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 8.644444444444445e-06,
+      "loss": 1.7446,
+      "step": 721
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 8.642424242424242e-06,
+      "loss": 1.4209,
+      "step": 722
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 8.640404040404041e-06,
+      "loss": 1.8225,
+      "step": 723
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 8.63838383838384e-06,
+      "loss": 1.6807,
+      "step": 724
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 8.636363636363637e-06,
+      "loss": 1.6718,
+      "step": 725
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 8.634343434343436e-06,
+      "loss": 1.4709,
+      "step": 726
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 8.632323232323233e-06,
+      "loss": 1.8182,
+      "step": 727
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 8.630303030303032e-06,
+      "loss": 1.4355,
+      "step": 728
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 8.628282828282829e-06,
+      "loss": 2.2152,
+      "step": 729
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 8.626262626262627e-06,
+      "loss": 1.6071,
+      "step": 730
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 8.624242424242424e-06,
+      "loss": 1.5322,
+      "step": 731
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 8.622222222222223e-06,
+      "loss": 1.8328,
+      "step": 732
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 8.62020202020202e-06,
+      "loss": 1.7799,
+      "step": 733
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 8.618181818181819e-06,
+      "loss": 1.6824,
+      "step": 734
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 8.616161616161618e-06,
+      "loss": 1.3896,
+      "step": 735
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 8.614141414141415e-06,
+      "loss": 1.5158,
+      "step": 736
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 8.612121212121213e-06,
+      "loss": 2.1152,
+      "step": 737
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 8.61010101010101e-06,
+      "loss": 1.8761,
+      "step": 738
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 8.608080808080809e-06,
+      "loss": 2.0583,
+      "step": 739
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 8.606060606060606e-06,
+      "loss": 1.7713,
+      "step": 740
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 8.604040404040405e-06,
+      "loss": 2.1803,
+      "step": 741
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 8.602020202020202e-06,
+      "loss": 1.7783,
+      "step": 742
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 8.6e-06,
+      "loss": 1.7986,
+      "step": 743
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 8.597979797979798e-06,
+      "loss": 1.7803,
+      "step": 744
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 8.595959595959596e-06,
+      "loss": 1.7083,
+      "step": 745
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 8.593939393939395e-06,
+      "loss": 1.835,
+      "step": 746
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 8.591919191919192e-06,
+      "loss": 1.8493,
+      "step": 747
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 8.589898989898991e-06,
+      "loss": 2.2879,
+      "step": 748
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 8.587878787878788e-06,
+      "loss": 1.7608,
+      "step": 749
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 8.585858585858587e-06,
+      "loss": 2.0727,
+      "step": 750
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 8.583838383838384e-06,
+      "loss": 1.5508,
+      "step": 751
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 8.581818181818183e-06,
+      "loss": 1.5714,
+      "step": 752
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 8.57979797979798e-06,
+      "loss": 1.6876,
+      "step": 753
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 8.577777777777778e-06,
+      "loss": 1.8579,
+      "step": 754
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 8.575757575757575e-06,
+      "loss": 1.5405,
+      "step": 755
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 8.573737373737374e-06,
+      "loss": 1.8325,
+      "step": 756
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 8.571717171717173e-06,
+      "loss": 1.8222,
+      "step": 757
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 8.56969696969697e-06,
+      "loss": 1.8166,
+      "step": 758
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 8.567676767676769e-06,
+      "loss": 1.5818,
+      "step": 759
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 8.565656565656566e-06,
+      "loss": 1.9261,
+      "step": 760
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 8.563636363636364e-06,
+      "loss": 2.1332,
+      "step": 761
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 8.561616161616161e-06,
+      "loss": 1.6806,
+      "step": 762
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 8.55959595959596e-06,
+      "loss": 2.1276,
+      "step": 763
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 8.557575757575757e-06,
+      "loss": 1.5969,
+      "step": 764
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 8.555555555555556e-06,
+      "loss": 1.6711,
+      "step": 765
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 8.553535353535355e-06,
+      "loss": 2.18,
+      "step": 766
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 8.551515151515152e-06,
+      "loss": 1.3844,
+      "step": 767
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 8.54949494949495e-06,
+      "loss": 2.0312,
+      "step": 768
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 8.547474747474747e-06,
+      "loss": 1.4345,
+      "step": 769
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 8.545454545454546e-06,
+      "loss": 1.4497,
+      "step": 770
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 8.543434343434343e-06,
+      "loss": 1.5907,
+      "step": 771
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 8.541414141414142e-06,
+      "loss": 1.8285,
+      "step": 772
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 8.539393939393939e-06,
+      "loss": 1.8619,
+      "step": 773
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 8.537373737373738e-06,
+      "loss": 1.5741,
+      "step": 774
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 8.535353535353535e-06,
+      "loss": 2.0152,
+      "step": 775
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 8.533333333333335e-06,
+      "loss": 2.0316,
+      "step": 776
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 8.531313131313132e-06,
+      "loss": 1.6834,
+      "step": 777
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 8.529292929292931e-06,
+      "loss": 1.7184,
+      "step": 778
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 8.527272727272728e-06,
+      "loss": 1.9087,
+      "step": 779
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 8.525252525252527e-06,
+      "loss": 1.6118,
+      "step": 780
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 8.523232323232324e-06,
+      "loss": 1.6897,
+      "step": 781
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 8.521212121212123e-06,
+      "loss": 1.8007,
+      "step": 782
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 8.51919191919192e-06,
+      "loss": 1.5489,
+      "step": 783
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 8.517171717171718e-06,
+      "loss": 1.583,
+      "step": 784
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 8.515151515151517e-06,
+      "loss": 1.4073,
+      "step": 785
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 8.513131313131314e-06,
+      "loss": 1.4817,
+      "step": 786
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 8.511111111111113e-06,
+      "loss": 1.593,
+      "step": 787
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 8.50909090909091e-06,
+      "loss": 1.8723,
+      "step": 788
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 8.507070707070709e-06,
+      "loss": 1.9124,
+      "step": 789
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 8.505050505050506e-06,
+      "loss": 1.7036,
+      "step": 790
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 8.503030303030304e-06,
+      "loss": 1.8998,
+      "step": 791
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 8.501010101010101e-06,
+      "loss": 1.7117,
+      "step": 792
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 8.4989898989899e-06,
+      "loss": 1.4787,
+      "step": 793
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 8.496969696969697e-06,
+      "loss": 1.3619,
+      "step": 794
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 8.494949494949496e-06,
+      "loss": 1.7042,
+      "step": 795
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 8.492929292929295e-06,
+      "loss": 1.7931,
+      "step": 796
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 8.490909090909092e-06,
+      "loss": 1.5717,
+      "step": 797
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 8.48888888888889e-06,
+      "loss": 2.3016,
+      "step": 798
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 8.486868686868687e-06,
+      "loss": 2.0792,
+      "step": 799
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 8.484848484848486e-06,
+      "loss": 1.5883,
+      "step": 800
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.48,
+      "eval_loss": 1.6317851543426514,
+      "eval_roc_auc": 0.8900486426988232,
+      "eval_runtime": 94.3838,
+      "eval_samples_per_second": 2.119,
+      "eval_steps_per_second": 0.53,
+      "step": 800
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 8.482828282828283e-06,
+      "loss": 1.5819,
+      "step": 801
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 8.480808080808082e-06,
+      "loss": 1.9615,
+      "step": 802
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 8.478787878787879e-06,
+      "loss": 1.6725,
+      "step": 803
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 8.476767676767678e-06,
+      "loss": 1.7433,
+      "step": 804
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 8.474747474747475e-06,
+      "loss": 1.7445,
+      "step": 805
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 8.472727272727274e-06,
+      "loss": 1.953,
+      "step": 806
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 8.470707070707072e-06,
+      "loss": 1.7869,
+      "step": 807
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 8.46868686868687e-06,
+      "loss": 1.8632,
+      "step": 808
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 8.466666666666668e-06,
+      "loss": 1.8337,
+      "step": 809
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 8.464646464646465e-06,
+      "loss": 1.7074,
+      "step": 810
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 8.462626262626264e-06,
+      "loss": 1.4912,
+      "step": 811
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 8.460606060606061e-06,
+      "loss": 1.4665,
+      "step": 812
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 8.45858585858586e-06,
+      "loss": 2.0954,
+      "step": 813
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 8.456565656565657e-06,
+      "loss": 1.3293,
+      "step": 814
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 8.454545454545455e-06,
+      "loss": 1.7759,
+      "step": 815
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 8.452525252525252e-06,
+      "loss": 1.8222,
+      "step": 816
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 8.450505050505051e-06,
+      "loss": 1.8386,
+      "step": 817
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 8.44848484848485e-06,
+      "loss": 1.5066,
+      "step": 818
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 8.446464646464647e-06,
+      "loss": 1.7193,
+      "step": 819
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 8.444444444444446e-06,
+      "loss": 2.1736,
+      "step": 820
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 8.442424242424243e-06,
+      "loss": 1.9739,
+      "step": 821
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 8.440404040404041e-06,
+      "loss": 2.0769,
+      "step": 822
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 8.438383838383838e-06,
+      "loss": 1.5874,
+      "step": 823
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 8.436363636363637e-06,
+      "loss": 1.379,
+      "step": 824
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 8.434343434343434e-06,
+      "loss": 1.7697,
+      "step": 825
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 8.432323232323233e-06,
+      "loss": 1.565,
+      "step": 826
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 8.43030303030303e-06,
+      "loss": 1.6453,
+      "step": 827
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 8.428282828282829e-06,
+      "loss": 1.2366,
+      "step": 828
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 8.426262626262628e-06,
+      "loss": 1.4356,
+      "step": 829
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 8.424242424242425e-06,
+      "loss": 1.4136,
+      "step": 830
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 8.422222222222223e-06,
+      "loss": 1.8168,
+      "step": 831
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 8.42020202020202e-06,
+      "loss": 1.7014,
+      "step": 832
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 8.418181818181819e-06,
+      "loss": 1.9966,
+      "step": 833
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 8.416161616161616e-06,
+      "loss": 1.7876,
+      "step": 834
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 8.414141414141415e-06,
+      "loss": 1.7393,
+      "step": 835
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 8.412121212121212e-06,
+      "loss": 1.696,
+      "step": 836
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 8.41010101010101e-06,
+      "loss": 2.1278,
+      "step": 837
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 8.40808080808081e-06,
+      "loss": 1.8949,
+      "step": 838
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 8.406060606060606e-06,
+      "loss": 1.3353,
+      "step": 839
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 8.404040404040405e-06,
+      "loss": 1.2373,
+      "step": 840
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 8.402020202020202e-06,
+      "loss": 1.7867,
+      "step": 841
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 1.2405,
+      "step": 842
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 8.397979797979798e-06,
+      "loss": 2.0386,
+      "step": 843
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 8.395959595959597e-06,
+      "loss": 1.9334,
+      "step": 844
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 8.393939393939394e-06,
+      "loss": 1.5761,
+      "step": 845
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 8.391919191919192e-06,
+      "loss": 2.0102,
+      "step": 846
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 8.38989898989899e-06,
+      "loss": 1.8049,
+      "step": 847
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 8.387878787878788e-06,
+      "loss": 1.3771,
+      "step": 848
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 8.385858585858587e-06,
+      "loss": 1.7976,
+      "step": 849
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 8.383838383838384e-06,
+      "loss": 1.9619,
+      "step": 850
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 8.381818181818183e-06,
+      "loss": 1.6369,
+      "step": 851
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 8.37979797979798e-06,
+      "loss": 1.5252,
+      "step": 852
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 8.377777777777779e-06,
+      "loss": 1.9233,
+      "step": 853
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 8.375757575757576e-06,
+      "loss": 1.7804,
+      "step": 854
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 8.373737373737374e-06,
+      "loss": 1.7506,
+      "step": 855
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 8.371717171717171e-06,
+      "loss": 1.5923,
+      "step": 856
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 8.36969696969697e-06,
+      "loss": 1.6559,
+      "step": 857
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 8.367676767676767e-06,
+      "loss": 1.6017,
+      "step": 858
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 8.365656565656566e-06,
+      "loss": 1.4229,
+      "step": 859
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 8.363636363636365e-06,
+      "loss": 1.8642,
+      "step": 860
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 8.361616161616163e-06,
+      "loss": 1.6416,
+      "step": 861
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 8.35959595959596e-06,
+      "loss": 1.7493,
+      "step": 862
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 8.357575757575759e-06,
+      "loss": 1.9163,
+      "step": 863
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 8.355555555555556e-06,
+      "loss": 1.6576,
+      "step": 864
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 8.353535353535355e-06,
+      "loss": 1.7268,
+      "step": 865
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 8.351515151515152e-06,
+      "loss": 1.6183,
+      "step": 866
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 8.34949494949495e-06,
+      "loss": 1.423,
+      "step": 867
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 8.34747474747475e-06,
+      "loss": 1.8499,
+      "step": 868
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 8.345454545454546e-06,
+      "loss": 1.4667,
+      "step": 869
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 8.343434343434345e-06,
+      "loss": 1.4169,
+      "step": 870
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 8.341414141414142e-06,
+      "loss": 2.1424,
+      "step": 871
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 8.339393939393941e-06,
+      "loss": 2.6316,
+      "step": 872
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 8.337373737373738e-06,
+      "loss": 1.8469,
+      "step": 873
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 8.335353535353537e-06,
+      "loss": 1.84,
+      "step": 874
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 1.5256,
+      "step": 875
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 8.331313131313132e-06,
+      "loss": 2.2097,
+      "step": 876
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 8.32929292929293e-06,
+      "loss": 1.6705,
+      "step": 877
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 8.327272727272728e-06,
+      "loss": 1.841,
+      "step": 878
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 8.325252525252527e-06,
+      "loss": 1.0743,
+      "step": 879
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 8.323232323232324e-06,
+      "loss": 1.7568,
+      "step": 880
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 8.321212121212123e-06,
+      "loss": 1.8788,
+      "step": 881
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 8.31919191919192e-06,
+      "loss": 1.674,
+      "step": 882
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 8.317171717171719e-06,
+      "loss": 1.3443,
+      "step": 883
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 8.315151515151516e-06,
+      "loss": 1.9045,
+      "step": 884
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 8.313131313131314e-06,
+      "loss": 1.3657,
+      "step": 885
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 8.311111111111111e-06,
+      "loss": 1.737,
+      "step": 886
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 8.30909090909091e-06,
+      "loss": 1.7678,
+      "step": 887
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 8.307070707070707e-06,
+      "loss": 1.9097,
+      "step": 888
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 8.305050505050506e-06,
+      "loss": 1.3329,
+      "step": 889
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 8.303030303030305e-06,
+      "loss": 1.5352,
+      "step": 890
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 8.301010101010102e-06,
+      "loss": 1.6685,
+      "step": 891
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 8.2989898989899e-06,
+      "loss": 1.7849,
+      "step": 892
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 8.296969696969697e-06,
+      "loss": 1.353,
+      "step": 893
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 8.294949494949496e-06,
+      "loss": 1.3734,
+      "step": 894
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 8.292929292929293e-06,
+      "loss": 1.7695,
+      "step": 895
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 8.290909090909092e-06,
+      "loss": 1.43,
+      "step": 896
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 8.288888888888889e-06,
+      "loss": 1.4631,
+      "step": 897
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 8.286868686868688e-06,
+      "loss": 1.3667,
+      "step": 898
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 8.284848484848486e-06,
+      "loss": 1.5946,
+      "step": 899
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 8.282828282828283e-06,
+      "loss": 1.9412,
+      "step": 900
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 8.280808080808082e-06,
+      "loss": 1.5134,
+      "step": 901
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 8.27878787878788e-06,
+      "loss": 1.5238,
+      "step": 902
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 8.276767676767678e-06,
+      "loss": 1.6126,
+      "step": 903
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 8.274747474747475e-06,
+      "loss": 1.5469,
+      "step": 904
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 8.272727272727274e-06,
+      "loss": 1.6669,
+      "step": 905
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 8.27070707070707e-06,
+      "loss": 1.7708,
+      "step": 906
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 8.26868686868687e-06,
+      "loss": 1.5447,
+      "step": 907
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 8.266666666666667e-06,
+      "loss": 1.6817,
+      "step": 908
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 8.264646464646465e-06,
+      "loss": 1.3813,
+      "step": 909
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 8.262626262626264e-06,
+      "loss": 1.6298,
+      "step": 910
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 8.260606060606061e-06,
+      "loss": 1.6147,
+      "step": 911
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 8.25858585858586e-06,
+      "loss": 1.7234,
+      "step": 912
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 8.256565656565657e-06,
+      "loss": 1.9258,
+      "step": 913
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 8.254545454545456e-06,
+      "loss": 1.5423,
+      "step": 914
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 8.252525252525253e-06,
+      "loss": 1.5438,
+      "step": 915
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 8.250505050505051e-06,
+      "loss": 1.8715,
+      "step": 916
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 8.248484848484848e-06,
+      "loss": 1.3605,
+      "step": 917
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 8.246464646464647e-06,
+      "loss": 1.6098,
+      "step": 918
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 8.244444444444444e-06,
+      "loss": 1.8543,
+      "step": 919
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 8.242424242424243e-06,
+      "loss": 1.6414,
+      "step": 920
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 8.240404040404042e-06,
+      "loss": 0.8965,
+      "step": 921
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 8.238383838383839e-06,
+      "loss": 1.1945,
+      "step": 922
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 8.236363636363637e-06,
+      "loss": 1.7522,
+      "step": 923
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 8.234343434343434e-06,
+      "loss": 1.8057,
+      "step": 924
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 8.232323232323233e-06,
+      "loss": 1.9094,
+      "step": 925
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 8.23030303030303e-06,
+      "loss": 1.7397,
+      "step": 926
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 8.228282828282829e-06,
+      "loss": 1.4837,
+      "step": 927
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 8.226262626262626e-06,
+      "loss": 1.6094,
+      "step": 928
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 8.224242424242425e-06,
+      "loss": 1.9702,
+      "step": 929
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 8.222222222222222e-06,
+      "loss": 1.9368,
+      "step": 930
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 8.22020202020202e-06,
+      "loss": 1.7571,
+      "step": 931
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 8.21818181818182e-06,
+      "loss": 1.4858,
+      "step": 932
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 8.216161616161616e-06,
+      "loss": 1.5127,
+      "step": 933
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 8.214141414141415e-06,
+      "loss": 1.782,
+      "step": 934
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 8.212121212121212e-06,
+      "loss": 1.8983,
+      "step": 935
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 8.21010101010101e-06,
+      "loss": 1.4163,
+      "step": 936
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 8.208080808080808e-06,
+      "loss": 1.5235,
+      "step": 937
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 8.206060606060607e-06,
+      "loss": 1.5048,
+      "step": 938
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 8.204040404040404e-06,
+      "loss": 2.0274,
+      "step": 939
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 8.202020202020202e-06,
+      "loss": 1.4352,
+      "step": 940
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 8.2e-06,
+      "loss": 1.7432,
+      "step": 941
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 8.197979797979798e-06,
+      "loss": 1.8162,
+      "step": 942
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 8.195959595959597e-06,
+      "loss": 1.6283,
+      "step": 943
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 8.193939393939394e-06,
+      "loss": 1.5304,
+      "step": 944
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 8.191919191919193e-06,
+      "loss": 2.1163,
+      "step": 945
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 8.18989898989899e-06,
+      "loss": 1.6713,
+      "step": 946
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 8.187878787878788e-06,
+      "loss": 1.2994,
+      "step": 947
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 8.185858585858587e-06,
+      "loss": 2.0603,
+      "step": 948
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 8.183838383838384e-06,
+      "loss": 1.7561,
+      "step": 949
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 8.181818181818183e-06,
+      "loss": 2.0067,
+      "step": 950
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 8.179797979797982e-06,
+      "loss": 1.6532,
+      "step": 951
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 8.177777777777779e-06,
+      "loss": 1.3186,
+      "step": 952
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 8.175757575757577e-06,
+      "loss": 1.5827,
+      "step": 953
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 8.173737373737375e-06,
+      "loss": 1.7459,
+      "step": 954
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 8.171717171717173e-06,
+      "loss": 1.3492,
+      "step": 955
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 8.16969696969697e-06,
+      "loss": 1.4013,
+      "step": 956
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 8.167676767676769e-06,
+      "loss": 2.0865,
+      "step": 957
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 8.165656565656566e-06,
+      "loss": 1.8408,
+      "step": 958
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 8.163636363636365e-06,
+      "loss": 1.1432,
+      "step": 959
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 8.161616161616162e-06,
+      "loss": 1.7224,
+      "step": 960
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 8.15959595959596e-06,
+      "loss": 1.5601,
+      "step": 961
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 8.15757575757576e-06,
+      "loss": 1.4813,
+      "step": 962
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 8.155555555555556e-06,
+      "loss": 1.4877,
+      "step": 963
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 8.153535353535355e-06,
+      "loss": 1.3356,
+      "step": 964
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 8.151515151515152e-06,
+      "loss": 1.28,
+      "step": 965
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 8.149494949494951e-06,
+      "loss": 1.4121,
+      "step": 966
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 8.147474747474748e-06,
+      "loss": 1.8524,
+      "step": 967
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 8.145454545454547e-06,
+      "loss": 1.839,
+      "step": 968
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 8.143434343434344e-06,
+      "loss": 1.8646,
+      "step": 969
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 8.141414141414142e-06,
+      "loss": 1.5101,
+      "step": 970
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 8.139393939393941e-06,
+      "loss": 1.5271,
+      "step": 971
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 8.137373737373738e-06,
+      "loss": 1.3243,
+      "step": 972
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 8.135353535353537e-06,
+      "loss": 1.6156,
+      "step": 973
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 8.133333333333334e-06,
+      "loss": 1.563,
+      "step": 974
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 8.131313131313133e-06,
+      "loss": 1.6073,
+      "step": 975
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 8.12929292929293e-06,
+      "loss": 1.5167,
+      "step": 976
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 8.127272727272728e-06,
+      "loss": 1.9048,
+      "step": 977
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 8.125252525252526e-06,
+      "loss": 1.3705,
+      "step": 978
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 8.123232323232324e-06,
+      "loss": 1.7965,
+      "step": 979
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 8.121212121212121e-06,
+      "loss": 1.974,
+      "step": 980
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 8.11919191919192e-06,
+      "loss": 1.7263,
+      "step": 981
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 8.117171717171719e-06,
+      "loss": 1.7301,
+      "step": 982
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 8.115151515151516e-06,
+      "loss": 1.8765,
+      "step": 983
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 8.113131313131315e-06,
+      "loss": 1.4722,
+      "step": 984
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 8.111111111111112e-06,
+      "loss": 1.9797,
+      "step": 985
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 8.10909090909091e-06,
+      "loss": 1.5385,
+      "step": 986
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 8.107070707070707e-06,
+      "loss": 1.7032,
+      "step": 987
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 8.105050505050506e-06,
+      "loss": 1.6707,
+      "step": 988
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 8.103030303030303e-06,
+      "loss": 1.797,
+      "step": 989
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 8.101010101010102e-06,
+      "loss": 1.9198,
+      "step": 990
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 8.098989898989899e-06,
+      "loss": 1.604,
+      "step": 991
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 8.096969696969698e-06,
+      "loss": 2.3061,
+      "step": 992
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 8.094949494949496e-06,
+      "loss": 1.5951,
+      "step": 993
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 8.092929292929293e-06,
+      "loss": 1.4968,
+      "step": 994
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 8.090909090909092e-06,
+      "loss": 1.7666,
+      "step": 995
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 8.08888888888889e-06,
+      "loss": 1.6348,
+      "step": 996
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 8.086868686868688e-06,
+      "loss": 1.6645,
+      "step": 997
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 8.084848484848485e-06,
+      "loss": 1.5372,
+      "step": 998
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 8.082828282828284e-06,
+      "loss": 1.6573,
+      "step": 999
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 8.08080808080808e-06,
+      "loss": 1.3683,
+      "step": 1000
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.465,
+      "eval_loss": 1.4937936067581177,
+      "eval_roc_auc": 0.9224649175963098,
+      "eval_runtime": 94.4011,
+      "eval_samples_per_second": 2.119,
+      "eval_steps_per_second": 0.53,
+      "step": 1000
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 8.07878787878788e-06,
+      "loss": 1.808,
+      "step": 1001
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 8.076767676767677e-06,
+      "loss": 1.413,
+      "step": 1002
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 8.074747474747475e-06,
+      "loss": 1.5851,
+      "step": 1003
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 8.072727272727274e-06,
+      "loss": 1.7919,
+      "step": 1004
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 8.070707070707071e-06,
+      "loss": 1.4952,
+      "step": 1005
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 8.06868686868687e-06,
+      "loss": 1.6555,
+      "step": 1006
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 8.066666666666667e-06,
+      "loss": 1.6528,
+      "step": 1007
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 8.064646464646466e-06,
+      "loss": 1.746,
+      "step": 1008
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 8.062626262626263e-06,
+      "loss": 1.4633,
+      "step": 1009
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 8.060606060606061e-06,
+      "loss": 1.4555,
+      "step": 1010
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 8.058585858585858e-06,
+      "loss": 1.8598,
+      "step": 1011
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 8.056565656565657e-06,
+      "loss": 1.9963,
+      "step": 1012
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 8.054545454545454e-06,
+      "loss": 1.7138,
+      "step": 1013
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 8.052525252525253e-06,
+      "loss": 1.643,
+      "step": 1014
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 8.050505050505052e-06,
+      "loss": 1.6401,
+      "step": 1015
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 8.048484848484849e-06,
+      "loss": 1.5452,
+      "step": 1016
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 8.046464646464647e-06,
+      "loss": 1.6658,
+      "step": 1017
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 8.044444444444444e-06,
+      "loss": 1.5085,
+      "step": 1018
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 8.042424242424243e-06,
+      "loss": 1.6656,
+      "step": 1019
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 8.04040404040404e-06,
+      "loss": 1.5249,
+      "step": 1020
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 8.038383838383839e-06,
+      "loss": 1.8615,
+      "step": 1021
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 8.036363636363636e-06,
+      "loss": 1.858,
+      "step": 1022
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 8.034343434343435e-06,
+      "loss": 1.7634,
+      "step": 1023
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 8.032323232323232e-06,
+      "loss": 1.8365,
+      "step": 1024
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 8.03030303030303e-06,
+      "loss": 1.2216,
+      "step": 1025
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 8.02828282828283e-06,
+      "loss": 1.5759,
+      "step": 1026
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 8.026262626262626e-06,
+      "loss": 1.1067,
+      "step": 1027
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 8.024242424242425e-06,
+      "loss": 1.4505,
+      "step": 1028
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 8.022222222222222e-06,
+      "loss": 1.7712,
+      "step": 1029
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 8.02020202020202e-06,
+      "loss": 1.8219,
+      "step": 1030
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 8.018181818181818e-06,
+      "loss": 1.7756,
+      "step": 1031
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 8.016161616161617e-06,
+      "loss": 1.5282,
+      "step": 1032
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 8.014141414141415e-06,
+      "loss": 1.2497,
+      "step": 1033
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 8.012121212121214e-06,
+      "loss": 1.3412,
+      "step": 1034
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 8.010101010101011e-06,
+      "loss": 1.2348,
+      "step": 1035
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 8.00808080808081e-06,
+      "loss": 1.6556,
+      "step": 1036
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 8.006060606060607e-06,
+      "loss": 1.2185,
+      "step": 1037
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 8.004040404040406e-06,
+      "loss": 1.8322,
+      "step": 1038
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 8.002020202020203e-06,
+      "loss": 1.6641,
+      "step": 1039
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.8441,
+      "step": 1040
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 7.997979797979798e-06,
+      "loss": 1.8821,
+      "step": 1041
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 7.995959595959597e-06,
+      "loss": 1.4581,
+      "step": 1042
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 7.993939393939396e-06,
+      "loss": 1.4762,
+      "step": 1043
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 7.991919191919193e-06,
+      "loss": 1.5577,
+      "step": 1044
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 7.989898989898992e-06,
+      "loss": 1.6603,
+      "step": 1045
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 7.987878787878789e-06,
+      "loss": 1.3383,
+      "step": 1046
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 7.985858585858587e-06,
+      "loss": 1.6181,
+      "step": 1047
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 7.983838383838384e-06,
+      "loss": 1.1631,
+      "step": 1048
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 7.981818181818183e-06,
+      "loss": 1.6121,
+      "step": 1049
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 7.97979797979798e-06,
+      "loss": 1.719,
+      "step": 1050
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 7.977777777777779e-06,
+      "loss": 1.2095,
+      "step": 1051
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 7.975757575757576e-06,
+      "loss": 1.3527,
+      "step": 1052
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 7.973737373737375e-06,
+      "loss": 1.7915,
+      "step": 1053
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 7.971717171717173e-06,
+      "loss": 1.5935,
+      "step": 1054
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 7.96969696969697e-06,
+      "loss": 1.2649,
+      "step": 1055
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 7.96767676767677e-06,
+      "loss": 1.9372,
+      "step": 1056
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 7.965656565656566e-06,
+      "loss": 1.8498,
+      "step": 1057
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 7.963636363636365e-06,
+      "loss": 1.2346,
+      "step": 1058
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 7.961616161616162e-06,
+      "loss": 1.7929,
+      "step": 1059
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 7.95959595959596e-06,
+      "loss": 1.9671,
+      "step": 1060
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 7.957575757575758e-06,
+      "loss": 1.5999,
+      "step": 1061
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 7.955555555555557e-06,
+      "loss": 1.5474,
+      "step": 1062
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 7.953535353535354e-06,
+      "loss": 1.4569,
+      "step": 1063
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 7.951515151515152e-06,
+      "loss": 1.3223,
+      "step": 1064
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 7.949494949494951e-06,
+      "loss": 1.7063,
+      "step": 1065
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 7.947474747474748e-06,
+      "loss": 1.6168,
+      "step": 1066
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 7.945454545454547e-06,
+      "loss": 1.7728,
+      "step": 1067
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 7.943434343434344e-06,
+      "loss": 1.9749,
+      "step": 1068
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 7.941414141414143e-06,
+      "loss": 1.5722,
+      "step": 1069
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 7.93939393939394e-06,
+      "loss": 1.8109,
+      "step": 1070
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 7.937373737373738e-06,
+      "loss": 1.5001,
+      "step": 1071
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 7.935353535353535e-06,
+      "loss": 1.175,
+      "step": 1072
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 7.933333333333334e-06,
+      "loss": 1.637,
+      "step": 1073
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 7.931313131313131e-06,
+      "loss": 1.8343,
+      "step": 1074
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 7.92929292929293e-06,
+      "loss": 1.5053,
+      "step": 1075
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 7.927272727272729e-06,
+      "loss": 1.5783,
+      "step": 1076
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 7.925252525252526e-06,
+      "loss": 1.0354,
+      "step": 1077
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 7.923232323232324e-06,
+      "loss": 1.486,
+      "step": 1078
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 7.921212121212122e-06,
+      "loss": 1.5888,
+      "step": 1079
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 7.91919191919192e-06,
+      "loss": 1.4969,
+      "step": 1080
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 7.917171717171717e-06,
+      "loss": 1.4519,
+      "step": 1081
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 7.915151515151516e-06,
+      "loss": 1.6706,
+      "step": 1082
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 7.913131313131313e-06,
+      "loss": 1.9074,
+      "step": 1083
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 7.911111111111112e-06,
+      "loss": 1.2433,
+      "step": 1084
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 7.909090909090909e-06,
+      "loss": 1.5971,
+      "step": 1085
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 7.907070707070708e-06,
+      "loss": 2.1256,
+      "step": 1086
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 7.905050505050506e-06,
+      "loss": 1.8915,
+      "step": 1087
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 7.903030303030303e-06,
+      "loss": 1.3233,
+      "step": 1088
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 7.901010101010102e-06,
+      "loss": 1.9943,
+      "step": 1089
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 7.898989898989899e-06,
+      "loss": 0.7403,
+      "step": 1090
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 7.896969696969698e-06,
+      "loss": 1.686,
+      "step": 1091
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 7.894949494949495e-06,
+      "loss": 1.2833,
+      "step": 1092
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 7.892929292929294e-06,
+      "loss": 1.3863,
+      "step": 1093
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 7.89090909090909e-06,
+      "loss": 1.7024,
+      "step": 1094
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 7.88888888888889e-06,
+      "loss": 1.4984,
+      "step": 1095
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 7.886868686868686e-06,
+      "loss": 1.7376,
+      "step": 1096
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 7.884848484848485e-06,
+      "loss": 1.5809,
+      "step": 1097
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 7.882828282828284e-06,
+      "loss": 1.9398,
+      "step": 1098
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 7.880808080808081e-06,
+      "loss": 1.1599,
+      "step": 1099
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 7.87878787878788e-06,
+      "loss": 1.6224,
+      "step": 1100
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 7.876767676767677e-06,
+      "loss": 1.4148,
+      "step": 1101
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 7.874747474747475e-06,
+      "loss": 1.771,
+      "step": 1102
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 7.872727272727273e-06,
+      "loss": 1.2999,
+      "step": 1103
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 7.870707070707071e-06,
+      "loss": 1.3769,
+      "step": 1104
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 7.868686868686868e-06,
+      "loss": 1.239,
+      "step": 1105
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 7.866666666666667e-06,
+      "loss": 1.5177,
+      "step": 1106
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 7.864646464646464e-06,
+      "loss": 1.015,
+      "step": 1107
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 7.862626262626263e-06,
+      "loss": 1.8084,
+      "step": 1108
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 7.860606060606062e-06,
+      "loss": 1.4093,
+      "step": 1109
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 7.858585858585859e-06,
+      "loss": 1.3053,
+      "step": 1110
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 7.856565656565657e-06,
+      "loss": 1.3188,
+      "step": 1111
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 7.854545454545454e-06,
+      "loss": 1.1476,
+      "step": 1112
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 7.852525252525253e-06,
+      "loss": 1.6501,
+      "step": 1113
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 7.85050505050505e-06,
+      "loss": 1.3974,
+      "step": 1114
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 7.848484848484849e-06,
+      "loss": 1.4001,
+      "step": 1115
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 7.846464646464646e-06,
+      "loss": 1.3374,
+      "step": 1116
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 7.844444444444446e-06,
+      "loss": 1.114,
+      "step": 1117
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 7.842424242424243e-06,
+      "loss": 1.5525,
+      "step": 1118
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 7.840404040404042e-06,
+      "loss": 1.4569,
+      "step": 1119
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 7.838383838383839e-06,
+      "loss": 1.2699,
+      "step": 1120
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 7.836363636363638e-06,
+      "loss": 1.5751,
+      "step": 1121
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 7.834343434343435e-06,
+      "loss": 1.2228,
+      "step": 1122
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 7.832323232323234e-06,
+      "loss": 0.9674,
+      "step": 1123
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 7.83030303030303e-06,
+      "loss": 1.4719,
+      "step": 1124
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 7.82828282828283e-06,
+      "loss": 1.6733,
+      "step": 1125
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 7.826262626262628e-06,
+      "loss": 1.8986,
+      "step": 1126
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 7.824242424242425e-06,
+      "loss": 1.4217,
+      "step": 1127
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 7.822222222222224e-06,
+      "loss": 1.3743,
+      "step": 1128
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 7.820202020202021e-06,
+      "loss": 1.7173,
+      "step": 1129
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 7.81818181818182e-06,
+      "loss": 1.0349,
+      "step": 1130
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 7.816161616161617e-06,
+      "loss": 1.552,
+      "step": 1131
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 7.814141414141415e-06,
+      "loss": 1.7526,
+      "step": 1132
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 7.812121212121213e-06,
+      "loss": 1.1192,
+      "step": 1133
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 7.810101010101011e-06,
+      "loss": 1.3526,
+      "step": 1134
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 7.808080808080808e-06,
+      "loss": 2.1168,
+      "step": 1135
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 7.806060606060607e-06,
+      "loss": 1.5744,
+      "step": 1136
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 7.804040404040406e-06,
+      "loss": 1.4374,
+      "step": 1137
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 7.802020202020203e-06,
+      "loss": 1.6945,
+      "step": 1138
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 7.800000000000002e-06,
+      "loss": 1.1393,
+      "step": 1139
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 7.797979797979799e-06,
+      "loss": 1.4361,
+      "step": 1140
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 7.795959595959597e-06,
+      "loss": 1.6594,
+      "step": 1141
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 7.793939393939394e-06,
+      "loss": 0.7759,
+      "step": 1142
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 7.791919191919193e-06,
+      "loss": 1.284,
+      "step": 1143
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 7.78989898989899e-06,
+      "loss": 1.2874,
+      "step": 1144
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 7.787878787878789e-06,
+      "loss": 1.6279,
+      "step": 1145
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 7.785858585858586e-06,
+      "loss": 1.697,
+      "step": 1146
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 7.783838383838385e-06,
+      "loss": 1.8136,
+      "step": 1147
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 7.781818181818183e-06,
+      "loss": 1.491,
+      "step": 1148
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 7.77979797979798e-06,
+      "loss": 2.0518,
+      "step": 1149
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 7.77777777777778e-06,
+      "loss": 1.1382,
+      "step": 1150
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 7.775757575757576e-06,
+      "loss": 1.4593,
+      "step": 1151
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 7.773737373737375e-06,
+      "loss": 1.2436,
+      "step": 1152
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 7.771717171717172e-06,
+      "loss": 1.6364,
+      "step": 1153
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 7.76969696969697e-06,
+      "loss": 1.837,
+      "step": 1154
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 7.767676767676768e-06,
+      "loss": 1.3268,
+      "step": 1155
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 7.765656565656566e-06,
+      "loss": 1.4753,
+      "step": 1156
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 7.763636363636364e-06,
+      "loss": 1.1689,
+      "step": 1157
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 7.761616161616162e-06,
+      "loss": 1.3778,
+      "step": 1158
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 7.759595959595961e-06,
+      "loss": 1.5626,
+      "step": 1159
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 7.757575757575758e-06,
+      "loss": 1.5306,
+      "step": 1160
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 7.755555555555557e-06,
+      "loss": 1.9524,
+      "step": 1161
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 7.753535353535354e-06,
+      "loss": 1.7318,
+      "step": 1162
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 7.751515151515153e-06,
+      "loss": 1.5586,
+      "step": 1163
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 7.74949494949495e-06,
+      "loss": 1.2951,
+      "step": 1164
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 7.747474747474748e-06,
+      "loss": 1.4736,
+      "step": 1165
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 7.745454545454545e-06,
+      "loss": 1.4006,
+      "step": 1166
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 7.743434343434344e-06,
+      "loss": 2.0625,
+      "step": 1167
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 7.741414141414141e-06,
+      "loss": 1.8369,
+      "step": 1168
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 7.73939393939394e-06,
+      "loss": 1.742,
+      "step": 1169
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 7.737373737373739e-06,
+      "loss": 1.639,
+      "step": 1170
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 7.735353535353536e-06,
+      "loss": 1.4681,
+      "step": 1171
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 7.733333333333334e-06,
+      "loss": 1.2993,
+      "step": 1172
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 7.731313131313131e-06,
+      "loss": 1.6361,
+      "step": 1173
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 7.72929292929293e-06,
+      "loss": 1.1108,
+      "step": 1174
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 7.727272727272727e-06,
+      "loss": 1.5705,
+      "step": 1175
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 7.725252525252526e-06,
+      "loss": 1.6962,
+      "step": 1176
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 7.723232323232323e-06,
+      "loss": 0.9462,
+      "step": 1177
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 7.721212121212122e-06,
+      "loss": 1.7139,
+      "step": 1178
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 7.719191919191919e-06,
+      "loss": 1.6192,
+      "step": 1179
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 7.717171717171717e-06,
+      "loss": 1.8822,
+      "step": 1180
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 7.715151515151516e-06,
+      "loss": 1.5869,
+      "step": 1181
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 7.713131313131313e-06,
+      "loss": 1.8346,
+      "step": 1182
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 7.711111111111112e-06,
+      "loss": 1.6463,
+      "step": 1183
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 7.709090909090909e-06,
+      "loss": 1.8482,
+      "step": 1184
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 7.707070707070708e-06,
+      "loss": 1.3378,
+      "step": 1185
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 7.705050505050505e-06,
+      "loss": 1.6229,
+      "step": 1186
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 7.703030303030304e-06,
+      "loss": 1.592,
+      "step": 1187
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 7.7010101010101e-06,
+      "loss": 2.1445,
+      "step": 1188
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 7.6989898989899e-06,
+      "loss": 1.251,
+      "step": 1189
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 7.696969696969696e-06,
+      "loss": 1.5814,
+      "step": 1190
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 7.694949494949495e-06,
+      "loss": 1.8232,
+      "step": 1191
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 7.692929292929294e-06,
+      "loss": 1.8317,
+      "step": 1192
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 7.690909090909091e-06,
+      "loss": 1.8274,
+      "step": 1193
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 7.68888888888889e-06,
+      "loss": 1.1979,
+      "step": 1194
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 7.686868686868687e-06,
+      "loss": 1.6625,
+      "step": 1195
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 7.684848484848485e-06,
+      "loss": 1.6434,
+      "step": 1196
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 7.682828282828282e-06,
+      "loss": 0.8596,
+      "step": 1197
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 7.680808080808081e-06,
+      "loss": 1.2386,
+      "step": 1198
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 7.678787878787878e-06,
+      "loss": 1.5361,
+      "step": 1199
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 7.676767676767677e-06,
+      "loss": 1.8529,
+      "step": 1200
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.585,
+      "eval_loss": 1.3470025062561035,
+      "eval_roc_auc": 0.936347526228635,
+      "eval_runtime": 94.9551,
+      "eval_samples_per_second": 2.106,
+      "eval_steps_per_second": 0.527,
+      "step": 1200
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 7.674747474747474e-06,
+      "loss": 1.5285,
+      "step": 1201
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 7.672727272727273e-06,
+      "loss": 1.277,
+      "step": 1202
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 7.670707070707071e-06,
+      "loss": 1.5494,
+      "step": 1203
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 7.66868686868687e-06,
+      "loss": 1.1058,
+      "step": 1204
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 7.666666666666667e-06,
+      "loss": 1.2355,
+      "step": 1205
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 7.664646464646466e-06,
+      "loss": 1.0584,
+      "step": 1206
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 7.662626262626263e-06,
+      "loss": 1.3426,
+      "step": 1207
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 7.660606060606062e-06,
+      "loss": 1.7641,
+      "step": 1208
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 7.65858585858586e-06,
+      "loss": 1.4092,
+      "step": 1209
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 7.656565656565658e-06,
+      "loss": 2.0569,
+      "step": 1210
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 7.654545454545456e-06,
+      "loss": 1.1731,
+      "step": 1211
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 7.652525252525253e-06,
+      "loss": 1.113,
+      "step": 1212
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 7.650505050505052e-06,
+      "loss": 2.3197,
+      "step": 1213
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 7.648484848484849e-06,
+      "loss": 1.9256,
+      "step": 1214
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 7.646464646464648e-06,
+      "loss": 1.4803,
+      "step": 1215
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 7.644444444444445e-06,
+      "loss": 1.458,
+      "step": 1216
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 7.642424242424244e-06,
+      "loss": 1.1163,
+      "step": 1217
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 7.64040404040404e-06,
+      "loss": 1.2428,
+      "step": 1218
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 7.63838383838384e-06,
+      "loss": 1.523,
+      "step": 1219
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 7.636363636363638e-06,
+      "loss": 1.5386,
+      "step": 1220
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 7.634343434343435e-06,
+      "loss": 1.4018,
+      "step": 1221
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 7.632323232323234e-06,
+      "loss": 2.1092,
+      "step": 1222
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 7.630303030303031e-06,
+      "loss": 1.3261,
+      "step": 1223
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 7.628282828282829e-06,
+      "loss": 1.1062,
+      "step": 1224
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 7.6262626262626275e-06,
+      "loss": 1.6079,
+      "step": 1225
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 7.6242424242424254e-06,
+      "loss": 1.7138,
+      "step": 1226
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 7.622222222222223e-06,
+      "loss": 1.0915,
+      "step": 1227
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 7.620202020202021e-06,
+      "loss": 1.1833,
+      "step": 1228
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 7.618181818181819e-06,
+      "loss": 1.8068,
+      "step": 1229
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 7.616161616161617e-06,
+      "loss": 1.5619,
+      "step": 1230
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 7.614141414141415e-06,
+      "loss": 1.5228,
+      "step": 1231
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 7.612121212121213e-06,
+      "loss": 1.3119,
+      "step": 1232
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 7.610101010101011e-06,
+      "loss": 1.2006,
+      "step": 1233
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 7.6080808080808085e-06,
+      "loss": 1.5752,
+      "step": 1234
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 7.606060606060606e-06,
+      "loss": 1.598,
+      "step": 1235
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 7.604040404040405e-06,
+      "loss": 1.4521,
+      "step": 1236
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 7.602020202020203e-06,
+      "loss": 1.2625,
+      "step": 1237
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 7.600000000000001e-06,
+      "loss": 1.8217,
+      "step": 1238
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 7.597979797979799e-06,
+      "loss": 1.544,
+      "step": 1239
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 7.595959595959597e-06,
+      "loss": 1.1732,
+      "step": 1240
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 7.593939393939395e-06,
+      "loss": 1.518,
+      "step": 1241
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 7.5919191919191925e-06,
+      "loss": 1.6382,
+      "step": 1242
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 7.58989898989899e-06,
+      "loss": 1.52,
+      "step": 1243
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 7.587878787878788e-06,
+      "loss": 1.2822,
+      "step": 1244
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 7.585858585858586e-06,
+      "loss": 1.9696,
+      "step": 1245
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 7.583838383838384e-06,
+      "loss": 1.7836,
+      "step": 1246
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 7.581818181818183e-06,
+      "loss": 1.3844,
+      "step": 1247
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 7.579797979797981e-06,
+      "loss": 1.0223,
+      "step": 1248
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 7.5777777777777785e-06,
+      "loss": 1.8815,
+      "step": 1249
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 7.5757575757575764e-06,
+      "loss": 1.7987,
+      "step": 1250
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 7.573737373737374e-06,
+      "loss": 2.1206,
+      "step": 1251
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 7.571717171717172e-06,
+      "loss": 1.4504,
+      "step": 1252
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 7.56969696969697e-06,
+      "loss": 1.513,
+      "step": 1253
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 7.567676767676768e-06,
+      "loss": 1.3635,
+      "step": 1254
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 7.565656565656566e-06,
+      "loss": 1.6046,
+      "step": 1255
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 7.563636363636364e-06,
+      "loss": 1.1637,
+      "step": 1256
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 7.5616161616161625e-06,
+      "loss": 1.5852,
+      "step": 1257
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 7.55959595959596e-06,
+      "loss": 1.3565,
+      "step": 1258
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 7.557575757575758e-06,
+      "loss": 1.772,
+      "step": 1259
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 7.555555555555556e-06,
+      "loss": 1.3623,
+      "step": 1260
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 7.553535353535354e-06,
+      "loss": 1.3771,
+      "step": 1261
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 7.551515151515152e-06,
+      "loss": 2.062,
+      "step": 1262
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 7.54949494949495e-06,
+      "loss": 1.3749,
+      "step": 1263
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 7.547474747474748e-06,
+      "loss": 1.5035,
+      "step": 1264
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 7.545454545454546e-06,
+      "loss": 1.32,
+      "step": 1265
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 7.5434343434343435e-06,
+      "loss": 1.3155,
+      "step": 1266
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 7.541414141414141e-06,
+      "loss": 1.6408,
+      "step": 1267
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 7.53939393939394e-06,
+      "loss": 1.7592,
+      "step": 1268
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 7.537373737373738e-06,
+      "loss": 1.9334,
+      "step": 1269
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 7.535353535353536e-06,
+      "loss": 0.977,
+      "step": 1270
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 7.533333333333334e-06,
+      "loss": 1.2041,
+      "step": 1271
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 7.531313131313132e-06,
+      "loss": 1.2428,
+      "step": 1272
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 7.5292929292929295e-06,
+      "loss": 0.9574,
+      "step": 1273
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 7.5272727272727274e-06,
+      "loss": 1.345,
+      "step": 1274
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 7.525252525252525e-06,
+      "loss": 1.1972,
+      "step": 1275
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 7.523232323232323e-06,
+      "loss": 0.9055,
+      "step": 1276
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 7.521212121212121e-06,
+      "loss": 1.1086,
+      "step": 1277
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 7.519191919191919e-06,
+      "loss": 1.7166,
+      "step": 1278
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 7.517171717171718e-06,
+      "loss": 1.3403,
+      "step": 1279
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 7.515151515151516e-06,
+      "loss": 2.1793,
+      "step": 1280
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 7.5131313131313135e-06,
+      "loss": 1.5098,
+      "step": 1281
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 7.511111111111111e-06,
+      "loss": 0.9458,
+      "step": 1282
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 7.509090909090909e-06,
+      "loss": 1.4548,
+      "step": 1283
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 7.507070707070707e-06,
+      "loss": 1.3854,
+      "step": 1284
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 7.505050505050505e-06,
+      "loss": 1.6548,
+      "step": 1285
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 7.503030303030303e-06,
+      "loss": 1.2717,
+      "step": 1286
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 7.501010101010101e-06,
+      "loss": 1.1454,
+      "step": 1287
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 7.4989898989899e-06,
+      "loss": 1.2925,
+      "step": 1288
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 7.496969696969698e-06,
+      "loss": 1.5973,
+      "step": 1289
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 7.494949494949496e-06,
+      "loss": 1.7364,
+      "step": 1290
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 7.492929292929294e-06,
+      "loss": 2.1067,
+      "step": 1291
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 7.490909090909092e-06,
+      "loss": 0.9372,
+      "step": 1292
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 7.48888888888889e-06,
+      "loss": 1.8539,
+      "step": 1293
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 7.486868686868688e-06,
+      "loss": 1.0118,
+      "step": 1294
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 7.484848484848486e-06,
+      "loss": 1.6056,
+      "step": 1295
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 7.4828282828282835e-06,
+      "loss": 1.1286,
+      "step": 1296
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 7.480808080808082e-06,
+      "loss": 1.2568,
+      "step": 1297
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 7.47878787878788e-06,
+      "loss": 1.396,
+      "step": 1298
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 7.476767676767678e-06,
+      "loss": 1.6033,
+      "step": 1299
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 7.474747474747476e-06,
+      "loss": 1.5758,
+      "step": 1300
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 7.472727272727274e-06,
+      "loss": 1.3823,
+      "step": 1301
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 7.470707070707072e-06,
+      "loss": 2.2595,
+      "step": 1302
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 7.4686868686868696e-06,
+      "loss": 1.7719,
+      "step": 1303
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 7.4666666666666675e-06,
+      "loss": 2.3978,
+      "step": 1304
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 7.464646464646465e-06,
+      "loss": 1.5772,
+      "step": 1305
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 7.462626262626263e-06,
+      "loss": 1.4508,
+      "step": 1306
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 7.460606060606061e-06,
+      "loss": 1.8534,
+      "step": 1307
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 7.45858585858586e-06,
+      "loss": 1.537,
+      "step": 1308
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 7.456565656565658e-06,
+      "loss": 1.4688,
+      "step": 1309
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 7.454545454545456e-06,
+      "loss": 1.3472,
+      "step": 1310
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 7.4525252525252535e-06,
+      "loss": 2.0754,
+      "step": 1311
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 7.450505050505051e-06,
+      "loss": 1.6746,
+      "step": 1312
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 7.448484848484849e-06,
+      "loss": 1.5522,
+      "step": 1313
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 7.446464646464647e-06,
+      "loss": 1.6491,
+      "step": 1314
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 7.444444444444445e-06,
+      "loss": 1.0223,
+      "step": 1315
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 7.442424242424243e-06,
+      "loss": 1.0353,
+      "step": 1316
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 7.440404040404041e-06,
+      "loss": 1.2082,
+      "step": 1317
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 7.438383838383839e-06,
+      "loss": 1.5613,
+      "step": 1318
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 7.4363636363636375e-06,
+      "loss": 1.6762,
+      "step": 1319
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 7.434343434343435e-06,
+      "loss": 1.4186,
+      "step": 1320
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 7.432323232323233e-06,
+      "loss": 1.2109,
+      "step": 1321
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 7.430303030303031e-06,
+      "loss": 1.7045,
+      "step": 1322
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 7.428282828282829e-06,
+      "loss": 1.0787,
+      "step": 1323
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 7.426262626262627e-06,
+      "loss": 1.4291,
+      "step": 1324
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 7.424242424242425e-06,
+      "loss": 1.5357,
+      "step": 1325
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 7.422222222222223e-06,
+      "loss": 1.2109,
+      "step": 1326
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 7.4202020202020206e-06,
+      "loss": 1.074,
+      "step": 1327
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 7.4181818181818185e-06,
+      "loss": 1.1418,
+      "step": 1328
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 7.416161616161617e-06,
+      "loss": 1.3834,
+      "step": 1329
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 7.414141414141415e-06,
+      "loss": 1.2168,
+      "step": 1330
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 7.412121212121213e-06,
+      "loss": 1.26,
+      "step": 1331
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 7.410101010101011e-06,
+      "loss": 1.5745,
+      "step": 1332
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 7.408080808080809e-06,
+      "loss": 1.33,
+      "step": 1333
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 7.406060606060607e-06,
+      "loss": 1.4802,
+      "step": 1334
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 7.4040404040404045e-06,
+      "loss": 1.5983,
+      "step": 1335
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 7.402020202020202e-06,
+      "loss": 1.646,
+      "step": 1336
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 7.4e-06,
+      "loss": 1.2289,
+      "step": 1337
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 7.397979797979798e-06,
+      "loss": 1.1523,
+      "step": 1338
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 7.395959595959596e-06,
+      "loss": 1.5441,
+      "step": 1339
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 7.393939393939395e-06,
+      "loss": 1.4787,
+      "step": 1340
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 7.391919191919193e-06,
+      "loss": 1.7624,
+      "step": 1341
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 7.389898989898991e-06,
+      "loss": 1.9072,
+      "step": 1342
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 7.3878787878787885e-06,
+      "loss": 1.1781,
+      "step": 1343
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 7.385858585858586e-06,
+      "loss": 1.0268,
+      "step": 1344
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 7.383838383838384e-06,
+      "loss": 1.3713,
+      "step": 1345
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 7.381818181818182e-06,
+      "loss": 1.238,
+      "step": 1346
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 7.37979797979798e-06,
+      "loss": 0.9514,
+      "step": 1347
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 7.377777777777778e-06,
+      "loss": 1.9475,
+      "step": 1348
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 7.375757575757576e-06,
+      "loss": 1.0095,
+      "step": 1349
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 7.373737373737374e-06,
+      "loss": 1.5383,
+      "step": 1350
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 7.3717171717171724e-06,
+      "loss": 1.3075,
+      "step": 1351
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 7.36969696969697e-06,
+      "loss": 0.5769,
+      "step": 1352
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 7.367676767676768e-06,
+      "loss": 1.3956,
+      "step": 1353
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 7.365656565656566e-06,
+      "loss": 1.3154,
+      "step": 1354
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 7.363636363636364e-06,
+      "loss": 1.7657,
+      "step": 1355
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 7.361616161616162e-06,
+      "loss": 1.2176,
+      "step": 1356
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 7.35959595959596e-06,
+      "loss": 1.4041,
+      "step": 1357
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 7.357575757575758e-06,
+      "loss": 1.2264,
+      "step": 1358
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 7.3555555555555555e-06,
+      "loss": 0.9306,
+      "step": 1359
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 7.353535353535353e-06,
+      "loss": 1.5451,
+      "step": 1360
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 7.351515151515151e-06,
+      "loss": 1.5243,
+      "step": 1361
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 7.34949494949495e-06,
+      "loss": 1.6033,
+      "step": 1362
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 7.347474747474748e-06,
+      "loss": 1.3845,
+      "step": 1363
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 7.345454545454546e-06,
+      "loss": 1.2536,
+      "step": 1364
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 7.343434343434344e-06,
+      "loss": 1.0877,
+      "step": 1365
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 7.341414141414142e-06,
+      "loss": 1.9936,
+      "step": 1366
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 7.3393939393939395e-06,
+      "loss": 1.6685,
+      "step": 1367
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 7.337373737373737e-06,
+      "loss": 1.4106,
+      "step": 1368
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 7.335353535353535e-06,
+      "loss": 1.866,
+      "step": 1369
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 7.333333333333333e-06,
+      "loss": 1.1782,
+      "step": 1370
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 7.331313131313131e-06,
+      "loss": 1.751,
+      "step": 1371
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 7.32929292929293e-06,
+      "loss": 1.6422,
+      "step": 1372
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 7.3272727272727285e-06,
+      "loss": 1.2792,
+      "step": 1373
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 7.325252525252526e-06,
+      "loss": 1.6476,
+      "step": 1374
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 7.323232323232324e-06,
+      "loss": 1.1966,
+      "step": 1375
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 7.321212121212122e-06,
+      "loss": 1.1125,
+      "step": 1376
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 7.31919191919192e-06,
+      "loss": 1.3246,
+      "step": 1377
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 7.317171717171718e-06,
+      "loss": 0.9332,
+      "step": 1378
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 7.315151515151516e-06,
+      "loss": 1.1779,
+      "step": 1379
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 7.3131313131313146e-06,
+      "loss": 1.5138,
+      "step": 1380
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 7.3111111111111125e-06,
+      "loss": 1.2724,
+      "step": 1381
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 7.30909090909091e-06,
+      "loss": 2.2569,
+      "step": 1382
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 7.307070707070708e-06,
+      "loss": 1.113,
+      "step": 1383
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 7.305050505050506e-06,
+      "loss": 1.5462,
+      "step": 1384
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 7.303030303030304e-06,
+      "loss": 1.1848,
+      "step": 1385
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 7.301010101010102e-06,
+      "loss": 1.5629,
+      "step": 1386
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 7.2989898989899e-06,
+      "loss": 1.4222,
+      "step": 1387
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 7.296969696969698e-06,
+      "loss": 1.6976,
+      "step": 1388
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 7.2949494949494956e-06,
+      "loss": 1.301,
+      "step": 1389
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 7.2929292929292934e-06,
+      "loss": 1.5467,
+      "step": 1390
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 7.290909090909092e-06,
+      "loss": 1.5533,
+      "step": 1391
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 7.28888888888889e-06,
+      "loss": 1.1694,
+      "step": 1392
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 7.286868686868688e-06,
+      "loss": 1.5809,
+      "step": 1393
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 7.284848484848486e-06,
+      "loss": 1.5647,
+      "step": 1394
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 7.282828282828284e-06,
+      "loss": 1.5498,
+      "step": 1395
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 7.280808080808082e-06,
+      "loss": 1.2162,
+      "step": 1396
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 7.2787878787878795e-06,
+      "loss": 1.0052,
+      "step": 1397
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 7.276767676767677e-06,
+      "loss": 1.4522,
+      "step": 1398
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 7.274747474747475e-06,
+      "loss": 1.6432,
+      "step": 1399
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 1.2732,
+      "step": 1400
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.665,
+      "eval_loss": 1.282926321029663,
+      "eval_roc_auc": 0.9384211055770695,
+      "eval_runtime": 92.9585,
+      "eval_samples_per_second": 2.151,
+      "eval_steps_per_second": 0.538,
+      "step": 1400
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 7.270707070707072e-06,
+      "loss": 1.1751,
+      "step": 1401
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 7.26868686868687e-06,
+      "loss": 1.1887,
+      "step": 1402
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 7.266666666666668e-06,
+      "loss": 1.0884,
+      "step": 1403
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 7.2646464646464656e-06,
+      "loss": 1.3561,
+      "step": 1404
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 7.2626262626262635e-06,
+      "loss": 1.799,
+      "step": 1405
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 7.260606060606061e-06,
+      "loss": 1.3387,
+      "step": 1406
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 7.258585858585859e-06,
+      "loss": 1.2252,
+      "step": 1407
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 7.256565656565657e-06,
+      "loss": 1.3716,
+      "step": 1408
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 7.254545454545455e-06,
+      "loss": 1.2939,
+      "step": 1409
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 7.252525252525253e-06,
+      "loss": 1.3663,
+      "step": 1410
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 7.250505050505051e-06,
+      "loss": 1.2638,
+      "step": 1411
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 7.2484848484848495e-06,
+      "loss": 1.2325,
+      "step": 1412
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 7.246464646464647e-06,
+      "loss": 1.5204,
+      "step": 1413
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 7.244444444444445e-06,
+      "loss": 1.9836,
+      "step": 1414
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 7.242424242424243e-06,
+      "loss": 1.2638,
+      "step": 1415
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 7.240404040404041e-06,
+      "loss": 1.4015,
+      "step": 1416
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 7.238383838383839e-06,
+      "loss": 1.3216,
+      "step": 1417
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 7.236363636363637e-06,
+      "loss": 1.4778,
+      "step": 1418
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 7.234343434343435e-06,
+      "loss": 0.6644,
+      "step": 1419
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 7.232323232323233e-06,
+      "loss": 0.8258,
+      "step": 1420
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 7.2303030303030305e-06,
+      "loss": 1.5164,
+      "step": 1421
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 7.228282828282828e-06,
+      "loss": 1.3517,
+      "step": 1422
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 7.226262626262627e-06,
+      "loss": 2.056,
+      "step": 1423
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 7.224242424242425e-06,
+      "loss": 1.3183,
+      "step": 1424
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 7.222222222222223e-06,
+      "loss": 1.1009,
+      "step": 1425
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 7.220202020202021e-06,
+      "loss": 1.1911,
+      "step": 1426
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 7.218181818181819e-06,
+      "loss": 0.9987,
+      "step": 1427
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 7.2161616161616166e-06,
+      "loss": 1.4897,
+      "step": 1428
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 7.2141414141414145e-06,
+      "loss": 1.4224,
+      "step": 1429
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 7.212121212121212e-06,
+      "loss": 1.3134,
+      "step": 1430
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 7.21010101010101e-06,
+      "loss": 1.4555,
+      "step": 1431
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 7.208080808080808e-06,
+      "loss": 2.0859,
+      "step": 1432
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 7.206060606060606e-06,
+      "loss": 1.4224,
+      "step": 1433
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 7.204040404040405e-06,
+      "loss": 1.3063,
+      "step": 1434
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 7.202020202020203e-06,
+      "loss": 1.4981,
+      "step": 1435
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 7.2000000000000005e-06,
+      "loss": 1.4924,
+      "step": 1436
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 7.197979797979798e-06,
+      "loss": 2.0093,
+      "step": 1437
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 7.195959595959596e-06,
+      "loss": 1.2063,
+      "step": 1438
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 7.193939393939394e-06,
+      "loss": 1.5618,
+      "step": 1439
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 7.191919191919192e-06,
+      "loss": 1.2191,
+      "step": 1440
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 7.18989898989899e-06,
+      "loss": 1.6965,
+      "step": 1441
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 7.187878787878788e-06,
+      "loss": 1.7918,
+      "step": 1442
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 7.185858585858586e-06,
+      "loss": 1.7872,
+      "step": 1443
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 7.1838383838383845e-06,
+      "loss": 0.9294,
+      "step": 1444
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 7.181818181818182e-06,
+      "loss": 1.7607,
+      "step": 1445
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 7.17979797979798e-06,
+      "loss": 1.5442,
+      "step": 1446
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 7.177777777777778e-06,
+      "loss": 1.2559,
+      "step": 1447
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 7.175757575757576e-06,
+      "loss": 1.2101,
+      "step": 1448
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 7.173737373737374e-06,
+      "loss": 2.5675,
+      "step": 1449
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 7.171717171717172e-06,
+      "loss": 1.7354,
+      "step": 1450
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 7.16969696969697e-06,
+      "loss": 1.6346,
+      "step": 1451
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 7.1676767676767676e-06,
+      "loss": 1.4588,
+      "step": 1452
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 7.1656565656565655e-06,
+      "loss": 1.2224,
+      "step": 1453
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 7.163636363636363e-06,
+      "loss": 1.1647,
+      "step": 1454
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 7.161616161616162e-06,
+      "loss": 1.3028,
+      "step": 1455
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 7.15959595959596e-06,
+      "loss": 1.2868,
+      "step": 1456
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 7.157575757575758e-06,
+      "loss": 1.0589,
+      "step": 1457
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 7.155555555555556e-06,
+      "loss": 1.5785,
+      "step": 1458
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 7.1535353535353545e-06,
+      "loss": 1.4672,
+      "step": 1459
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 7.151515151515152e-06,
+      "loss": 1.0977,
+      "step": 1460
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 7.14949494949495e-06,
+      "loss": 2.0653,
+      "step": 1461
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 7.147474747474748e-06,
+      "loss": 1.5091,
+      "step": 1462
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 7.145454545454547e-06,
+      "loss": 1.5985,
+      "step": 1463
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 7.143434343434345e-06,
+      "loss": 1.1101,
+      "step": 1464
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 7.141414141414143e-06,
+      "loss": 1.1279,
+      "step": 1465
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 7.1393939393939405e-06,
+      "loss": 1.2193,
+      "step": 1466
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 7.1373737373737384e-06,
+      "loss": 1.3293,
+      "step": 1467
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.135353535353536e-06,
+      "loss": 1.1414,
+      "step": 1468
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.133333333333334e-06,
+      "loss": 1.2182,
+      "step": 1469
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.131313131313132e-06,
+      "loss": 1.7236,
+      "step": 1470
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.12929292929293e-06,
+      "loss": 1.2492,
+      "step": 1471
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.127272727272728e-06,
+      "loss": 1.0685,
+      "step": 1472
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.125252525252527e-06,
+      "loss": 1.4576,
+      "step": 1473
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.1232323232323245e-06,
+      "loss": 1.292,
+      "step": 1474
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.121212121212122e-06,
+      "loss": 0.8224,
+      "step": 1475
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.11919191919192e-06,
+      "loss": 1.6779,
+      "step": 1476
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.117171717171718e-06,
+      "loss": 1.288,
+      "step": 1477
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.115151515151516e-06,
+      "loss": 1.5006,
+      "step": 1478
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.113131313131314e-06,
+      "loss": 1.6287,
+      "step": 1479
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.111111111111112e-06,
+      "loss": 1.2954,
+      "step": 1480
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.10909090909091e-06,
+      "loss": 1.3753,
+      "step": 1481
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.107070707070708e-06,
+      "loss": 1.3002,
+      "step": 1482
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.1050505050505055e-06,
+      "loss": 1.8752,
+      "step": 1483
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.103030303030304e-06,
+      "loss": 1.6734,
+      "step": 1484
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.101010101010102e-06,
+      "loss": 1.6016,
+      "step": 1485
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.0989898989899e-06,
+      "loss": 1.0589,
+      "step": 1486
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.096969696969698e-06,
+      "loss": 0.8662,
+      "step": 1487
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.094949494949496e-06,
+      "loss": 1.4192,
+      "step": 1488
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.092929292929294e-06,
+      "loss": 1.3288,
+      "step": 1489
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.0909090909090916e-06,
+      "loss": 2.0141,
+      "step": 1490
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.0888888888888894e-06,
+      "loss": 0.989,
+      "step": 1491
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.086868686868687e-06,
+      "loss": 2.0162,
+      "step": 1492
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.084848484848485e-06,
+      "loss": 1.4205,
+      "step": 1493
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.082828282828283e-06,
+      "loss": 1.7305,
+      "step": 1494
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.080808080808082e-06,
+      "loss": 1.3829,
+      "step": 1495
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.07878787878788e-06,
+      "loss": 1.3108,
+      "step": 1496
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.076767676767678e-06,
+      "loss": 1.3915,
+      "step": 1497
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.0747474747474755e-06,
+      "loss": 1.4199,
+      "step": 1498
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.072727272727273e-06,
+      "loss": 1.3824,
+      "step": 1499
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.070707070707071e-06,
+      "loss": 1.1658,
+      "step": 1500
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.068686868686869e-06,
+      "loss": 1.324,
+      "step": 1501
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.066666666666667e-06,
+      "loss": 1.3482,
+      "step": 1502
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.064646464646465e-06,
+      "loss": 0.8384,
+      "step": 1503
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.062626262626263e-06,
+      "loss": 1.1765,
+      "step": 1504
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.060606060606061e-06,
+      "loss": 1.7822,
+      "step": 1505
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.0585858585858595e-06,
+      "loss": 1.6511,
+      "step": 1506
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.056565656565657e-06,
+      "loss": 0.8849,
+      "step": 1507
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.054545454545455e-06,
+      "loss": 1.5862,
+      "step": 1508
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.052525252525253e-06,
+      "loss": 1.7671,
+      "step": 1509
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.050505050505051e-06,
+      "loss": 1.9733,
+      "step": 1510
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.048484848484849e-06,
+      "loss": 1.1513,
+      "step": 1511
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.046464646464647e-06,
+      "loss": 0.9944,
+      "step": 1512
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.044444444444445e-06,
+      "loss": 1.1296,
+      "step": 1513
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.0424242424242426e-06,
+      "loss": 1.0453,
+      "step": 1514
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.0404040404040404e-06,
+      "loss": 1.2724,
+      "step": 1515
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.038383838383839e-06,
+      "loss": 0.7716,
+      "step": 1516
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.036363636363637e-06,
+      "loss": 1.468,
+      "step": 1517
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.034343434343435e-06,
+      "loss": 1.4605,
+      "step": 1518
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.032323232323233e-06,
+      "loss": 0.8205,
+      "step": 1519
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.030303030303031e-06,
+      "loss": 1.59,
+      "step": 1520
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.028282828282829e-06,
+      "loss": 2.3756,
+      "step": 1521
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.0262626262626265e-06,
+      "loss": 1.4783,
+      "step": 1522
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.024242424242424e-06,
+      "loss": 1.7351,
+      "step": 1523
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.022222222222222e-06,
+      "loss": 1.31,
+      "step": 1524
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.02020202020202e-06,
+      "loss": 1.1801,
+      "step": 1525
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.018181818181818e-06,
+      "loss": 1.08,
+      "step": 1526
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.016161616161617e-06,
+      "loss": 1.8928,
+      "step": 1527
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.014141414141415e-06,
+      "loss": 1.3002,
+      "step": 1528
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.0121212121212126e-06,
+      "loss": 1.3222,
+      "step": 1529
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.0101010101010105e-06,
+      "loss": 1.4274,
+      "step": 1530
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.008080808080808e-06,
+      "loss": 1.6285,
+      "step": 1531
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.006060606060606e-06,
+      "loss": 1.6379,
+      "step": 1532
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.004040404040404e-06,
+      "loss": 1.9267,
+      "step": 1533
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.002020202020202e-06,
+      "loss": 1.3668,
+      "step": 1534
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7e-06,
+      "loss": 1.4606,
+      "step": 1535
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.997979797979798e-06,
+      "loss": 1.2523,
+      "step": 1536
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.995959595959596e-06,
+      "loss": 2.0036,
+      "step": 1537
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.993939393939394e-06,
+      "loss": 1.8413,
+      "step": 1538
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.991919191919192e-06,
+      "loss": 1.8602,
+      "step": 1539
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.98989898989899e-06,
+      "loss": 1.8174,
+      "step": 1540
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.987878787878788e-06,
+      "loss": 1.5288,
+      "step": 1541
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.985858585858586e-06,
+      "loss": 1.3808,
+      "step": 1542
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.983838383838384e-06,
+      "loss": 1.4514,
+      "step": 1543
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.981818181818183e-06,
+      "loss": 1.442,
+      "step": 1544
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.979797979797981e-06,
+      "loss": 0.8834,
+      "step": 1545
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.977777777777779e-06,
+      "loss": 1.0988,
+      "step": 1546
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.975757575757577e-06,
+      "loss": 0.956,
+      "step": 1547
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.973737373737375e-06,
+      "loss": 1.8031,
+      "step": 1548
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.971717171717173e-06,
+      "loss": 1.2739,
+      "step": 1549
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.969696969696971e-06,
+      "loss": 1.3796,
+      "step": 1550
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.967676767676769e-06,
+      "loss": 0.9096,
+      "step": 1551
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.9656565656565665e-06,
+      "loss": 0.9328,
+      "step": 1552
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.963636363636364e-06,
+      "loss": 1.1742,
+      "step": 1553
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.961616161616162e-06,
+      "loss": 1.4301,
+      "step": 1554
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.95959595959596e-06,
+      "loss": 1.6312,
+      "step": 1555
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.957575757575759e-06,
+      "loss": 1.5976,
+      "step": 1556
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.955555555555557e-06,
+      "loss": 1.3034,
+      "step": 1557
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.953535353535355e-06,
+      "loss": 1.0983,
+      "step": 1558
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.951515151515153e-06,
+      "loss": 0.9962,
+      "step": 1559
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.9494949494949505e-06,
+      "loss": 1.224,
+      "step": 1560
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.947474747474748e-06,
+      "loss": 1.6469,
+      "step": 1561
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.945454545454546e-06,
+      "loss": 0.8921,
+      "step": 1562
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.943434343434344e-06,
+      "loss": 1.0275,
+      "step": 1563
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.941414141414142e-06,
+      "loss": 1.6263,
+      "step": 1564
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.93939393939394e-06,
+      "loss": 1.3184,
+      "step": 1565
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.937373737373738e-06,
+      "loss": 1.7499,
+      "step": 1566
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.9353535353535365e-06,
+      "loss": 1.8123,
+      "step": 1567
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.9333333333333344e-06,
+      "loss": 1.2929,
+      "step": 1568
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.931313131313132e-06,
+      "loss": 1.4878,
+      "step": 1569
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.92929292929293e-06,
+      "loss": 2.4559,
+      "step": 1570
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.927272727272728e-06,
+      "loss": 0.8237,
+      "step": 1571
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.925252525252526e-06,
+      "loss": 2.0085,
+      "step": 1572
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.923232323232324e-06,
+      "loss": 1.2327,
+      "step": 1573
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.921212121212122e-06,
+      "loss": 1.0233,
+      "step": 1574
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.91919191919192e-06,
+      "loss": 1.305,
+      "step": 1575
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.9171717171717175e-06,
+      "loss": 0.9139,
+      "step": 1576
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.915151515151515e-06,
+      "loss": 0.9938,
+      "step": 1577
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.913131313131314e-06,
+      "loss": 1.0041,
+      "step": 1578
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.911111111111112e-06,
+      "loss": 0.9896,
+      "step": 1579
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.90909090909091e-06,
+      "loss": 1.4315,
+      "step": 1580
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.907070707070708e-06,
+      "loss": 1.619,
+      "step": 1581
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.905050505050506e-06,
+      "loss": 1.3026,
+      "step": 1582
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.903030303030304e-06,
+      "loss": 1.4077,
+      "step": 1583
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.9010101010101015e-06,
+      "loss": 1.5026,
+      "step": 1584
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.898989898989899e-06,
+      "loss": 0.6397,
+      "step": 1585
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.896969696969697e-06,
+      "loss": 1.0362,
+      "step": 1586
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.894949494949495e-06,
+      "loss": 1.2736,
+      "step": 1587
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.892929292929294e-06,
+      "loss": 1.1822,
+      "step": 1588
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.890909090909092e-06,
+      "loss": 1.0438,
+      "step": 1589
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.88888888888889e-06,
+      "loss": 1.3373,
+      "step": 1590
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.8868686868686875e-06,
+      "loss": 1.6304,
+      "step": 1591
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.8848484848484854e-06,
+      "loss": 1.6677,
+      "step": 1592
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.882828282828283e-06,
+      "loss": 1.3821,
+      "step": 1593
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.880808080808081e-06,
+      "loss": 1.6867,
+      "step": 1594
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.878787878787879e-06,
+      "loss": 1.0938,
+      "step": 1595
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.876767676767677e-06,
+      "loss": 0.8114,
+      "step": 1596
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.874747474747475e-06,
+      "loss": 1.2522,
+      "step": 1597
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.872727272727273e-06,
+      "loss": 1.605,
+      "step": 1598
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.8707070707070715e-06,
+      "loss": 1.2873,
+      "step": 1599
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.868686868686869e-06,
+      "loss": 0.9384,
+      "step": 1600
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.585,
+      "eval_loss": 1.246170997619629,
+      "eval_roc_auc": 0.931168413307331,
+      "eval_runtime": 92.402,
+      "eval_samples_per_second": 2.164,
+      "eval_steps_per_second": 0.541,
+      "step": 1600
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 6.866666666666667e-06,
+      "loss": 1.1077,
+      "step": 1601
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 6.864646464646465e-06,
+      "loss": 1.2072,
+      "step": 1602
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 6.862626262626263e-06,
+      "loss": 1.1132,
+      "step": 1603
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 6.860606060606061e-06,
+      "loss": 1.5235,
+      "step": 1604
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 6.858585858585859e-06,
+      "loss": 1.1861,
+      "step": 1605
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 6.856565656565657e-06,
+      "loss": 1.371,
+      "step": 1606
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 6.854545454545455e-06,
+      "loss": 0.9065,
+      "step": 1607
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 6.8525252525252525e-06,
+      "loss": 1.5701,
+      "step": 1608
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 6.85050505050505e-06,
+      "loss": 1.1888,
+      "step": 1609
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 6.848484848484849e-06,
+      "loss": 1.1125,
+      "step": 1610
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 6.846464646464647e-06,
+      "loss": 1.3599,
+      "step": 1611
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 6.844444444444445e-06,
+      "loss": 1.3324,
+      "step": 1612
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 6.842424242424243e-06,
+      "loss": 1.742,
+      "step": 1613
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 6.840404040404041e-06,
+      "loss": 1.3685,
+      "step": 1614
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 6.8383838383838386e-06,
+      "loss": 1.0725,
+      "step": 1615
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 6.8363636363636364e-06,
+      "loss": 0.7376,
+      "step": 1616
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 6.834343434343434e-06,
+      "loss": 1.251,
+      "step": 1617
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 6.832323232323232e-06,
+      "loss": 1.4218,
+      "step": 1618
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 6.83030303030303e-06,
+      "loss": 1.375,
+      "step": 1619
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 6.828282828282828e-06,
+      "loss": 1.4484,
+      "step": 1620
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 6.826262626262627e-06,
+      "loss": 0.8822,
+      "step": 1621
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 6.824242424242425e-06,
+      "loss": 1.4168,
+      "step": 1622
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 6.8222222222222225e-06,
+      "loss": 0.7218,
+      "step": 1623
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 6.82020202020202e-06,
+      "loss": 1.265,
+      "step": 1624
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 6.818181818181818e-06,
+      "loss": 0.7462,
+      "step": 1625
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 6.816161616161616e-06,
+      "loss": 1.6033,
+      "step": 1626
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 6.814141414141414e-06,
+      "loss": 1.236,
+      "step": 1627
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 6.812121212121212e-06,
+      "loss": 1.0812,
+      "step": 1628
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 6.8101010101010115e-06,
+      "loss": 0.9514,
+      "step": 1629
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 6.808080808080809e-06,
+      "loss": 1.8934,
+      "step": 1630
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 6.806060606060607e-06,
+      "loss": 1.3443,
+      "step": 1631
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 6.804040404040405e-06,
+      "loss": 1.5116,
+      "step": 1632
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 6.802020202020203e-06,
+      "loss": 1.7176,
+      "step": 1633
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 6.800000000000001e-06,
+      "loss": 0.7609,
+      "step": 1634
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 6.797979797979799e-06,
+      "loss": 1.0045,
+      "step": 1635
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 6.795959595959597e-06,
+      "loss": 1.4546,
+      "step": 1636
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 6.793939393939395e-06,
+      "loss": 1.4116,
+      "step": 1637
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 6.7919191919191925e-06,
+      "loss": 0.564,
+      "step": 1638
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 6.789898989898991e-06,
+      "loss": 1.6268,
+      "step": 1639
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 6.787878787878789e-06,
+      "loss": 0.8816,
+      "step": 1640
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 6.785858585858587e-06,
+      "loss": 1.2385,
+      "step": 1641
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 6.783838383838385e-06,
+      "loss": 0.9631,
+      "step": 1642
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 6.781818181818183e-06,
+      "loss": 0.8525,
+      "step": 1643
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 6.779797979797981e-06,
+      "loss": 0.9875,
+      "step": 1644
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 6.777777777777779e-06,
+      "loss": 1.2599,
+      "step": 1645
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 6.7757575757575765e-06,
+      "loss": 2.1503,
+      "step": 1646
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 6.773737373737374e-06,
+      "loss": 1.3779,
+      "step": 1647
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 6.771717171717172e-06,
+      "loss": 1.2723,
+      "step": 1648
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 6.76969696969697e-06,
+      "loss": 1.0852,
+      "step": 1649
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 6.767676767676769e-06,
+      "loss": 1.114,
+      "step": 1650
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 6.765656565656567e-06,
+      "loss": 0.8786,
+      "step": 1651
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 6.763636363636365e-06,
+      "loss": 1.122,
+      "step": 1652
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 6.7616161616161625e-06,
+      "loss": 0.7567,
+      "step": 1653
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 6.75959595959596e-06,
+      "loss": 1.1124,
+      "step": 1654
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 6.757575757575758e-06,
+      "loss": 1.4733,
+      "step": 1655
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 6.755555555555556e-06,
+      "loss": 1.6524,
+      "step": 1656
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 6.753535353535354e-06,
+      "loss": 1.1319,
+      "step": 1657
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 6.751515151515152e-06,
+      "loss": 0.9542,
+      "step": 1658
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 6.74949494949495e-06,
+      "loss": 1.2349,
+      "step": 1659
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 6.747474747474749e-06,
+      "loss": 1.6881,
+      "step": 1660
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 6.7454545454545465e-06,
+      "loss": 1.9885,
+      "step": 1661
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 6.743434343434344e-06,
+      "loss": 0.5821,
+      "step": 1662
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 6.741414141414142e-06,
+      "loss": 1.2624,
+      "step": 1663
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 6.73939393939394e-06,
+      "loss": 0.9667,
+      "step": 1664
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 6.737373737373738e-06,
+      "loss": 0.8157,
+      "step": 1665
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 6.735353535353536e-06,
+      "loss": 1.5268,
+      "step": 1666
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 6.733333333333334e-06,
+      "loss": 1.594,
+      "step": 1667
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 6.731313131313132e-06,
+      "loss": 1.4089,
+      "step": 1668
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 6.72929292929293e-06,
+      "loss": 1.6676,
+      "step": 1669
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 6.7272727272727275e-06,
+      "loss": 0.9183,
+      "step": 1670
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 6.725252525252526e-06,
+      "loss": 1.5985,
+      "step": 1671
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 6.723232323232324e-06,
+      "loss": 1.2862,
+      "step": 1672
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 6.721212121212122e-06,
+      "loss": 0.9301,
+      "step": 1673
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 6.71919191919192e-06,
+      "loss": 1.2369,
+      "step": 1674
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 6.717171717171718e-06,
+      "loss": 1.3135,
+      "step": 1675
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 6.715151515151516e-06,
+      "loss": 1.4862,
+      "step": 1676
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 6.7131313131313135e-06,
+      "loss": 1.1906,
+      "step": 1677
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 6.711111111111111e-06,
+      "loss": 1.1198,
+      "step": 1678
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 6.709090909090909e-06,
+      "loss": 1.3242,
+      "step": 1679
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 6.707070707070707e-06,
+      "loss": 1.45,
+      "step": 1680
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 6.705050505050505e-06,
+      "loss": 1.1418,
+      "step": 1681
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 6.703030303030304e-06,
+      "loss": 1.2163,
+      "step": 1682
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 6.701010101010102e-06,
+      "loss": 1.418,
+      "step": 1683
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 6.6989898989899e-06,
+      "loss": 1.5978,
+      "step": 1684
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 6.6969696969696975e-06,
+      "loss": 1.088,
+      "step": 1685
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 6.694949494949495e-06,
+      "loss": 1.2705,
+      "step": 1686
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 6.692929292929293e-06,
+      "loss": 1.3211,
+      "step": 1687
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 6.690909090909091e-06,
+      "loss": 1.7667,
+      "step": 1688
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 6.688888888888889e-06,
+      "loss": 1.1492,
+      "step": 1689
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 6.686868686868687e-06,
+      "loss": 1.3455,
+      "step": 1690
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 6.684848484848485e-06,
+      "loss": 1.4911,
+      "step": 1691
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 6.682828282828283e-06,
+      "loss": 1.7435,
+      "step": 1692
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 6.6808080808080814e-06,
+      "loss": 0.8091,
+      "step": 1693
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 6.678787878787879e-06,
+      "loss": 1.1538,
+      "step": 1694
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 6.676767676767677e-06,
+      "loss": 1.0412,
+      "step": 1695
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 6.674747474747475e-06,
+      "loss": 0.9635,
+      "step": 1696
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 6.672727272727273e-06,
+      "loss": 1.2381,
+      "step": 1697
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 6.670707070707071e-06,
+      "loss": 1.7944,
+      "step": 1698
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 6.668686868686869e-06,
+      "loss": 1.2845,
+      "step": 1699
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.8508,
+      "step": 1700
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 6.6646464646464645e-06,
+      "loss": 1.2024,
+      "step": 1701
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 6.662626262626262e-06,
+      "loss": 1.8188,
+      "step": 1702
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 6.660606060606061e-06,
+      "loss": 1.6477,
+      "step": 1703
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 6.658585858585859e-06,
+      "loss": 1.3396,
+      "step": 1704
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 6.656565656565657e-06,
+      "loss": 0.6748,
+      "step": 1705
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 6.654545454545455e-06,
+      "loss": 1.0297,
+      "step": 1706
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 6.652525252525253e-06,
+      "loss": 0.9621,
+      "step": 1707
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 6.650505050505051e-06,
+      "loss": 1.5749,
+      "step": 1708
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 6.6484848484848485e-06,
+      "loss": 0.9847,
+      "step": 1709
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 6.646464646464646e-06,
+      "loss": 0.8154,
+      "step": 1710
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 6.644444444444444e-06,
+      "loss": 0.9875,
+      "step": 1711
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 6.642424242424242e-06,
+      "loss": 1.2152,
+      "step": 1712
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 6.64040404040404e-06,
+      "loss": 0.9205,
+      "step": 1713
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 6.638383838383839e-06,
+      "loss": 1.6657,
+      "step": 1714
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 6.6363636363636375e-06,
+      "loss": 0.878,
+      "step": 1715
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 6.634343434343435e-06,
+      "loss": 1.254,
+      "step": 1716
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 6.632323232323233e-06,
+      "loss": 1.8,
+      "step": 1717
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 6.630303030303031e-06,
+      "loss": 0.9255,
+      "step": 1718
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 6.628282828282829e-06,
+      "loss": 0.9442,
+      "step": 1719
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 6.626262626262627e-06,
+      "loss": 1.2682,
+      "step": 1720
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 6.624242424242425e-06,
+      "loss": 2.5503,
+      "step": 1721
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 6.6222222222222236e-06,
+      "loss": 1.1709,
+      "step": 1722
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 6.6202020202020215e-06,
+      "loss": 0.8702,
+      "step": 1723
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 6.618181818181819e-06,
+      "loss": 1.4774,
+      "step": 1724
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 6.616161616161617e-06,
+      "loss": 0.8964,
+      "step": 1725
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 6.614141414141415e-06,
+      "loss": 1.0635,
+      "step": 1726
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 6.612121212121213e-06,
+      "loss": 1.4086,
+      "step": 1727
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 6.610101010101011e-06,
+      "loss": 1.1251,
+      "step": 1728
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 6.608080808080809e-06,
+      "loss": 1.4975,
+      "step": 1729
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 6.606060606060607e-06,
+      "loss": 1.0618,
+      "step": 1730
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 6.6040404040404046e-06,
+      "loss": 1.2507,
+      "step": 1731
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 6.602020202020203e-06,
+      "loss": 1.4101,
+      "step": 1732
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 6.600000000000001e-06,
+      "loss": 1.2253,
+      "step": 1733
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 6.597979797979799e-06,
+      "loss": 1.4657,
+      "step": 1734
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 6.595959595959597e-06,
+      "loss": 1.797,
+      "step": 1735
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 6.593939393939395e-06,
+      "loss": 1.6298,
+      "step": 1736
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 6.591919191919193e-06,
+      "loss": 1.2422,
+      "step": 1737
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 6.589898989898991e-06,
+      "loss": 1.2923,
+      "step": 1738
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 6.5878787878787885e-06,
+      "loss": 1.654,
+      "step": 1739
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 6.585858585858586e-06,
+      "loss": 1.485,
+      "step": 1740
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 6.583838383838384e-06,
+      "loss": 1.1117,
+      "step": 1741
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 6.581818181818182e-06,
+      "loss": 1.4387,
+      "step": 1742
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 6.579797979797981e-06,
+      "loss": 0.9215,
+      "step": 1743
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 6.577777777777779e-06,
+      "loss": 1.5164,
+      "step": 1744
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 6.575757575757577e-06,
+      "loss": 1.0548,
+      "step": 1745
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 6.5737373737373746e-06,
+      "loss": 1.377,
+      "step": 1746
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 6.5717171717171725e-06,
+      "loss": 1.2152,
+      "step": 1747
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 6.56969696969697e-06,
+      "loss": 1.5759,
+      "step": 1748
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 6.567676767676768e-06,
+      "loss": 1.1471,
+      "step": 1749
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 6.565656565656566e-06,
+      "loss": 1.4082,
+      "step": 1750
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 6.563636363636364e-06,
+      "loss": 0.547,
+      "step": 1751
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 6.561616161616162e-06,
+      "loss": 0.7779,
+      "step": 1752
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 6.55959595959596e-06,
+      "loss": 1.3438,
+      "step": 1753
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 6.5575757575757585e-06,
+      "loss": 0.7829,
+      "step": 1754
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 6.555555555555556e-06,
+      "loss": 1.3523,
+      "step": 1755
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 6.553535353535354e-06,
+      "loss": 1.1423,
+      "step": 1756
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 6.551515151515152e-06,
+      "loss": 1.136,
+      "step": 1757
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 6.54949494949495e-06,
+      "loss": 1.02,
+      "step": 1758
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 6.547474747474748e-06,
+      "loss": 1.9447,
+      "step": 1759
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 6.545454545454546e-06,
+      "loss": 0.4985,
+      "step": 1760
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 6.543434343434344e-06,
+      "loss": 1.836,
+      "step": 1761
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 6.541414141414142e-06,
+      "loss": 1.8359,
+      "step": 1762
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 6.5393939393939395e-06,
+      "loss": 1.2954,
+      "step": 1763
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 6.537373737373737e-06,
+      "loss": 1.1861,
+      "step": 1764
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 6.535353535353536e-06,
+      "loss": 0.8441,
+      "step": 1765
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 6.533333333333334e-06,
+      "loss": 1.1389,
+      "step": 1766
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 6.531313131313132e-06,
+      "loss": 0.9691,
+      "step": 1767
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 6.52929292929293e-06,
+      "loss": 1.6235,
+      "step": 1768
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 6.527272727272728e-06,
+      "loss": 1.5921,
+      "step": 1769
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 6.525252525252526e-06,
+      "loss": 0.988,
+      "step": 1770
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 6.5232323232323235e-06,
+      "loss": 2.1679,
+      "step": 1771
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 6.521212121212121e-06,
+      "loss": 1.891,
+      "step": 1772
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 6.519191919191919e-06,
+      "loss": 1.3592,
+      "step": 1773
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 6.517171717171717e-06,
+      "loss": 1.0358,
+      "step": 1774
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 6.515151515151516e-06,
+      "loss": 1.0147,
+      "step": 1775
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 6.513131313131314e-06,
+      "loss": 1.5351,
+      "step": 1776
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 6.511111111111112e-06,
+      "loss": 1.2916,
+      "step": 1777
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 6.5090909090909095e-06,
+      "loss": 1.6195,
+      "step": 1778
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 6.507070707070707e-06,
+      "loss": 0.9766,
+      "step": 1779
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 6.505050505050505e-06,
+      "loss": 1.206,
+      "step": 1780
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 6.503030303030303e-06,
+      "loss": 0.5074,
+      "step": 1781
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 6.501010101010101e-06,
+      "loss": 0.7408,
+      "step": 1782
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 6.498989898989899e-06,
+      "loss": 1.4632,
+      "step": 1783
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 6.496969696969697e-06,
+      "loss": 0.9874,
+      "step": 1784
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 6.494949494949495e-06,
+      "loss": 1.924,
+      "step": 1785
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 6.4929292929292935e-06,
+      "loss": 1.5919,
+      "step": 1786
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 6.490909090909091e-06,
+      "loss": 1.3703,
+      "step": 1787
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 6.488888888888889e-06,
+      "loss": 1.4174,
+      "step": 1788
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 6.486868686868687e-06,
+      "loss": 1.3984,
+      "step": 1789
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 6.484848484848485e-06,
+      "loss": 1.906,
+      "step": 1790
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 6.482828282828283e-06,
+      "loss": 0.8845,
+      "step": 1791
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 6.480808080808081e-06,
+      "loss": 1.2354,
+      "step": 1792
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 6.478787878787879e-06,
+      "loss": 1.1372,
+      "step": 1793
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 6.476767676767677e-06,
+      "loss": 1.5015,
+      "step": 1794
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 6.4747474747474745e-06,
+      "loss": 0.6601,
+      "step": 1795
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 6.472727272727272e-06,
+      "loss": 1.1124,
+      "step": 1796
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 6.470707070707071e-06,
+      "loss": 1.0178,
+      "step": 1797
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 6.468686868686869e-06,
+      "loss": 1.215,
+      "step": 1798
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 6.466666666666667e-06,
+      "loss": 0.9997,
+      "step": 1799
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 6.464646464646466e-06,
+      "loss": 1.9395,
+      "step": 1800
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.66,
+      "eval_loss": 1.1244035959243774,
+      "eval_roc_auc": 0.948620602892373,
+      "eval_runtime": 92.752,
+      "eval_samples_per_second": 2.156,
+      "eval_steps_per_second": 0.539,
+      "step": 1800
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 6.4626262626262635e-06,
+      "loss": 1.4603,
+      "step": 1801
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 6.460606060606061e-06,
+      "loss": 1.1541,
+      "step": 1802
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 6.458585858585859e-06,
+      "loss": 1.0563,
+      "step": 1803
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 6.456565656565658e-06,
+      "loss": 2.1503,
+      "step": 1804
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 6.454545454545456e-06,
+      "loss": 1.1938,
+      "step": 1805
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 6.452525252525254e-06,
+      "loss": 1.2509,
+      "step": 1806
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 6.450505050505052e-06,
+      "loss": 1.1588,
+      "step": 1807
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 6.4484848484848496e-06,
+      "loss": 0.9768,
+      "step": 1808
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 6.4464646464646474e-06,
+      "loss": 1.3555,
+      "step": 1809
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 6.444444444444445e-06,
+      "loss": 0.7321,
+      "step": 1810
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 6.442424242424243e-06,
+      "loss": 1.0138,
+      "step": 1811
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 6.440404040404041e-06,
+      "loss": 1.2336,
+      "step": 1812
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 6.438383838383839e-06,
+      "loss": 0.5523,
+      "step": 1813
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 6.436363636363637e-06,
+      "loss": 1.1271,
+      "step": 1814
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 6.434343434343436e-06,
+      "loss": 1.3674,
+      "step": 1815
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 6.4323232323232335e-06,
+      "loss": 0.8627,
+      "step": 1816
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 6.430303030303031e-06,
+      "loss": 1.1283,
+      "step": 1817
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 6.428282828282829e-06,
+      "loss": 1.3127,
+      "step": 1818
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 6.426262626262627e-06,
+      "loss": 1.1041,
+      "step": 1819
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 6.424242424242425e-06,
+      "loss": 1.1063,
+      "step": 1820
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 6.422222222222223e-06,
+      "loss": 1.0435,
+      "step": 1821
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 6.420202020202021e-06,
+      "loss": 1.0739,
+      "step": 1822
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 6.418181818181819e-06,
+      "loss": 1.3367,
+      "step": 1823
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 6.416161616161617e-06,
+      "loss": 1.0739,
+      "step": 1824
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 6.4141414141414145e-06,
+      "loss": 1.4018,
+      "step": 1825
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 6.412121212121213e-06,
+      "loss": 0.978,
+      "step": 1826
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 6.410101010101011e-06,
+      "loss": 1.3539,
+      "step": 1827
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 6.408080808080809e-06,
+      "loss": 1.4601,
+      "step": 1828
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 6.406060606060607e-06,
+      "loss": 1.4293,
+      "step": 1829
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 6.404040404040405e-06,
+      "loss": 1.5111,
+      "step": 1830
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 6.402020202020203e-06,
+      "loss": 0.5228,
+      "step": 1831
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 1.6467,
+      "step": 1832
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 6.3979797979797984e-06,
+      "loss": 1.1987,
+      "step": 1833
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 6.395959595959596e-06,
+      "loss": 0.8557,
+      "step": 1834
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 6.393939393939394e-06,
+      "loss": 0.7849,
+      "step": 1835
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 6.391919191919192e-06,
+      "loss": 1.2873,
+      "step": 1836
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 6.389898989898991e-06,
+      "loss": 1.2987,
+      "step": 1837
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 6.387878787878789e-06,
+      "loss": 0.9742,
+      "step": 1838
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 6.385858585858587e-06,
+      "loss": 0.8977,
+      "step": 1839
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 6.3838383838383845e-06,
+      "loss": 0.9261,
+      "step": 1840
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 6.381818181818182e-06,
+      "loss": 1.7393,
+      "step": 1841
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 6.37979797979798e-06,
+      "loss": 0.9767,
+      "step": 1842
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 6.377777777777778e-06,
+      "loss": 1.0733,
+      "step": 1843
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 6.375757575757576e-06,
+      "loss": 2.1037,
+      "step": 1844
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 6.373737373737374e-06,
+      "loss": 0.6599,
+      "step": 1845
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 6.371717171717172e-06,
+      "loss": 0.8912,
+      "step": 1846
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 6.3696969696969706e-06,
+      "loss": 0.9732,
+      "step": 1847
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 6.3676767676767685e-06,
+      "loss": 1.2723,
+      "step": 1848
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 6.365656565656566e-06,
+      "loss": 0.756,
+      "step": 1849
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 6.363636363636364e-06,
+      "loss": 1.7068,
+      "step": 1850
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 6.361616161616162e-06,
+      "loss": 1.4224,
+      "step": 1851
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 6.35959595959596e-06,
+      "loss": 1.4477,
+      "step": 1852
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 6.357575757575758e-06,
+      "loss": 1.6129,
+      "step": 1853
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 6.355555555555556e-06,
+      "loss": 0.9888,
+      "step": 1854
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 6.353535353535354e-06,
+      "loss": 0.7877,
+      "step": 1855
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 6.3515151515151516e-06,
+      "loss": 1.4787,
+      "step": 1856
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 6.3494949494949494e-06,
+      "loss": 0.8055,
+      "step": 1857
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 6.347474747474748e-06,
+      "loss": 0.9103,
+      "step": 1858
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 6.345454545454546e-06,
+      "loss": 1.1366,
+      "step": 1859
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 6.343434343434344e-06,
+      "loss": 0.7223,
+      "step": 1860
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 6.341414141414142e-06,
+      "loss": 0.8694,
+      "step": 1861
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 6.33939393939394e-06,
+      "loss": 1.1275,
+      "step": 1862
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 6.337373737373738e-06,
+      "loss": 1.5934,
+      "step": 1863
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 6.3353535353535355e-06,
+      "loss": 0.9037,
+      "step": 1864
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 6.333333333333333e-06,
+      "loss": 1.4933,
+      "step": 1865
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 6.331313131313131e-06,
+      "loss": 1.4159,
+      "step": 1866
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 6.329292929292929e-06,
+      "loss": 1.1407,
+      "step": 1867
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 6.327272727272727e-06,
+      "loss": 0.989,
+      "step": 1868
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 6.325252525252526e-06,
+      "loss": 1.5903,
+      "step": 1869
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 6.323232323232324e-06,
+      "loss": 0.8766,
+      "step": 1870
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 6.3212121212121216e-06,
+      "loss": 1.3342,
+      "step": 1871
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 6.3191919191919195e-06,
+      "loss": 0.9599,
+      "step": 1872
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 6.317171717171717e-06,
+      "loss": 1.2847,
+      "step": 1873
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 6.315151515151515e-06,
+      "loss": 0.7196,
+      "step": 1874
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 6.313131313131313e-06,
+      "loss": 1.3024,
+      "step": 1875
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 6.311111111111111e-06,
+      "loss": 1.3035,
+      "step": 1876
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 6.309090909090909e-06,
+      "loss": 1.4009,
+      "step": 1877
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 6.307070707070707e-06,
+      "loss": 1.1533,
+      "step": 1878
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 6.305050505050505e-06,
+      "loss": 1.2925,
+      "step": 1879
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 6.303030303030303e-06,
+      "loss": 0.9174,
+      "step": 1880
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 6.301010101010101e-06,
+      "loss": 0.9669,
+      "step": 1881
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 6.298989898989899e-06,
+      "loss": 1.1683,
+      "step": 1882
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 6.296969696969697e-06,
+      "loss": 0.7672,
+      "step": 1883
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 6.294949494949495e-06,
+      "loss": 1.9884,
+      "step": 1884
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 6.292929292929294e-06,
+      "loss": 1.2339,
+      "step": 1885
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 6.290909090909092e-06,
+      "loss": 1.5695,
+      "step": 1886
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 6.28888888888889e-06,
+      "loss": 0.9821,
+      "step": 1887
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 6.286868686868688e-06,
+      "loss": 0.744,
+      "step": 1888
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 6.284848484848486e-06,
+      "loss": 0.9664,
+      "step": 1889
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 6.282828282828284e-06,
+      "loss": 0.8308,
+      "step": 1890
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 6.280808080808082e-06,
+      "loss": 0.9851,
+      "step": 1891
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 6.27878787878788e-06,
+      "loss": 1.2763,
+      "step": 1892
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 6.276767676767678e-06,
+      "loss": 1.1255,
+      "step": 1893
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 6.2747474747474755e-06,
+      "loss": 1.5663,
+      "step": 1894
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 6.2727272727272734e-06,
+      "loss": 0.6567,
+      "step": 1895
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 6.270707070707071e-06,
+      "loss": 1.3764,
+      "step": 1896
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 6.268686868686869e-06,
+      "loss": 1.5109,
+      "step": 1897
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 6.266666666666668e-06,
+      "loss": 0.8455,
+      "step": 1898
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 6.264646464646466e-06,
+      "loss": 1.0275,
+      "step": 1899
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 6.262626262626264e-06,
+      "loss": 1.5876,
+      "step": 1900
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 6.260606060606062e-06,
+      "loss": 1.2757,
+      "step": 1901
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 6.2585858585858595e-06,
+      "loss": 0.8745,
+      "step": 1902
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 6.256565656565657e-06,
+      "loss": 1.3587,
+      "step": 1903
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 6.254545454545455e-06,
+      "loss": 1.3963,
+      "step": 1904
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 6.252525252525253e-06,
+      "loss": 1.3207,
+      "step": 1905
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 6.250505050505051e-06,
+      "loss": 1.0528,
+      "step": 1906
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 6.248484848484849e-06,
+      "loss": 0.8853,
+      "step": 1907
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 6.246464646464647e-06,
+      "loss": 1.334,
+      "step": 1908
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 6.2444444444444456e-06,
+      "loss": 1.1084,
+      "step": 1909
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 6.2424242424242434e-06,
+      "loss": 0.8688,
+      "step": 1910
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 6.240404040404041e-06,
+      "loss": 1.4175,
+      "step": 1911
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 6.238383838383839e-06,
+      "loss": 1.1521,
+      "step": 1912
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 6.236363636363637e-06,
+      "loss": 1.3246,
+      "step": 1913
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 6.234343434343435e-06,
+      "loss": 1.8526,
+      "step": 1914
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 6.232323232323233e-06,
+      "loss": 1.1917,
+      "step": 1915
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 6.230303030303031e-06,
+      "loss": 1.4823,
+      "step": 1916
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 6.228282828282829e-06,
+      "loss": 1.0115,
+      "step": 1917
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 6.2262626262626265e-06,
+      "loss": 1.5093,
+      "step": 1918
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 6.224242424242425e-06,
+      "loss": 1.2525,
+      "step": 1919
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 6.222222222222223e-06,
+      "loss": 1.4834,
+      "step": 1920
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 6.220202020202021e-06,
+      "loss": 1.1005,
+      "step": 1921
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 6.218181818181819e-06,
+      "loss": 0.8547,
+      "step": 1922
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 6.216161616161617e-06,
+      "loss": 0.814,
+      "step": 1923
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 6.214141414141415e-06,
+      "loss": 1.3476,
+      "step": 1924
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 6.212121212121213e-06,
+      "loss": 0.8048,
+      "step": 1925
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 6.2101010101010105e-06,
+      "loss": 1.1417,
+      "step": 1926
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 6.208080808080808e-06,
+      "loss": 0.788,
+      "step": 1927
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 6.206060606060606e-06,
+      "loss": 1.4804,
+      "step": 1928
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 6.204040404040404e-06,
+      "loss": 1.1498,
+      "step": 1929
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 6.202020202020203e-06,
+      "loss": 1.3295,
+      "step": 1930
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 6.200000000000001e-06,
+      "loss": 1.3156,
+      "step": 1931
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 6.197979797979799e-06,
+      "loss": 1.4877,
+      "step": 1932
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 6.1959595959595966e-06,
+      "loss": 1.267,
+      "step": 1933
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 6.1939393939393944e-06,
+      "loss": 1.5335,
+      "step": 1934
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 6.191919191919192e-06,
+      "loss": 1.1038,
+      "step": 1935
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 6.18989898989899e-06,
+      "loss": 1.1042,
+      "step": 1936
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 6.187878787878788e-06,
+      "loss": 1.4715,
+      "step": 1937
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 6.185858585858586e-06,
+      "loss": 0.9374,
+      "step": 1938
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 6.183838383838384e-06,
+      "loss": 1.6596,
+      "step": 1939
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 6.181818181818182e-06,
+      "loss": 0.8745,
+      "step": 1940
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 6.1797979797979805e-06,
+      "loss": 1.1153,
+      "step": 1941
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 6.177777777777778e-06,
+      "loss": 1.065,
+      "step": 1942
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 6.175757575757576e-06,
+      "loss": 1.3358,
+      "step": 1943
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 6.173737373737374e-06,
+      "loss": 1.4458,
+      "step": 1944
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 6.171717171717172e-06,
+      "loss": 1.0028,
+      "step": 1945
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 6.16969696969697e-06,
+      "loss": 1.2845,
+      "step": 1946
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 6.167676767676768e-06,
+      "loss": 1.9714,
+      "step": 1947
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 6.165656565656566e-06,
+      "loss": 2.0366,
+      "step": 1948
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 6.163636363636364e-06,
+      "loss": 1.0181,
+      "step": 1949
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 6.1616161616161615e-06,
+      "loss": 1.3067,
+      "step": 1950
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 6.159595959595959e-06,
+      "loss": 2.1505,
+      "step": 1951
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 6.157575757575758e-06,
+      "loss": 0.9164,
+      "step": 1952
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.155555555555556e-06,
+      "loss": 1.7701,
+      "step": 1953
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.153535353535354e-06,
+      "loss": 1.0989,
+      "step": 1954
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.151515151515152e-06,
+      "loss": 1.7579,
+      "step": 1955
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.14949494949495e-06,
+      "loss": 1.5755,
+      "step": 1956
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.1474747474747476e-06,
+      "loss": 1.1359,
+      "step": 1957
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.1454545454545454e-06,
+      "loss": 1.2325,
+      "step": 1958
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.143434343434343e-06,
+      "loss": 1.6992,
+      "step": 1959
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.141414141414141e-06,
+      "loss": 1.0762,
+      "step": 1960
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.139393939393939e-06,
+      "loss": 0.6751,
+      "step": 1961
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 6.137373737373738e-06,
+      "loss": 1.1812,
+      "step": 1962
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 6.135353535353536e-06,
+      "loss": 0.7995,
+      "step": 1963
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 6.133333333333334e-06,
+      "loss": 1.4167,
+      "step": 1964
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 6.1313131313131315e-06,
+      "loss": 0.9542,
+      "step": 1965
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 6.129292929292929e-06,
+      "loss": 2.0929,
+      "step": 1966
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 6.127272727272727e-06,
+      "loss": 0.6531,
+      "step": 1967
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 6.125252525252525e-06,
+      "loss": 0.7607,
+      "step": 1968
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 6.123232323232323e-06,
+      "loss": 0.8754,
+      "step": 1969
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 6.121212121212121e-06,
+      "loss": 1.0028,
+      "step": 1970
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 6.1191919191919205e-06,
+      "loss": 0.878,
+      "step": 1971
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 6.117171717171718e-06,
+      "loss": 1.0971,
+      "step": 1972
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 6.115151515151516e-06,
+      "loss": 1.1677,
+      "step": 1973
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 6.113131313131314e-06,
+      "loss": 1.4913,
+      "step": 1974
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 6.111111111111112e-06,
+      "loss": 1.1385,
+      "step": 1975
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 6.10909090909091e-06,
+      "loss": 1.0038,
+      "step": 1976
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 6.107070707070708e-06,
+      "loss": 1.2087,
+      "step": 1977
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 6.105050505050506e-06,
+      "loss": 0.6339,
+      "step": 1978
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 6.103030303030304e-06,
+      "loss": 0.9977,
+      "step": 1979
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 6.1010101010101015e-06,
+      "loss": 1.1065,
+      "step": 1980
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 6.0989898989899e-06,
+      "loss": 0.8033,
+      "step": 1981
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 6.096969696969698e-06,
+      "loss": 1.284,
+      "step": 1982
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 6.094949494949496e-06,
+      "loss": 0.7703,
+      "step": 1983
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 6.092929292929294e-06,
+      "loss": 1.1111,
+      "step": 1984
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 6.090909090909092e-06,
+      "loss": 0.7807,
+      "step": 1985
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 6.08888888888889e-06,
+      "loss": 1.3853,
+      "step": 1986
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 6.086868686868688e-06,
+      "loss": 1.2594,
+      "step": 1987
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 6.0848484848484855e-06,
+      "loss": 0.9705,
+      "step": 1988
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 6.082828282828283e-06,
+      "loss": 0.8583,
+      "step": 1989
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 6.080808080808081e-06,
+      "loss": 1.4743,
+      "step": 1990
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 6.07878787878788e-06,
+      "loss": 1.422,
+      "step": 1991
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 6.076767676767678e-06,
+      "loss": 1.4463,
+      "step": 1992
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 6.074747474747476e-06,
+      "loss": 1.2611,
+      "step": 1993
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 6.072727272727274e-06,
+      "loss": 0.7474,
+      "step": 1994
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 6.0707070707070715e-06,
+      "loss": 1.7372,
+      "step": 1995
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.068686868686869e-06,
+      "loss": 0.6365,
+      "step": 1996
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.066666666666667e-06,
+      "loss": 1.3309,
+      "step": 1997
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 6.064646464646465e-06,
+      "loss": 0.5639,
+      "step": 1998
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 6.062626262626263e-06,
+      "loss": 1.017,
+      "step": 1999
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 6.060606060606061e-06,
+      "loss": 2.1052,
+      "step": 2000
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.685,
+      "eval_loss": 1.1022549867630005,
+      "eval_roc_auc": 0.9442341458262987,
+      "eval_runtime": 92.7307,
+      "eval_samples_per_second": 2.157,
+      "eval_steps_per_second": 0.539,
+      "step": 2000
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 6.058585858585859e-06,
+      "loss": 1.4227,
+      "step": 2001
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 6.056565656565658e-06,
+      "loss": 0.9868,
+      "step": 2002
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 6.0545454545454555e-06,
+      "loss": 1.6677,
+      "step": 2003
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 6.052525252525253e-06,
+      "loss": 1.2493,
+      "step": 2004
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 6.050505050505051e-06,
+      "loss": 0.8231,
+      "step": 2005
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 6.048484848484849e-06,
+      "loss": 1.0001,
+      "step": 2006
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 6.046464646464647e-06,
+      "loss": 1.1578,
+      "step": 2007
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 6.044444444444445e-06,
+      "loss": 0.9009,
+      "step": 2008
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 6.042424242424243e-06,
+      "loss": 1.4627,
+      "step": 2009
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 6.040404040404041e-06,
+      "loss": 1.088,
+      "step": 2010
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 6.038383838383839e-06,
+      "loss": 1.6456,
+      "step": 2011
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 6.0363636363636365e-06,
+      "loss": 1.0127,
+      "step": 2012
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 6.034343434343435e-06,
+      "loss": 1.6338,
+      "step": 2013
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 6.032323232323233e-06,
+      "loss": 1.6141,
+      "step": 2014
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 6.030303030303031e-06,
+      "loss": 1.1422,
+      "step": 2015
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 6.028282828282829e-06,
+      "loss": 2.0381,
+      "step": 2016
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 6.026262626262627e-06,
+      "loss": 1.5389,
+      "step": 2017
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 6.024242424242425e-06,
+      "loss": 0.9011,
+      "step": 2018
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 6.0222222222222225e-06,
+      "loss": 1.4352,
+      "step": 2019
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 6.0202020202020204e-06,
+      "loss": 1.1347,
+      "step": 2020
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 6.018181818181818e-06,
+      "loss": 0.8076,
+      "step": 2021
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 6.016161616161616e-06,
+      "loss": 0.971,
+      "step": 2022
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 6.014141414141414e-06,
+      "loss": 0.8322,
+      "step": 2023
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 6.012121212121213e-06,
+      "loss": 1.1994,
+      "step": 2024
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 6.010101010101011e-06,
+      "loss": 1.3917,
+      "step": 2025
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 6.008080808080809e-06,
+      "loss": 0.6523,
+      "step": 2026
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 6.0060606060606065e-06,
+      "loss": 0.9325,
+      "step": 2027
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 6.004040404040404e-06,
+      "loss": 0.973,
+      "step": 2028
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 6.002020202020202e-06,
+      "loss": 1.2415,
+      "step": 2029
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 6e-06,
+      "loss": 0.8554,
+      "step": 2030
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 5.997979797979798e-06,
+      "loss": 1.0195,
+      "step": 2031
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 5.995959595959596e-06,
+      "loss": 0.803,
+      "step": 2032
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 5.993939393939394e-06,
+      "loss": 1.0436,
+      "step": 2033
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 5.9919191919191926e-06,
+      "loss": 0.5034,
+      "step": 2034
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 5.9898989898989904e-06,
+      "loss": 1.1497,
+      "step": 2035
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 5.987878787878788e-06,
+      "loss": 1.1263,
+      "step": 2036
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 5.985858585858586e-06,
+      "loss": 0.7859,
+      "step": 2037
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 5.983838383838384e-06,
+      "loss": 1.3079,
+      "step": 2038
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 5.981818181818182e-06,
+      "loss": 1.6768,
+      "step": 2039
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 5.97979797979798e-06,
+      "loss": 1.3087,
+      "step": 2040
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 5.977777777777778e-06,
+      "loss": 0.5769,
+      "step": 2041
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 5.975757575757576e-06,
+      "loss": 1.1326,
+      "step": 2042
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 5.9737373737373735e-06,
+      "loss": 1.4033,
+      "step": 2043
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 5.9717171717171714e-06,
+      "loss": 1.2424,
+      "step": 2044
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 5.96969696969697e-06,
+      "loss": 0.6403,
+      "step": 2045
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 5.967676767676768e-06,
+      "loss": 1.1359,
+      "step": 2046
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 5.965656565656566e-06,
+      "loss": 1.2865,
+      "step": 2047
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 5.963636363636364e-06,
+      "loss": 0.9206,
+      "step": 2048
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 5.961616161616162e-06,
+      "loss": 0.9522,
+      "step": 2049
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 5.95959595959596e-06,
+      "loss": 0.7603,
+      "step": 2050
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 5.9575757575757575e-06,
+      "loss": 0.6133,
+      "step": 2051
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 5.955555555555555e-06,
+      "loss": 1.7857,
+      "step": 2052
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 5.953535353535353e-06,
+      "loss": 1.6765,
+      "step": 2053
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 5.951515151515151e-06,
+      "loss": 1.2022,
+      "step": 2054
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 5.949494949494949e-06,
+      "loss": 0.8348,
+      "step": 2055
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 5.947474747474749e-06,
+      "loss": 1.126,
+      "step": 2056
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 5.9454545454545465e-06,
+      "loss": 0.8385,
+      "step": 2057
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 5.943434343434344e-06,
+      "loss": 1.2192,
+      "step": 2058
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 5.941414141414142e-06,
+      "loss": 1.7461,
+      "step": 2059
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 5.93939393939394e-06,
+      "loss": 1.1598,
+      "step": 2060
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 5.937373737373738e-06,
+      "loss": 1.2341,
+      "step": 2061
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 5.935353535353536e-06,
+      "loss": 0.9402,
+      "step": 2062
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 5.933333333333335e-06,
+      "loss": 1.2022,
+      "step": 2063
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 5.9313131313131326e-06,
+      "loss": 1.078,
+      "step": 2064
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 5.9292929292929305e-06,
+      "loss": 0.9426,
+      "step": 2065
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 5.927272727272728e-06,
+      "loss": 0.6294,
+      "step": 2066
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 5.925252525252526e-06,
+      "loss": 1.3163,
+      "step": 2067
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 5.923232323232324e-06,
+      "loss": 0.7188,
+      "step": 2068
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 5.921212121212122e-06,
+      "loss": 0.9474,
+      "step": 2069
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 5.91919191919192e-06,
+      "loss": 0.7721,
+      "step": 2070
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 5.917171717171718e-06,
+      "loss": 0.6637,
+      "step": 2071
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 5.915151515151516e-06,
+      "loss": 0.7912,
+      "step": 2072
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 5.9131313131313136e-06,
+      "loss": 1.2061,
+      "step": 2073
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 5.911111111111112e-06,
+      "loss": 1.0913,
+      "step": 2074
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 5.90909090909091e-06,
+      "loss": 1.4181,
+      "step": 2075
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 5.907070707070708e-06,
+      "loss": 0.5473,
+      "step": 2076
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 5.905050505050506e-06,
+      "loss": 1.274,
+      "step": 2077
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 5.903030303030304e-06,
+      "loss": 1.4427,
+      "step": 2078
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 5.901010101010102e-06,
+      "loss": 0.7137,
+      "step": 2079
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 5.8989898989899e-06,
+      "loss": 0.9376,
+      "step": 2080
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 5.8969696969696975e-06,
+      "loss": 0.9813,
+      "step": 2081
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 5.894949494949495e-06,
+      "loss": 0.8271,
+      "step": 2082
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 5.892929292929293e-06,
+      "loss": 0.8751,
+      "step": 2083
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 5.890909090909091e-06,
+      "loss": 0.8884,
+      "step": 2084
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 5.88888888888889e-06,
+      "loss": 1.3278,
+      "step": 2085
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 5.886868686868688e-06,
+      "loss": 0.8314,
+      "step": 2086
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 5.884848484848486e-06,
+      "loss": 1.0427,
+      "step": 2087
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 5.882828282828284e-06,
+      "loss": 1.0275,
+      "step": 2088
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 5.8808080808080815e-06,
+      "loss": 1.1004,
+      "step": 2089
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 5.878787878787879e-06,
+      "loss": 0.6313,
+      "step": 2090
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 5.876767676767677e-06,
+      "loss": 1.0961,
+      "step": 2091
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 5.874747474747475e-06,
+      "loss": 1.2885,
+      "step": 2092
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 5.872727272727273e-06,
+      "loss": 0.7394,
+      "step": 2093
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 5.870707070707071e-06,
+      "loss": 0.8665,
+      "step": 2094
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 5.868686868686869e-06,
+      "loss": 1.3462,
+      "step": 2095
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 5.8666666666666675e-06,
+      "loss": 1.4747,
+      "step": 2096
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 5.864646464646465e-06,
+      "loss": 0.3716,
+      "step": 2097
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 5.862626262626263e-06,
+      "loss": 0.7498,
+      "step": 2098
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 5.860606060606061e-06,
+      "loss": 0.9101,
+      "step": 2099
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 5.858585858585859e-06,
+      "loss": 0.8957,
+      "step": 2100
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 5.856565656565657e-06,
+      "loss": 1.1034,
+      "step": 2101
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 5.854545454545455e-06,
+      "loss": 1.2119,
+      "step": 2102
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 5.852525252525253e-06,
+      "loss": 1.0677,
+      "step": 2103
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 5.850505050505051e-06,
+      "loss": 0.9187,
+      "step": 2104
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 5.8484848484848485e-06,
+      "loss": 1.5606,
+      "step": 2105
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 5.846464646464647e-06,
+      "loss": 1.0399,
+      "step": 2106
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 5.844444444444445e-06,
+      "loss": 1.5945,
+      "step": 2107
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 5.842424242424243e-06,
+      "loss": 0.9385,
+      "step": 2108
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 5.840404040404041e-06,
+      "loss": 0.93,
+      "step": 2109
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 5.838383838383839e-06,
+      "loss": 0.7843,
+      "step": 2110
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 5.836363636363637e-06,
+      "loss": 0.9328,
+      "step": 2111
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 5.834343434343435e-06,
+      "loss": 1.1142,
+      "step": 2112
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 5.8323232323232325e-06,
+      "loss": 1.5553,
+      "step": 2113
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 5.83030303030303e-06,
+      "loss": 0.6005,
+      "step": 2114
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 5.828282828282828e-06,
+      "loss": 0.8463,
+      "step": 2115
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 5.826262626262626e-06,
+      "loss": 1.3489,
+      "step": 2116
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 5.824242424242425e-06,
+      "loss": 1.0468,
+      "step": 2117
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 5.822222222222223e-06,
+      "loss": 1.3091,
+      "step": 2118
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 5.820202020202021e-06,
+      "loss": 0.8379,
+      "step": 2119
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 5.8181818181818185e-06,
+      "loss": 0.8651,
+      "step": 2120
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 5.816161616161616e-06,
+      "loss": 1.0405,
+      "step": 2121
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 5.814141414141414e-06,
+      "loss": 0.9927,
+      "step": 2122
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 5.812121212121212e-06,
+      "loss": 0.6287,
+      "step": 2123
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 5.81010101010101e-06,
+      "loss": 0.5028,
+      "step": 2124
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 5.808080808080808e-06,
+      "loss": 1.1561,
+      "step": 2125
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 5.806060606060606e-06,
+      "loss": 1.1451,
+      "step": 2126
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 5.804040404040404e-06,
+      "loss": 0.6231,
+      "step": 2127
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 5.8020202020202025e-06,
+      "loss": 1.6975,
+      "step": 2128
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 5.8e-06,
+      "loss": 1.1213,
+      "step": 2129
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 5.797979797979798e-06,
+      "loss": 1.4936,
+      "step": 2130
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 5.795959595959596e-06,
+      "loss": 1.0111,
+      "step": 2131
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 5.793939393939394e-06,
+      "loss": 1.3302,
+      "step": 2132
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 5.791919191919192e-06,
+      "loss": 1.8226,
+      "step": 2133
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 5.78989898989899e-06,
+      "loss": 0.6903,
+      "step": 2134
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 5.787878787878788e-06,
+      "loss": 1.0404,
+      "step": 2135
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 5.785858585858586e-06,
+      "loss": 1.2322,
+      "step": 2136
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 5.7838383838383835e-06,
+      "loss": 1.2338,
+      "step": 2137
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 5.781818181818181e-06,
+      "loss": 1.0005,
+      "step": 2138
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 5.77979797979798e-06,
+      "loss": 1.1222,
+      "step": 2139
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 5.777777777777778e-06,
+      "loss": 1.373,
+      "step": 2140
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 5.775757575757577e-06,
+      "loss": 1.0682,
+      "step": 2141
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 5.773737373737375e-06,
+      "loss": 0.6656,
+      "step": 2142
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 5.7717171717171725e-06,
+      "loss": 1.6672,
+      "step": 2143
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 5.76969696969697e-06,
+      "loss": 0.8385,
+      "step": 2144
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 5.767676767676768e-06,
+      "loss": 1.4881,
+      "step": 2145
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 5.765656565656567e-06,
+      "loss": 1.3319,
+      "step": 2146
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 5.763636363636365e-06,
+      "loss": 0.7948,
+      "step": 2147
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 5.761616161616163e-06,
+      "loss": 1.8383,
+      "step": 2148
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 5.759595959595961e-06,
+      "loss": 0.7742,
+      "step": 2149
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 5.7575757575757586e-06,
+      "loss": 2.2225,
+      "step": 2150
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 5.7555555555555564e-06,
+      "loss": 1.8902,
+      "step": 2151
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 5.753535353535354e-06,
+      "loss": 0.7364,
+      "step": 2152
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 5.751515151515152e-06,
+      "loss": 1.275,
+      "step": 2153
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 5.74949494949495e-06,
+      "loss": 1.267,
+      "step": 2154
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 5.747474747474748e-06,
+      "loss": 1.4583,
+      "step": 2155
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 5.745454545454546e-06,
+      "loss": 0.7851,
+      "step": 2156
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 5.743434343434345e-06,
+      "loss": 0.552,
+      "step": 2157
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 5.7414141414141425e-06,
+      "loss": 0.9845,
+      "step": 2158
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 5.73939393939394e-06,
+      "loss": 0.8385,
+      "step": 2159
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 5.737373737373738e-06,
+      "loss": 0.8233,
+      "step": 2160
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 5.735353535353536e-06,
+      "loss": 1.3389,
+      "step": 2161
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 5.733333333333334e-06,
+      "loss": 0.588,
+      "step": 2162
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 5.731313131313132e-06,
+      "loss": 0.7044,
+      "step": 2163
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 5.72929292929293e-06,
+      "loss": 0.9109,
+      "step": 2164
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 5.727272727272728e-06,
+      "loss": 0.8041,
+      "step": 2165
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 5.725252525252526e-06,
+      "loss": 1.1183,
+      "step": 2166
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 5.7232323232323235e-06,
+      "loss": 1.3309,
+      "step": 2167
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 5.721212121212122e-06,
+      "loss": 0.7522,
+      "step": 2168
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 5.71919191919192e-06,
+      "loss": 1.4093,
+      "step": 2169
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 5.717171717171718e-06,
+      "loss": 1.4376,
+      "step": 2170
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 5.715151515151516e-06,
+      "loss": 1.5648,
+      "step": 2171
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 5.713131313131314e-06,
+      "loss": 0.8545,
+      "step": 2172
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 5.711111111111112e-06,
+      "loss": 1.0117,
+      "step": 2173
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 5.7090909090909096e-06,
+      "loss": 0.8118,
+      "step": 2174
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 5.7070707070707075e-06,
+      "loss": 1.7187,
+      "step": 2175
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 5.705050505050505e-06,
+      "loss": 1.0142,
+      "step": 2176
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 5.703030303030303e-06,
+      "loss": 1.1789,
+      "step": 2177
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 5.701010101010102e-06,
+      "loss": 1.1622,
+      "step": 2178
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 5.6989898989899e-06,
+      "loss": 1.1146,
+      "step": 2179
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 5.696969696969698e-06,
+      "loss": 1.3084,
+      "step": 2180
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 5.694949494949496e-06,
+      "loss": 1.1384,
+      "step": 2181
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 5.6929292929292935e-06,
+      "loss": 1.9135,
+      "step": 2182
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 5.690909090909091e-06,
+      "loss": 1.359,
+      "step": 2183
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 5.688888888888889e-06,
+      "loss": 1.4462,
+      "step": 2184
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 5.686868686868687e-06,
+      "loss": 1.0856,
+      "step": 2185
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 5.684848484848485e-06,
+      "loss": 0.6898,
+      "step": 2186
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 5.682828282828283e-06,
+      "loss": 0.3621,
+      "step": 2187
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 5.680808080808081e-06,
+      "loss": 1.4696,
+      "step": 2188
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 5.67878787878788e-06,
+      "loss": 1.2334,
+      "step": 2189
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 5.6767676767676775e-06,
+      "loss": 1.2357,
+      "step": 2190
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 5.674747474747475e-06,
+      "loss": 1.1476,
+      "step": 2191
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 5.672727272727273e-06,
+      "loss": 1.4025,
+      "step": 2192
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 5.670707070707071e-06,
+      "loss": 0.9347,
+      "step": 2193
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 5.668686868686869e-06,
+      "loss": 0.6623,
+      "step": 2194
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 5.666666666666667e-06,
+      "loss": 0.523,
+      "step": 2195
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 5.664646464646465e-06,
+      "loss": 0.9558,
+      "step": 2196
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 5.662626262626263e-06,
+      "loss": 0.5039,
+      "step": 2197
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 5.6606060606060606e-06,
+      "loss": 0.8328,
+      "step": 2198
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 5.6585858585858585e-06,
+      "loss": 0.5865,
+      "step": 2199
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 5.656565656565657e-06,
+      "loss": 1.0731,
+      "step": 2200
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.715,
+      "eval_loss": 0.978618323802948,
+      "eval_roc_auc": 0.9551393498195008,
+      "eval_runtime": 92.745,
+      "eval_samples_per_second": 2.156,
+      "eval_steps_per_second": 0.539,
+      "step": 2200
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 5.654545454545455e-06,
+      "loss": 0.6291,
+      "step": 2201
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 5.652525252525253e-06,
+      "loss": 1.4081,
+      "step": 2202
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 5.650505050505051e-06,
+      "loss": 0.6111,
+      "step": 2203
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 5.648484848484849e-06,
+      "loss": 0.5658,
+      "step": 2204
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 5.646464646464647e-06,
+      "loss": 1.2012,
+      "step": 2205
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 5.6444444444444445e-06,
+      "loss": 1.5018,
+      "step": 2206
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 5.642424242424242e-06,
+      "loss": 0.7732,
+      "step": 2207
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 5.64040404040404e-06,
+      "loss": 0.7005,
+      "step": 2208
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 5.638383838383838e-06,
+      "loss": 1.0035,
+      "step": 2209
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 5.636363636363636e-06,
+      "loss": 0.9318,
+      "step": 2210
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 5.634343434343435e-06,
+      "loss": 0.6199,
+      "step": 2211
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 5.632323232323233e-06,
+      "loss": 1.7591,
+      "step": 2212
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 5.630303030303031e-06,
+      "loss": 1.9962,
+      "step": 2213
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 5.6282828282828285e-06,
+      "loss": 1.3348,
+      "step": 2214
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 5.626262626262626e-06,
+      "loss": 1.7911,
+      "step": 2215
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 5.624242424242424e-06,
+      "loss": 1.0028,
+      "step": 2216
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 5.622222222222222e-06,
+      "loss": 1.3888,
+      "step": 2217
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 5.62020202020202e-06,
+      "loss": 0.8333,
+      "step": 2218
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 5.618181818181818e-06,
+      "loss": 1.0672,
+      "step": 2219
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 5.616161616161616e-06,
+      "loss": 0.7361,
+      "step": 2220
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 5.6141414141414145e-06,
+      "loss": 0.7851,
+      "step": 2221
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 5.612121212121212e-06,
+      "loss": 0.6996,
+      "step": 2222
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 5.61010101010101e-06,
+      "loss": 1.1642,
+      "step": 2223
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 5.608080808080808e-06,
+      "loss": 1.6974,
+      "step": 2224
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 5.606060606060606e-06,
+      "loss": 0.4947,
+      "step": 2225
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 5.604040404040404e-06,
+      "loss": 0.6871,
+      "step": 2226
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 5.602020202020203e-06,
+      "loss": 1.1459,
+      "step": 2227
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 1.2439,
+      "step": 2228
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 5.597979797979799e-06,
+      "loss": 1.0447,
+      "step": 2229
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 5.595959595959597e-06,
+      "loss": 0.9234,
+      "step": 2230
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 5.593939393939395e-06,
+      "loss": 1.1281,
+      "step": 2231
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 5.591919191919193e-06,
+      "loss": 1.3693,
+      "step": 2232
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 5.589898989898991e-06,
+      "loss": 0.9281,
+      "step": 2233
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 5.587878787878789e-06,
+      "loss": 0.7568,
+      "step": 2234
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 5.585858585858587e-06,
+      "loss": 1.1839,
+      "step": 2235
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 5.5838383838383845e-06,
+      "loss": 0.9233,
+      "step": 2236
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 5.5818181818181824e-06,
+      "loss": 0.9163,
+      "step": 2237
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 5.57979797979798e-06,
+      "loss": 1.0797,
+      "step": 2238
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 5.577777777777778e-06,
+      "loss": 0.9486,
+      "step": 2239
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 5.575757575757577e-06,
+      "loss": 1.247,
+      "step": 2240
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 5.573737373737375e-06,
+      "loss": 0.8484,
+      "step": 2241
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 5.571717171717173e-06,
+      "loss": 1.1594,
+      "step": 2242
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 5.569696969696971e-06,
+      "loss": 0.6296,
+      "step": 2243
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 5.5676767676767685e-06,
+      "loss": 1.5354,
+      "step": 2244
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 5.565656565656566e-06,
+      "loss": 0.8907,
+      "step": 2245
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 5.563636363636364e-06,
+      "loss": 1.058,
+      "step": 2246
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 5.561616161616162e-06,
+      "loss": 1.9962,
+      "step": 2247
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 5.55959595959596e-06,
+      "loss": 0.7232,
+      "step": 2248
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 5.557575757575758e-06,
+      "loss": 1.3111,
+      "step": 2249
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 5.555555555555557e-06,
+      "loss": 1.2579,
+      "step": 2250
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 5.5535353535353546e-06,
+      "loss": 1.8402,
+      "step": 2251
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 5.5515151515151524e-06,
+      "loss": 1.1367,
+      "step": 2252
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 5.54949494949495e-06,
+      "loss": 1.5862,
+      "step": 2253
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 5.547474747474748e-06,
+      "loss": 1.8962,
+      "step": 2254
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 5.545454545454546e-06,
+      "loss": 2.0138,
+      "step": 2255
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 5.543434343434344e-06,
+      "loss": 0.4173,
+      "step": 2256
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 5.541414141414142e-06,
+      "loss": 1.2775,
+      "step": 2257
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 5.53939393939394e-06,
+      "loss": 1.1728,
+      "step": 2258
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 5.537373737373738e-06,
+      "loss": 1.1195,
+      "step": 2259
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 5.5353535353535355e-06,
+      "loss": 1.7153,
+      "step": 2260
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 5.533333333333334e-06,
+      "loss": 0.9603,
+      "step": 2261
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 5.531313131313132e-06,
+      "loss": 1.4393,
+      "step": 2262
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 5.52929292929293e-06,
+      "loss": 1.1346,
+      "step": 2263
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 5.527272727272728e-06,
+      "loss": 2.334,
+      "step": 2264
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 5.525252525252526e-06,
+      "loss": 1.4455,
+      "step": 2265
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 5.523232323232324e-06,
+      "loss": 0.6484,
+      "step": 2266
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 5.521212121212122e-06,
+      "loss": 1.7721,
+      "step": 2267
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 5.5191919191919195e-06,
+      "loss": 0.6985,
+      "step": 2268
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 5.517171717171717e-06,
+      "loss": 0.6557,
+      "step": 2269
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 5.515151515151515e-06,
+      "loss": 1.4508,
+      "step": 2270
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 5.513131313131313e-06,
+      "loss": 1.4433,
+      "step": 2271
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 5.511111111111112e-06,
+      "loss": 0.7746,
+      "step": 2272
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 5.50909090909091e-06,
+      "loss": 1.8263,
+      "step": 2273
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 5.507070707070708e-06,
+      "loss": 1.4724,
+      "step": 2274
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 5.5050505050505056e-06,
+      "loss": 0.8633,
+      "step": 2275
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 5.5030303030303034e-06,
+      "loss": 1.1502,
+      "step": 2276
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 5.501010101010101e-06,
+      "loss": 0.7949,
+      "step": 2277
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 5.498989898989899e-06,
+      "loss": 1.135,
+      "step": 2278
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 5.496969696969697e-06,
+      "loss": 0.8594,
+      "step": 2279
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 5.494949494949495e-06,
+      "loss": 1.1268,
+      "step": 2280
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 5.492929292929293e-06,
+      "loss": 0.9269,
+      "step": 2281
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 5.490909090909091e-06,
+      "loss": 0.7575,
+      "step": 2282
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 5.4888888888888895e-06,
+      "loss": 0.5046,
+      "step": 2283
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 5.486868686868687e-06,
+      "loss": 1.034,
+      "step": 2284
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 5.484848484848485e-06,
+      "loss": 1.1943,
+      "step": 2285
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 5.482828282828283e-06,
+      "loss": 1.1939,
+      "step": 2286
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 5.480808080808081e-06,
+      "loss": 0.8344,
+      "step": 2287
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 5.478787878787879e-06,
+      "loss": 1.037,
+      "step": 2288
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 5.476767676767677e-06,
+      "loss": 0.6487,
+      "step": 2289
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 5.474747474747475e-06,
+      "loss": 1.3345,
+      "step": 2290
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 5.472727272727273e-06,
+      "loss": 0.9941,
+      "step": 2291
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 5.4707070707070705e-06,
+      "loss": 1.0636,
+      "step": 2292
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 5.468686868686869e-06,
+      "loss": 1.4844,
+      "step": 2293
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 5.466666666666667e-06,
+      "loss": 1.7755,
+      "step": 2294
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 5.464646464646465e-06,
+      "loss": 0.5156,
+      "step": 2295
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 5.462626262626263e-06,
+      "loss": 0.8952,
+      "step": 2296
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 5.460606060606061e-06,
+      "loss": 0.9811,
+      "step": 2297
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 5.458585858585859e-06,
+      "loss": 1.2664,
+      "step": 2298
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 5.4565656565656566e-06,
+      "loss": 1.2107,
+      "step": 2299
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 5.4545454545454545e-06,
+      "loss": 0.8582,
+      "step": 2300
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 5.452525252525252e-06,
+      "loss": 0.6426,
+      "step": 2301
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 5.45050505050505e-06,
+      "loss": 0.7112,
+      "step": 2302
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 5.448484848484848e-06,
+      "loss": 1.0609,
+      "step": 2303
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 5.446464646464647e-06,
+      "loss": 1.8529,
+      "step": 2304
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 5.444444444444445e-06,
+      "loss": 0.621,
+      "step": 2305
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 5.442424242424243e-06,
+      "loss": 1.0904,
+      "step": 2306
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 5.4404040404040405e-06,
+      "loss": 0.6904,
+      "step": 2307
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 5.438383838383838e-06,
+      "loss": 1.1797,
+      "step": 2308
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 5.436363636363636e-06,
+      "loss": 1.104,
+      "step": 2309
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 5.434343434343434e-06,
+      "loss": 0.8147,
+      "step": 2310
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 5.432323232323232e-06,
+      "loss": 1.0121,
+      "step": 2311
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 5.430303030303032e-06,
+      "loss": 0.6133,
+      "step": 2312
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 5.4282828282828295e-06,
+      "loss": 0.6044,
+      "step": 2313
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 5.4262626262626274e-06,
+      "loss": 0.6593,
+      "step": 2314
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 5.424242424242425e-06,
+      "loss": 1.1951,
+      "step": 2315
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 5.422222222222223e-06,
+      "loss": 0.7497,
+      "step": 2316
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 5.420202020202021e-06,
+      "loss": 1.7507,
+      "step": 2317
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 5.418181818181819e-06,
+      "loss": 0.5175,
+      "step": 2318
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 5.416161616161617e-06,
+      "loss": 1.0152,
+      "step": 2319
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 5.414141414141415e-06,
+      "loss": 1.062,
+      "step": 2320
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 5.412121212121213e-06,
+      "loss": 1.2148,
+      "step": 2321
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 5.410101010101011e-06,
+      "loss": 0.5876,
+      "step": 2322
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 5.408080808080809e-06,
+      "loss": 1.3048,
+      "step": 2323
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 5.406060606060607e-06,
+      "loss": 0.4434,
+      "step": 2324
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 5.404040404040405e-06,
+      "loss": 1.4586,
+      "step": 2325
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 5.402020202020203e-06,
+      "loss": 1.341,
+      "step": 2326
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 5.400000000000001e-06,
+      "loss": 0.7213,
+      "step": 2327
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 5.397979797979799e-06,
+      "loss": 1.1202,
+      "step": 2328
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 5.395959595959597e-06,
+      "loss": 0.8382,
+      "step": 2329
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 5.3939393939393945e-06,
+      "loss": 0.831,
+      "step": 2330
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 5.391919191919192e-06,
+      "loss": 1.0154,
+      "step": 2331
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 5.38989898989899e-06,
+      "loss": 1.0036,
+      "step": 2332
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 5.387878787878789e-06,
+      "loss": 1.4485,
+      "step": 2333
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 5.385858585858587e-06,
+      "loss": 0.6352,
+      "step": 2334
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 5.383838383838385e-06,
+      "loss": 0.825,
+      "step": 2335
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 5.381818181818183e-06,
+      "loss": 0.4636,
+      "step": 2336
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 5.3797979797979805e-06,
+      "loss": 1.2336,
+      "step": 2337
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 5.3777777777777784e-06,
+      "loss": 1.1043,
+      "step": 2338
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 5.375757575757576e-06,
+      "loss": 1.915,
+      "step": 2339
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 5.373737373737374e-06,
+      "loss": 1.0021,
+      "step": 2340
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 5.371717171717172e-06,
+      "loss": 0.7061,
+      "step": 2341
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 5.36969696969697e-06,
+      "loss": 0.6468,
+      "step": 2342
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 5.367676767676768e-06,
+      "loss": 0.719,
+      "step": 2343
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 5.365656565656567e-06,
+      "loss": 1.4511,
+      "step": 2344
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 5.3636363636363645e-06,
+      "loss": 0.9725,
+      "step": 2345
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 5.361616161616162e-06,
+      "loss": 1.1892,
+      "step": 2346
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 5.35959595959596e-06,
+      "loss": 0.851,
+      "step": 2347
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 5.357575757575758e-06,
+      "loss": 1.1306,
+      "step": 2348
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 5.355555555555556e-06,
+      "loss": 0.4612,
+      "step": 2349
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 5.353535353535354e-06,
+      "loss": 1.1062,
+      "step": 2350
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 5.351515151515152e-06,
+      "loss": 0.8347,
+      "step": 2351
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 5.34949494949495e-06,
+      "loss": 0.7687,
+      "step": 2352
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 5.347474747474748e-06,
+      "loss": 0.787,
+      "step": 2353
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 5.3454545454545455e-06,
+      "loss": 0.9578,
+      "step": 2354
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 5.343434343434344e-06,
+      "loss": 0.4967,
+      "step": 2355
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 5.341414141414142e-06,
+      "loss": 0.8238,
+      "step": 2356
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 5.33939393939394e-06,
+      "loss": 0.8988,
+      "step": 2357
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 5.337373737373738e-06,
+      "loss": 1.0675,
+      "step": 2358
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 5.335353535353536e-06,
+      "loss": 0.584,
+      "step": 2359
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 5.333333333333334e-06,
+      "loss": 0.5438,
+      "step": 2360
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 5.3313131313131315e-06,
+      "loss": 0.9494,
+      "step": 2361
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 5.3292929292929294e-06,
+      "loss": 0.97,
+      "step": 2362
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 5.327272727272727e-06,
+      "loss": 0.8431,
+      "step": 2363
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 5.325252525252525e-06,
+      "loss": 1.4164,
+      "step": 2364
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 5.323232323232324e-06,
+      "loss": 1.2717,
+      "step": 2365
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 5.321212121212122e-06,
+      "loss": 1.5975,
+      "step": 2366
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 5.31919191919192e-06,
+      "loss": 1.4165,
+      "step": 2367
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 5.317171717171718e-06,
+      "loss": 1.3602,
+      "step": 2368
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 5.3151515151515155e-06,
+      "loss": 0.6754,
+      "step": 2369
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 5.313131313131313e-06,
+      "loss": 1.1299,
+      "step": 2370
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 5.311111111111111e-06,
+      "loss": 1.8249,
+      "step": 2371
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 5.309090909090909e-06,
+      "loss": 1.226,
+      "step": 2372
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 5.307070707070707e-06,
+      "loss": 1.107,
+      "step": 2373
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 5.305050505050505e-06,
+      "loss": 0.3583,
+      "step": 2374
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 5.303030303030303e-06,
+      "loss": 0.6951,
+      "step": 2375
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 5.3010101010101016e-06,
+      "loss": 1.2418,
+      "step": 2376
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 5.2989898989898994e-06,
+      "loss": 0.6794,
+      "step": 2377
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 5.296969696969697e-06,
+      "loss": 0.726,
+      "step": 2378
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 5.294949494949495e-06,
+      "loss": 1.3202,
+      "step": 2379
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 5.292929292929293e-06,
+      "loss": 0.5351,
+      "step": 2380
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 5.290909090909091e-06,
+      "loss": 1.6404,
+      "step": 2381
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 5.288888888888889e-06,
+      "loss": 1.3909,
+      "step": 2382
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 5.286868686868687e-06,
+      "loss": 1.0947,
+      "step": 2383
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 5.284848484848485e-06,
+      "loss": 0.5925,
+      "step": 2384
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 5.2828282828282825e-06,
+      "loss": 0.7895,
+      "step": 2385
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 5.2808080808080804e-06,
+      "loss": 0.5827,
+      "step": 2386
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 5.278787878787879e-06,
+      "loss": 1.0064,
+      "step": 2387
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 5.276767676767677e-06,
+      "loss": 0.5784,
+      "step": 2388
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 5.274747474747475e-06,
+      "loss": 0.7919,
+      "step": 2389
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 5.272727272727273e-06,
+      "loss": 1.0083,
+      "step": 2390
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 5.270707070707071e-06,
+      "loss": 0.7702,
+      "step": 2391
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 5.268686868686869e-06,
+      "loss": 0.5504,
+      "step": 2392
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 5.2666666666666665e-06,
+      "loss": 1.0769,
+      "step": 2393
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 5.264646464646464e-06,
+      "loss": 1.5117,
+      "step": 2394
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 5.262626262626262e-06,
+      "loss": 1.0804,
+      "step": 2395
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 5.26060606060606e-06,
+      "loss": 1.2362,
+      "step": 2396
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 5.25858585858586e-06,
+      "loss": 1.639,
+      "step": 2397
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 5.256565656565658e-06,
+      "loss": 0.7572,
+      "step": 2398
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 5.2545454545454555e-06,
+      "loss": 0.6502,
+      "step": 2399
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 5.252525252525253e-06,
+      "loss": 1.3848,
+      "step": 2400
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.73,
+      "eval_loss": 0.9378707408905029,
+      "eval_roc_auc": 0.9604709205637219,
+      "eval_runtime": 92.6597,
+      "eval_samples_per_second": 2.158,
+      "eval_steps_per_second": 0.54,
+      "step": 2400
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 5.250505050505051e-06,
+      "loss": 0.6288,
+      "step": 2401
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 5.248484848484849e-06,
+      "loss": 0.7871,
+      "step": 2402
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 5.246464646464647e-06,
+      "loss": 1.4582,
+      "step": 2403
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 5.244444444444445e-06,
+      "loss": 0.7712,
+      "step": 2404
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 5.242424242424244e-06,
+      "loss": 1.0339,
+      "step": 2405
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 5.240404040404042e-06,
+      "loss": 1.7462,
+      "step": 2406
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 5.2383838383838395e-06,
+      "loss": 1.4401,
+      "step": 2407
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 5.236363636363637e-06,
+      "loss": 0.7702,
+      "step": 2408
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 5.234343434343435e-06,
+      "loss": 1.0265,
+      "step": 2409
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 5.232323232323233e-06,
+      "loss": 0.8468,
+      "step": 2410
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 5.230303030303031e-06,
+      "loss": 1.1625,
+      "step": 2411
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 5.228282828282829e-06,
+      "loss": 1.2059,
+      "step": 2412
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 5.226262626262627e-06,
+      "loss": 0.9711,
+      "step": 2413
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 5.224242424242425e-06,
+      "loss": 0.8397,
+      "step": 2414
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 5.2222222222222226e-06,
+      "loss": 0.9561,
+      "step": 2415
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 5.220202020202021e-06,
+      "loss": 0.9161,
+      "step": 2416
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 5.218181818181819e-06,
+      "loss": 1.0942,
+      "step": 2417
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 5.216161616161617e-06,
+      "loss": 1.2512,
+      "step": 2418
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 5.214141414141415e-06,
+      "loss": 0.6749,
+      "step": 2419
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 5.212121212121213e-06,
+      "loss": 1.9947,
+      "step": 2420
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 5.210101010101011e-06,
+      "loss": 1.4307,
+      "step": 2421
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 5.208080808080809e-06,
+      "loss": 1.4534,
+      "step": 2422
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 5.2060606060606065e-06,
+      "loss": 0.5042,
+      "step": 2423
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 5.204040404040404e-06,
+      "loss": 1.1976,
+      "step": 2424
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 5.202020202020202e-06,
+      "loss": 1.2722,
+      "step": 2425
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 5.2e-06,
+      "loss": 0.3441,
+      "step": 2426
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 5.197979797979799e-06,
+      "loss": 0.9591,
+      "step": 2427
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 5.195959595959597e-06,
+      "loss": 0.7513,
+      "step": 2428
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 5.193939393939395e-06,
+      "loss": 0.5837,
+      "step": 2429
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 5.191919191919193e-06,
+      "loss": 1.9432,
+      "step": 2430
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 5.1898989898989905e-06,
+      "loss": 0.7846,
+      "step": 2431
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 5.187878787878788e-06,
+      "loss": 0.593,
+      "step": 2432
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 5.185858585858586e-06,
+      "loss": 0.8016,
+      "step": 2433
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 5.183838383838384e-06,
+      "loss": 0.6764,
+      "step": 2434
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 5.181818181818182e-06,
+      "loss": 1.3374,
+      "step": 2435
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 5.17979797979798e-06,
+      "loss": 0.5448,
+      "step": 2436
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 5.177777777777779e-06,
+      "loss": 0.5585,
+      "step": 2437
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 5.1757575757575765e-06,
+      "loss": 0.7576,
+      "step": 2438
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 5.1737373737373744e-06,
+      "loss": 1.657,
+      "step": 2439
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 5.171717171717172e-06,
+      "loss": 1.144,
+      "step": 2440
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 5.16969696969697e-06,
+      "loss": 1.5122,
+      "step": 2441
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 5.167676767676768e-06,
+      "loss": 0.7271,
+      "step": 2442
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 5.165656565656566e-06,
+      "loss": 0.8013,
+      "step": 2443
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 5.163636363636364e-06,
+      "loss": 0.4336,
+      "step": 2444
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 5.161616161616162e-06,
+      "loss": 1.0787,
+      "step": 2445
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 5.15959595959596e-06,
+      "loss": 1.5404,
+      "step": 2446
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 5.1575757575757575e-06,
+      "loss": 1.1365,
+      "step": 2447
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 5.155555555555556e-06,
+      "loss": 0.4781,
+      "step": 2448
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 5.153535353535354e-06,
+      "loss": 0.973,
+      "step": 2449
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 5.151515151515152e-06,
+      "loss": 0.9384,
+      "step": 2450
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 5.14949494949495e-06,
+      "loss": 1.0128,
+      "step": 2451
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 5.147474747474748e-06,
+      "loss": 0.9746,
+      "step": 2452
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 5.145454545454546e-06,
+      "loss": 0.2915,
+      "step": 2453
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 5.143434343434344e-06,
+      "loss": 1.3564,
+      "step": 2454
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 5.1414141414141415e-06,
+      "loss": 0.5925,
+      "step": 2455
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 5.139393939393939e-06,
+      "loss": 1.0303,
+      "step": 2456
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 5.137373737373737e-06,
+      "loss": 0.9599,
+      "step": 2457
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 5.135353535353535e-06,
+      "loss": 1.4191,
+      "step": 2458
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 5.133333333333334e-06,
+      "loss": 0.5812,
+      "step": 2459
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 5.131313131313132e-06,
+      "loss": 0.95,
+      "step": 2460
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 5.12929292929293e-06,
+      "loss": 1.298,
+      "step": 2461
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 5.1272727272727275e-06,
+      "loss": 0.9872,
+      "step": 2462
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 5.1252525252525254e-06,
+      "loss": 0.8533,
+      "step": 2463
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 5.123232323232323e-06,
+      "loss": 0.7834,
+      "step": 2464
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 5.121212121212121e-06,
+      "loss": 0.326,
+      "step": 2465
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 5.119191919191919e-06,
+      "loss": 0.9791,
+      "step": 2466
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 5.117171717171717e-06,
+      "loss": 1.3773,
+      "step": 2467
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 5.115151515151515e-06,
+      "loss": 0.9681,
+      "step": 2468
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 5.113131313131313e-06,
+      "loss": 1.1159,
+      "step": 2469
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 5.1111111111111115e-06,
+      "loss": 0.999,
+      "step": 2470
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 5.109090909090909e-06,
+      "loss": 0.7469,
+      "step": 2471
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 5.107070707070707e-06,
+      "loss": 1.3121,
+      "step": 2472
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 5.105050505050505e-06,
+      "loss": 0.8989,
+      "step": 2473
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 5.103030303030303e-06,
+      "loss": 1.3475,
+      "step": 2474
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 5.101010101010101e-06,
+      "loss": 0.5503,
+      "step": 2475
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 5.098989898989899e-06,
+      "loss": 0.8575,
+      "step": 2476
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 5.096969696969697e-06,
+      "loss": 0.5704,
+      "step": 2477
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 5.094949494949495e-06,
+      "loss": 0.629,
+      "step": 2478
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 5.0929292929292925e-06,
+      "loss": 0.7765,
+      "step": 2479
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 5.090909090909091e-06,
+      "loss": 0.6862,
+      "step": 2480
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 5.088888888888889e-06,
+      "loss": 0.5322,
+      "step": 2481
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 5.086868686868687e-06,
+      "loss": 1.3679,
+      "step": 2482
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 5.084848484848486e-06,
+      "loss": 1.6839,
+      "step": 2483
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 5.082828282828284e-06,
+      "loss": 0.6693,
+      "step": 2484
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 5.0808080808080815e-06,
+      "loss": 0.828,
+      "step": 2485
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 5.078787878787879e-06,
+      "loss": 0.7878,
+      "step": 2486
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 5.076767676767677e-06,
+      "loss": 1.5082,
+      "step": 2487
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 5.074747474747476e-06,
+      "loss": 1.3718,
+      "step": 2488
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 5.072727272727274e-06,
+      "loss": 0.974,
+      "step": 2489
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 5.070707070707072e-06,
+      "loss": 1.145,
+      "step": 2490
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 5.06868686868687e-06,
+      "loss": 1.125,
+      "step": 2491
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 5.0666666666666676e-06,
+      "loss": 0.9638,
+      "step": 2492
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 5.0646464646464655e-06,
+      "loss": 0.648,
+      "step": 2493
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 5.062626262626263e-06,
+      "loss": 1.3058,
+      "step": 2494
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 5.060606060606061e-06,
+      "loss": 0.5196,
+      "step": 2495
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 5.058585858585859e-06,
+      "loss": 0.7084,
+      "step": 2496
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 5.056565656565657e-06,
+      "loss": 0.7116,
+      "step": 2497
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 5.054545454545455e-06,
+      "loss": 0.6016,
+      "step": 2498
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 5.052525252525254e-06,
+      "loss": 0.7252,
+      "step": 2499
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 5.0505050505050515e-06,
+      "loss": 0.7304,
+      "step": 2500
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 5.048484848484849e-06,
+      "loss": 1.3834,
+      "step": 2501
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 5.046464646464647e-06,
+      "loss": 1.3052,
+      "step": 2502
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 5.044444444444445e-06,
+      "loss": 0.6097,
+      "step": 2503
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 5.042424242424243e-06,
+      "loss": 0.5411,
+      "step": 2504
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 5.040404040404041e-06,
+      "loss": 0.761,
+      "step": 2505
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 5.038383838383839e-06,
+      "loss": 1.3016,
+      "step": 2506
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 5.036363636363637e-06,
+      "loss": 0.9388,
+      "step": 2507
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 5.034343434343435e-06,
+      "loss": 1.0036,
+      "step": 2508
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 5.032323232323233e-06,
+      "loss": 0.7624,
+      "step": 2509
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 5.030303030303031e-06,
+      "loss": 1.0526,
+      "step": 2510
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 5.028282828282829e-06,
+      "loss": 0.9275,
+      "step": 2511
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 5.026262626262627e-06,
+      "loss": 0.7494,
+      "step": 2512
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 5.024242424242425e-06,
+      "loss": 1.2384,
+      "step": 2513
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 5.022222222222223e-06,
+      "loss": 1.0713,
+      "step": 2514
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 5.020202020202021e-06,
+      "loss": 2.2485,
+      "step": 2515
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 5.0181818181818186e-06,
+      "loss": 1.0349,
+      "step": 2516
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 5.0161616161616165e-06,
+      "loss": 0.6196,
+      "step": 2517
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 5.014141414141414e-06,
+      "loss": 0.7799,
+      "step": 2518
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 5.012121212121212e-06,
+      "loss": 0.3146,
+      "step": 2519
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 5.010101010101011e-06,
+      "loss": 0.5669,
+      "step": 2520
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 5.008080808080809e-06,
+      "loss": 1.103,
+      "step": 2521
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 5.006060606060607e-06,
+      "loss": 0.8829,
+      "step": 2522
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 5.004040404040405e-06,
+      "loss": 0.8517,
+      "step": 2523
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 5.0020202020202025e-06,
+      "loss": 0.5966,
+      "step": 2524
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 5e-06,
+      "loss": 0.715,
+      "step": 2525
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 4.997979797979798e-06,
+      "loss": 0.6288,
+      "step": 2526
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 4.995959595959596e-06,
+      "loss": 0.817,
+      "step": 2527
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 4.993939393939394e-06,
+      "loss": 0.8488,
+      "step": 2528
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 4.991919191919192e-06,
+      "loss": 1.33,
+      "step": 2529
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 4.98989898989899e-06,
+      "loss": 0.7859,
+      "step": 2530
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 4.987878787878789e-06,
+      "loss": 1.1282,
+      "step": 2531
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 4.9858585858585865e-06,
+      "loss": 1.4033,
+      "step": 2532
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 4.983838383838384e-06,
+      "loss": 1.0414,
+      "step": 2533
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 4.981818181818182e-06,
+      "loss": 0.9503,
+      "step": 2534
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 4.97979797979798e-06,
+      "loss": 0.542,
+      "step": 2535
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 4.977777777777778e-06,
+      "loss": 0.8323,
+      "step": 2536
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 4.975757575757576e-06,
+      "loss": 0.9942,
+      "step": 2537
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 4.973737373737374e-06,
+      "loss": 0.9438,
+      "step": 2538
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 4.971717171717172e-06,
+      "loss": 0.6143,
+      "step": 2539
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 4.9696969696969696e-06,
+      "loss": 0.8206,
+      "step": 2540
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 4.9676767676767675e-06,
+      "loss": 0.816,
+      "step": 2541
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 4.965656565656566e-06,
+      "loss": 0.8879,
+      "step": 2542
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 4.963636363636364e-06,
+      "loss": 0.8087,
+      "step": 2543
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 4.961616161616162e-06,
+      "loss": 0.614,
+      "step": 2544
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 4.95959595959596e-06,
+      "loss": 0.6494,
+      "step": 2545
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 4.957575757575758e-06,
+      "loss": 1.5708,
+      "step": 2546
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 4.9555555555555565e-06,
+      "loss": 0.982,
+      "step": 2547
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 4.953535353535354e-06,
+      "loss": 1.135,
+      "step": 2548
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 4.951515151515152e-06,
+      "loss": 0.4104,
+      "step": 2549
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 4.94949494949495e-06,
+      "loss": 1.1427,
+      "step": 2550
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 4.947474747474748e-06,
+      "loss": 0.6236,
+      "step": 2551
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 4.945454545454546e-06,
+      "loss": 0.5989,
+      "step": 2552
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 4.943434343434344e-06,
+      "loss": 1.0735,
+      "step": 2553
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 4.941414141414142e-06,
+      "loss": 0.8536,
+      "step": 2554
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 4.93939393939394e-06,
+      "loss": 1.5821,
+      "step": 2555
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 4.937373737373738e-06,
+      "loss": 1.2803,
+      "step": 2556
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 4.935353535353536e-06,
+      "loss": 1.0957,
+      "step": 2557
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 4.933333333333334e-06,
+      "loss": 0.5505,
+      "step": 2558
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 4.931313131313132e-06,
+      "loss": 1.1259,
+      "step": 2559
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 4.92929292929293e-06,
+      "loss": 0.425,
+      "step": 2560
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 4.927272727272728e-06,
+      "loss": 1.1329,
+      "step": 2561
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 4.925252525252526e-06,
+      "loss": 0.9524,
+      "step": 2562
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 4.9232323232323235e-06,
+      "loss": 2.7224,
+      "step": 2563
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 4.9212121212121214e-06,
+      "loss": 1.2368,
+      "step": 2564
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 4.919191919191919e-06,
+      "loss": 1.1427,
+      "step": 2565
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 4.917171717171717e-06,
+      "loss": 0.5483,
+      "step": 2566
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 4.915151515151516e-06,
+      "loss": 0.5427,
+      "step": 2567
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 4.913131313131314e-06,
+      "loss": 0.8848,
+      "step": 2568
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 4.911111111111112e-06,
+      "loss": 0.7803,
+      "step": 2569
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 4.90909090909091e-06,
+      "loss": 0.9745,
+      "step": 2570
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 4.9070707070707075e-06,
+      "loss": 0.7611,
+      "step": 2571
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 4.905050505050505e-06,
+      "loss": 0.6629,
+      "step": 2572
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 4.903030303030303e-06,
+      "loss": 1.182,
+      "step": 2573
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 4.901010101010101e-06,
+      "loss": 0.9688,
+      "step": 2574
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 4.898989898989899e-06,
+      "loss": 0.8067,
+      "step": 2575
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 4.896969696969697e-06,
+      "loss": 0.5568,
+      "step": 2576
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 4.894949494949495e-06,
+      "loss": 1.2614,
+      "step": 2577
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 4.8929292929292936e-06,
+      "loss": 0.9548,
+      "step": 2578
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 4.8909090909090914e-06,
+      "loss": 1.7856,
+      "step": 2579
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 4.888888888888889e-06,
+      "loss": 0.4044,
+      "step": 2580
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 4.886868686868687e-06,
+      "loss": 0.681,
+      "step": 2581
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 4.884848484848485e-06,
+      "loss": 1.0299,
+      "step": 2582
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 4.882828282828283e-06,
+      "loss": 1.5085,
+      "step": 2583
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 4.880808080808081e-06,
+      "loss": 1.3632,
+      "step": 2584
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 4.878787878787879e-06,
+      "loss": 0.7866,
+      "step": 2585
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 4.876767676767677e-06,
+      "loss": 1.0065,
+      "step": 2586
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 4.8747474747474745e-06,
+      "loss": 0.9533,
+      "step": 2587
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 4.872727272727273e-06,
+      "loss": 0.8273,
+      "step": 2588
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 4.870707070707071e-06,
+      "loss": 0.8317,
+      "step": 2589
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 4.868686868686869e-06,
+      "loss": 0.4922,
+      "step": 2590
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 4.866666666666667e-06,
+      "loss": 1.4377,
+      "step": 2591
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 4.864646464646466e-06,
+      "loss": 0.9837,
+      "step": 2592
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 4.8626262626262636e-06,
+      "loss": 1.7168,
+      "step": 2593
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 4.8606060606060615e-06,
+      "loss": 2.1525,
+      "step": 2594
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 4.858585858585859e-06,
+      "loss": 1.1136,
+      "step": 2595
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 4.856565656565657e-06,
+      "loss": 1.5899,
+      "step": 2596
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 4.854545454545455e-06,
+      "loss": 1.1278,
+      "step": 2597
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 4.852525252525253e-06,
+      "loss": 0.7326,
+      "step": 2598
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 4.850505050505051e-06,
+      "loss": 1.3308,
+      "step": 2599
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 4.848484848484849e-06,
+      "loss": 0.9078,
+      "step": 2600
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.715,
+      "eval_loss": 0.9693288207054138,
+      "eval_roc_auc": 0.9448425014189976,
+      "eval_runtime": 93.456,
+      "eval_samples_per_second": 2.14,
+      "eval_steps_per_second": 0.535,
+      "step": 2600
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 4.846464646464647e-06,
+      "loss": 0.9309,
+      "step": 2601
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 4.8444444444444446e-06,
+      "loss": 0.6079,
+      "step": 2602
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 4.842424242424243e-06,
+      "loss": 0.8862,
+      "step": 2603
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 4.840404040404041e-06,
+      "loss": 0.942,
+      "step": 2604
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 4.838383838383839e-06,
+      "loss": 1.2439,
+      "step": 2605
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 4.836363636363637e-06,
+      "loss": 0.4157,
+      "step": 2606
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 4.834343434343435e-06,
+      "loss": 0.5949,
+      "step": 2607
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 4.832323232323233e-06,
+      "loss": 0.7575,
+      "step": 2608
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 4.830303030303031e-06,
+      "loss": 0.9069,
+      "step": 2609
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 4.8282828282828285e-06,
+      "loss": 0.6027,
+      "step": 2610
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 4.826262626262626e-06,
+      "loss": 0.6315,
+      "step": 2611
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 4.824242424242424e-06,
+      "loss": 0.8099,
+      "step": 2612
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 4.822222222222222e-06,
+      "loss": 0.3384,
+      "step": 2613
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 4.820202020202021e-06,
+      "loss": 1.0532,
+      "step": 2614
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 4.818181818181819e-06,
+      "loss": 0.7944,
+      "step": 2615
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 4.816161616161617e-06,
+      "loss": 0.617,
+      "step": 2616
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 4.8141414141414146e-06,
+      "loss": 0.4652,
+      "step": 2617
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 4.8121212121212125e-06,
+      "loss": 0.701,
+      "step": 2618
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 4.81010101010101e-06,
+      "loss": 0.5121,
+      "step": 2619
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 4.808080808080808e-06,
+      "loss": 1.2161,
+      "step": 2620
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 4.806060606060606e-06,
+      "loss": 0.7594,
+      "step": 2621
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 4.804040404040404e-06,
+      "loss": 0.9837,
+      "step": 2622
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 4.802020202020202e-06,
+      "loss": 2.1606,
+      "step": 2623
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 1.2748,
+      "step": 2624
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 4.7979797979797985e-06,
+      "loss": 1.0731,
+      "step": 2625
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 4.795959595959596e-06,
+      "loss": 0.9299,
+      "step": 2626
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 4.793939393939394e-06,
+      "loss": 0.8577,
+      "step": 2627
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 4.791919191919192e-06,
+      "loss": 1.5976,
+      "step": 2628
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 4.78989898989899e-06,
+      "loss": 0.7515,
+      "step": 2629
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 4.787878787878788e-06,
+      "loss": 0.5501,
+      "step": 2630
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 4.785858585858586e-06,
+      "loss": 0.7516,
+      "step": 2631
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 4.783838383838385e-06,
+      "loss": 0.5218,
+      "step": 2632
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 4.7818181818181825e-06,
+      "loss": 0.7855,
+      "step": 2633
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 4.77979797979798e-06,
+      "loss": 0.337,
+      "step": 2634
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 4.777777777777778e-06,
+      "loss": 1.2105,
+      "step": 2635
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 4.775757575757576e-06,
+      "loss": 1.6369,
+      "step": 2636
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 4.773737373737374e-06,
+      "loss": 1.2424,
+      "step": 2637
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 4.771717171717172e-06,
+      "loss": 0.9863,
+      "step": 2638
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 4.769696969696971e-06,
+      "loss": 0.7197,
+      "step": 2639
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 4.7676767676767685e-06,
+      "loss": 0.7352,
+      "step": 2640
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 4.765656565656566e-06,
+      "loss": 0.8249,
+      "step": 2641
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 4.763636363636364e-06,
+      "loss": 1.1496,
+      "step": 2642
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 4.761616161616162e-06,
+      "loss": 0.3653,
+      "step": 2643
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 4.75959595959596e-06,
+      "loss": 0.775,
+      "step": 2644
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 4.757575757575758e-06,
+      "loss": 0.7251,
+      "step": 2645
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 4.755555555555556e-06,
+      "loss": 1.0161,
+      "step": 2646
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 4.753535353535354e-06,
+      "loss": 0.7035,
+      "step": 2647
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 4.751515151515152e-06,
+      "loss": 1.1299,
+      "step": 2648
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 4.7494949494949495e-06,
+      "loss": 0.8096,
+      "step": 2649
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 4.747474747474748e-06,
+      "loss": 0.6828,
+      "step": 2650
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 4.745454545454546e-06,
+      "loss": 1.6178,
+      "step": 2651
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 4.743434343434344e-06,
+      "loss": 0.6944,
+      "step": 2652
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 4.741414141414142e-06,
+      "loss": 0.5605,
+      "step": 2653
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 4.73939393939394e-06,
+      "loss": 1.2115,
+      "step": 2654
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 4.737373737373738e-06,
+      "loss": 0.4938,
+      "step": 2655
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 4.735353535353536e-06,
+      "loss": 0.6724,
+      "step": 2656
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 4.7333333333333335e-06,
+      "loss": 0.9456,
+      "step": 2657
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 4.731313131313131e-06,
+      "loss": 0.8553,
+      "step": 2658
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 4.729292929292929e-06,
+      "loss": 0.9154,
+      "step": 2659
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 4.727272727272728e-06,
+      "loss": 0.637,
+      "step": 2660
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 4.725252525252526e-06,
+      "loss": 0.7265,
+      "step": 2661
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 4.723232323232324e-06,
+      "loss": 1.2994,
+      "step": 2662
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 4.721212121212122e-06,
+      "loss": 0.3251,
+      "step": 2663
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 4.7191919191919195e-06,
+      "loss": 0.9257,
+      "step": 2664
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 4.717171717171717e-06,
+      "loss": 1.085,
+      "step": 2665
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 4.715151515151515e-06,
+      "loss": 1.1147,
+      "step": 2666
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 4.713131313131313e-06,
+      "loss": 0.4137,
+      "step": 2667
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 4.711111111111111e-06,
+      "loss": 2.0956,
+      "step": 2668
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 4.709090909090909e-06,
+      "loss": 1.0884,
+      "step": 2669
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 4.707070707070707e-06,
+      "loss": 1.2468,
+      "step": 2670
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 4.705050505050506e-06,
+      "loss": 1.3473,
+      "step": 2671
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 4.7030303030303035e-06,
+      "loss": 0.4476,
+      "step": 2672
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 4.701010101010101e-06,
+      "loss": 0.6412,
+      "step": 2673
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 4.698989898989899e-06,
+      "loss": 0.7264,
+      "step": 2674
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 4.696969696969698e-06,
+      "loss": 0.8344,
+      "step": 2675
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 4.694949494949496e-06,
+      "loss": 0.8276,
+      "step": 2676
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 4.692929292929294e-06,
+      "loss": 2.0351,
+      "step": 2677
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 4.690909090909092e-06,
+      "loss": 0.7135,
+      "step": 2678
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 4.6888888888888895e-06,
+      "loss": 0.5388,
+      "step": 2679
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 4.6868686868686874e-06,
+      "loss": 0.8666,
+      "step": 2680
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 4.684848484848485e-06,
+      "loss": 0.7439,
+      "step": 2681
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 4.682828282828283e-06,
+      "loss": 0.5785,
+      "step": 2682
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 4.680808080808081e-06,
+      "loss": 0.5038,
+      "step": 2683
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 4.678787878787879e-06,
+      "loss": 1.1107,
+      "step": 2684
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 4.676767676767677e-06,
+      "loss": 1.356,
+      "step": 2685
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 4.674747474747476e-06,
+      "loss": 1.23,
+      "step": 2686
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 4.6727272727272735e-06,
+      "loss": 0.6556,
+      "step": 2687
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 4.670707070707071e-06,
+      "loss": 0.8859,
+      "step": 2688
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 4.668686868686869e-06,
+      "loss": 0.4854,
+      "step": 2689
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 4.666666666666667e-06,
+      "loss": 0.8733,
+      "step": 2690
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 4.664646464646465e-06,
+      "loss": 1.0327,
+      "step": 2691
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 4.662626262626263e-06,
+      "loss": 0.9579,
+      "step": 2692
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 4.660606060606061e-06,
+      "loss": 0.5322,
+      "step": 2693
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 4.658585858585859e-06,
+      "loss": 1.3983,
+      "step": 2694
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 4.656565656565657e-06,
+      "loss": 0.8622,
+      "step": 2695
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 4.654545454545455e-06,
+      "loss": 0.7959,
+      "step": 2696
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 4.652525252525253e-06,
+      "loss": 1.4391,
+      "step": 2697
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 4.650505050505051e-06,
+      "loss": 0.6075,
+      "step": 2698
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 4.648484848484849e-06,
+      "loss": 0.9574,
+      "step": 2699
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 4.646464646464647e-06,
+      "loss": 0.7249,
+      "step": 2700
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 4.644444444444445e-06,
+      "loss": 0.5335,
+      "step": 2701
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 4.642424242424243e-06,
+      "loss": 0.4873,
+      "step": 2702
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 4.6404040404040406e-06,
+      "loss": 1.7045,
+      "step": 2703
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 4.6383838383838384e-06,
+      "loss": 0.713,
+      "step": 2704
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 4.636363636363636e-06,
+      "loss": 1.0056,
+      "step": 2705
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 4.634343434343434e-06,
+      "loss": 1.5834,
+      "step": 2706
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 4.632323232323233e-06,
+      "loss": 1.5886,
+      "step": 2707
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 4.630303030303031e-06,
+      "loss": 0.5229,
+      "step": 2708
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 4.628282828282829e-06,
+      "loss": 1.4257,
+      "step": 2709
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 4.626262626262627e-06,
+      "loss": 0.8699,
+      "step": 2710
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 4.6242424242424245e-06,
+      "loss": 1.9244,
+      "step": 2711
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 4.622222222222222e-06,
+      "loss": 1.121,
+      "step": 2712
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 4.62020202020202e-06,
+      "loss": 1.0661,
+      "step": 2713
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 4.618181818181818e-06,
+      "loss": 0.7242,
+      "step": 2714
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 4.616161616161616e-06,
+      "loss": 1.0415,
+      "step": 2715
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 4.614141414141414e-06,
+      "loss": 0.7325,
+      "step": 2716
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 4.612121212121212e-06,
+      "loss": 0.7019,
+      "step": 2717
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 4.6101010101010106e-06,
+      "loss": 0.6352,
+      "step": 2718
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 4.6080808080808085e-06,
+      "loss": 0.8934,
+      "step": 2719
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 4.606060606060606e-06,
+      "loss": 0.7027,
+      "step": 2720
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 4.604040404040404e-06,
+      "loss": 0.9991,
+      "step": 2721
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 4.602020202020203e-06,
+      "loss": 0.7274,
+      "step": 2722
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 0.9128,
+      "step": 2723
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 4.597979797979799e-06,
+      "loss": 0.9311,
+      "step": 2724
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 4.595959595959597e-06,
+      "loss": 1.9084,
+      "step": 2725
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 4.5939393939393945e-06,
+      "loss": 1.001,
+      "step": 2726
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 4.591919191919192e-06,
+      "loss": 0.521,
+      "step": 2727
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 4.58989898989899e-06,
+      "loss": 1.4438,
+      "step": 2728
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 4.587878787878788e-06,
+      "loss": 0.845,
+      "step": 2729
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 4.585858585858586e-06,
+      "loss": 2.1989,
+      "step": 2730
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 4.583838383838384e-06,
+      "loss": 0.7665,
+      "step": 2731
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 4.581818181818183e-06,
+      "loss": 0.7116,
+      "step": 2732
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 4.579797979797981e-06,
+      "loss": 1.0826,
+      "step": 2733
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 4.5777777777777785e-06,
+      "loss": 0.8713,
+      "step": 2734
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 4.575757575757576e-06,
+      "loss": 0.5465,
+      "step": 2735
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 4.573737373737374e-06,
+      "loss": 0.7161,
+      "step": 2736
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 4.571717171717172e-06,
+      "loss": 1.5261,
+      "step": 2737
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 4.56969696969697e-06,
+      "loss": 1.8705,
+      "step": 2738
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 4.567676767676768e-06,
+      "loss": 0.7267,
+      "step": 2739
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 4.565656565656566e-06,
+      "loss": 0.6813,
+      "step": 2740
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 4.563636363636364e-06,
+      "loss": 1.3732,
+      "step": 2741
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 4.5616161616161616e-06,
+      "loss": 0.81,
+      "step": 2742
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 4.55959595959596e-06,
+      "loss": 0.7636,
+      "step": 2743
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 4.557575757575758e-06,
+      "loss": 0.6444,
+      "step": 2744
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 4.555555555555556e-06,
+      "loss": 0.658,
+      "step": 2745
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 4.553535353535354e-06,
+      "loss": 0.8259,
+      "step": 2746
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 4.551515151515152e-06,
+      "loss": 0.6385,
+      "step": 2747
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 4.54949494949495e-06,
+      "loss": 1.5268,
+      "step": 2748
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 4.547474747474748e-06,
+      "loss": 0.6562,
+      "step": 2749
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 4.5454545454545455e-06,
+      "loss": 0.8436,
+      "step": 2750
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 4.543434343434343e-06,
+      "loss": 0.3661,
+      "step": 2751
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 4.541414141414141e-06,
+      "loss": 0.7206,
+      "step": 2752
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 4.539393939393939e-06,
+      "loss": 0.6081,
+      "step": 2753
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 4.537373737373738e-06,
+      "loss": 0.6723,
+      "step": 2754
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 4.535353535353536e-06,
+      "loss": 0.8044,
+      "step": 2755
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 4.533333333333334e-06,
+      "loss": 1.7323,
+      "step": 2756
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 4.531313131313132e-06,
+      "loss": 0.6618,
+      "step": 2757
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 4.5292929292929295e-06,
+      "loss": 0.5044,
+      "step": 2758
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 4.527272727272727e-06,
+      "loss": 1.3708,
+      "step": 2759
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 4.525252525252526e-06,
+      "loss": 0.9281,
+      "step": 2760
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 4.523232323232324e-06,
+      "loss": 0.6275,
+      "step": 2761
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 4.521212121212122e-06,
+      "loss": 0.3496,
+      "step": 2762
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 4.51919191919192e-06,
+      "loss": 0.2517,
+      "step": 2763
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 4.517171717171718e-06,
+      "loss": 0.2934,
+      "step": 2764
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 4.5151515151515155e-06,
+      "loss": 1.2888,
+      "step": 2765
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 4.513131313131313e-06,
+      "loss": 0.7528,
+      "step": 2766
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 4.511111111111111e-06,
+      "loss": 1.4203,
+      "step": 2767
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 4.50909090909091e-06,
+      "loss": 1.6167,
+      "step": 2768
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 4.507070707070708e-06,
+      "loss": 1.0386,
+      "step": 2769
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 4.505050505050506e-06,
+      "loss": 0.774,
+      "step": 2770
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 4.503030303030304e-06,
+      "loss": 1.1637,
+      "step": 2771
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 4.501010101010102e-06,
+      "loss": 1.1607,
+      "step": 2772
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 4.4989898989898995e-06,
+      "loss": 1.0942,
+      "step": 2773
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 4.496969696969697e-06,
+      "loss": 1.0283,
+      "step": 2774
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 4.494949494949495e-06,
+      "loss": 0.4139,
+      "step": 2775
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 4.492929292929293e-06,
+      "loss": 0.5364,
+      "step": 2776
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 4.490909090909091e-06,
+      "loss": 0.5863,
+      "step": 2777
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 4.488888888888889e-06,
+      "loss": 0.3303,
+      "step": 2778
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 4.486868686868688e-06,
+      "loss": 0.6582,
+      "step": 2779
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 4.4848484848484855e-06,
+      "loss": 1.5709,
+      "step": 2780
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 4.4828282828282834e-06,
+      "loss": 1.2704,
+      "step": 2781
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 4.480808080808081e-06,
+      "loss": 0.3199,
+      "step": 2782
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 4.478787878787879e-06,
+      "loss": 0.7726,
+      "step": 2783
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 4.476767676767677e-06,
+      "loss": 0.562,
+      "step": 2784
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 4.474747474747475e-06,
+      "loss": 0.4225,
+      "step": 2785
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 4.472727272727273e-06,
+      "loss": 0.7335,
+      "step": 2786
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 4.470707070707071e-06,
+      "loss": 1.0428,
+      "step": 2787
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 4.468686868686869e-06,
+      "loss": 0.9216,
+      "step": 2788
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 4.4666666666666665e-06,
+      "loss": 1.309,
+      "step": 2789
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 4.464646464646465e-06,
+      "loss": 0.9163,
+      "step": 2790
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 4.462626262626263e-06,
+      "loss": 0.6938,
+      "step": 2791
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 4.460606060606061e-06,
+      "loss": 0.615,
+      "step": 2792
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 4.458585858585859e-06,
+      "loss": 1.2773,
+      "step": 2793
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 4.456565656565657e-06,
+      "loss": 1.4348,
+      "step": 2794
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 4.454545454545455e-06,
+      "loss": 1.2602,
+      "step": 2795
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 4.452525252525253e-06,
+      "loss": 1.1798,
+      "step": 2796
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 4.4505050505050505e-06,
+      "loss": 0.5505,
+      "step": 2797
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 4.448484848484848e-06,
+      "loss": 0.6484,
+      "step": 2798
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 4.446464646464646e-06,
+      "loss": 0.8797,
+      "step": 2799
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 4.444444444444444e-06,
+      "loss": 1.6524,
+      "step": 2800
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.72,
+      "eval_loss": 0.9333136081695557,
+      "eval_roc_auc": 0.9495847682858164,
+      "eval_runtime": 91.5061,
+      "eval_samples_per_second": 2.186,
+      "eval_steps_per_second": 0.546,
+      "step": 2800
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 4.442424242424243e-06,
+      "loss": 0.6077,
+      "step": 2801
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 4.440404040404041e-06,
+      "loss": 0.8502,
+      "step": 2802
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 4.438383838383839e-06,
+      "loss": 0.6796,
+      "step": 2803
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 4.436363636363637e-06,
+      "loss": 0.7211,
+      "step": 2804
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 4.434343434343435e-06,
+      "loss": 1.0487,
+      "step": 2805
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 4.432323232323233e-06,
+      "loss": 0.7676,
+      "step": 2806
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 4.430303030303031e-06,
+      "loss": 0.86,
+      "step": 2807
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 4.428282828282829e-06,
+      "loss": 0.5751,
+      "step": 2808
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 4.426262626262627e-06,
+      "loss": 1.1329,
+      "step": 2809
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 4.424242424242425e-06,
+      "loss": 0.7446,
+      "step": 2810
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 4.422222222222223e-06,
+      "loss": 1.1863,
+      "step": 2811
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 4.4202020202020205e-06,
+      "loss": 1.5304,
+      "step": 2812
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 4.418181818181818e-06,
+      "loss": 1.0871,
+      "step": 2813
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 4.416161616161616e-06,
+      "loss": 0.8416,
+      "step": 2814
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 4.414141414141415e-06,
+      "loss": 0.4256,
+      "step": 2815
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 4.412121212121213e-06,
+      "loss": 0.5639,
+      "step": 2816
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 4.410101010101011e-06,
+      "loss": 0.7508,
+      "step": 2817
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 4.408080808080809e-06,
+      "loss": 1.5642,
+      "step": 2818
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 4.4060606060606066e-06,
+      "loss": 1.1873,
+      "step": 2819
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 4.4040404040404044e-06,
+      "loss": 0.4959,
+      "step": 2820
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 4.402020202020202e-06,
+      "loss": 0.782,
+      "step": 2821
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 4.4e-06,
+      "loss": 0.4862,
+      "step": 2822
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 4.397979797979798e-06,
+      "loss": 1.727,
+      "step": 2823
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 4.395959595959596e-06,
+      "loss": 0.3639,
+      "step": 2824
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 4.393939393939394e-06,
+      "loss": 0.9373,
+      "step": 2825
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 4.391919191919193e-06,
+      "loss": 1.527,
+      "step": 2826
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 4.3898989898989905e-06,
+      "loss": 1.473,
+      "step": 2827
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 4.387878787878788e-06,
+      "loss": 0.7149,
+      "step": 2828
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 4.385858585858586e-06,
+      "loss": 0.8874,
+      "step": 2829
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 4.383838383838384e-06,
+      "loss": 0.9616,
+      "step": 2830
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 4.381818181818182e-06,
+      "loss": 0.7886,
+      "step": 2831
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 4.37979797979798e-06,
+      "loss": 0.7545,
+      "step": 2832
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 4.377777777777778e-06,
+      "loss": 1.3287,
+      "step": 2833
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 4.375757575757576e-06,
+      "loss": 0.8255,
+      "step": 2834
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 4.373737373737374e-06,
+      "loss": 1.0221,
+      "step": 2835
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 4.3717171717171715e-06,
+      "loss": 0.7006,
+      "step": 2836
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 4.36969696969697e-06,
+      "loss": 0.2831,
+      "step": 2837
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 4.367676767676768e-06,
+      "loss": 1.2727,
+      "step": 2838
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 4.365656565656566e-06,
+      "loss": 1.0753,
+      "step": 2839
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 4.363636363636364e-06,
+      "loss": 1.321,
+      "step": 2840
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 4.361616161616162e-06,
+      "loss": 0.8696,
+      "step": 2841
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 4.35959595959596e-06,
+      "loss": 0.9835,
+      "step": 2842
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 4.3575757575757576e-06,
+      "loss": 0.8719,
+      "step": 2843
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 4.3555555555555555e-06,
+      "loss": 1.0896,
+      "step": 2844
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 4.353535353535353e-06,
+      "loss": 0.9562,
+      "step": 2845
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 4.351515151515152e-06,
+      "loss": 1.395,
+      "step": 2846
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 4.34949494949495e-06,
+      "loss": 0.8114,
+      "step": 2847
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 4.347474747474748e-06,
+      "loss": 0.6779,
+      "step": 2848
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 4.345454545454546e-06,
+      "loss": 0.6374,
+      "step": 2849
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 4.343434343434344e-06,
+      "loss": 0.8503,
+      "step": 2850
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 4.341414141414142e-06,
+      "loss": 1.3479,
+      "step": 2851
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 4.33939393939394e-06,
+      "loss": 1.0793,
+      "step": 2852
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 4.337373737373738e-06,
+      "loss": 1.3441,
+      "step": 2853
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 4.335353535353536e-06,
+      "loss": 0.7558,
+      "step": 2854
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 4.333333333333334e-06,
+      "loss": 0.672,
+      "step": 2855
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 4.331313131313132e-06,
+      "loss": 1.0334,
+      "step": 2856
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 4.32929292929293e-06,
+      "loss": 0.7363,
+      "step": 2857
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 4.327272727272728e-06,
+      "loss": 0.4105,
+      "step": 2858
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 4.3252525252525255e-06,
+      "loss": 0.6627,
+      "step": 2859
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 4.323232323232323e-06,
+      "loss": 0.5998,
+      "step": 2860
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 4.321212121212121e-06,
+      "loss": 1.0636,
+      "step": 2861
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 4.31919191919192e-06,
+      "loss": 1.0235,
+      "step": 2862
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 4.317171717171718e-06,
+      "loss": 1.1727,
+      "step": 2863
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 4.315151515151516e-06,
+      "loss": 0.5119,
+      "step": 2864
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 4.313131313131314e-06,
+      "loss": 0.8592,
+      "step": 2865
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 4.3111111111111115e-06,
+      "loss": 0.6352,
+      "step": 2866
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 4.309090909090909e-06,
+      "loss": 0.6388,
+      "step": 2867
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 4.307070707070707e-06,
+      "loss": 1.2654,
+      "step": 2868
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 4.305050505050505e-06,
+      "loss": 0.7732,
+      "step": 2869
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 4.303030303030303e-06,
+      "loss": 0.4567,
+      "step": 2870
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 4.301010101010101e-06,
+      "loss": 0.1676,
+      "step": 2871
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 4.298989898989899e-06,
+      "loss": 1.5982,
+      "step": 2872
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 4.296969696969698e-06,
+      "loss": 0.9477,
+      "step": 2873
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 4.2949494949494955e-06,
+      "loss": 0.8248,
+      "step": 2874
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 4.292929292929293e-06,
+      "loss": 0.8126,
+      "step": 2875
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 4.290909090909091e-06,
+      "loss": 0.8441,
+      "step": 2876
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 4.288888888888889e-06,
+      "loss": 0.2647,
+      "step": 2877
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 4.286868686868687e-06,
+      "loss": 0.9693,
+      "step": 2878
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 4.284848484848485e-06,
+      "loss": 0.8248,
+      "step": 2879
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 4.282828282828283e-06,
+      "loss": 0.4498,
+      "step": 2880
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 4.280808080808081e-06,
+      "loss": 0.625,
+      "step": 2881
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 4.278787878787879e-06,
+      "loss": 1.0219,
+      "step": 2882
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 4.276767676767677e-06,
+      "loss": 0.5392,
+      "step": 2883
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 4.274747474747475e-06,
+      "loss": 0.7574,
+      "step": 2884
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 4.272727272727273e-06,
+      "loss": 0.7474,
+      "step": 2885
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 4.270707070707071e-06,
+      "loss": 0.7665,
+      "step": 2886
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 4.268686868686869e-06,
+      "loss": 0.4037,
+      "step": 2887
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 4.266666666666668e-06,
+      "loss": 0.2997,
+      "step": 2888
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 4.2646464646464655e-06,
+      "loss": 0.8117,
+      "step": 2889
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 4.262626262626263e-06,
+      "loss": 0.4673,
+      "step": 2890
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 4.260606060606061e-06,
+      "loss": 0.9325,
+      "step": 2891
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 4.258585858585859e-06,
+      "loss": 0.6915,
+      "step": 2892
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 4.256565656565657e-06,
+      "loss": 1.38,
+      "step": 2893
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 4.254545454545455e-06,
+      "loss": 0.7299,
+      "step": 2894
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 4.252525252525253e-06,
+      "loss": 1.1214,
+      "step": 2895
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 4.250505050505051e-06,
+      "loss": 0.5503,
+      "step": 2896
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 4.248484848484849e-06,
+      "loss": 1.3365,
+      "step": 2897
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 4.246464646464647e-06,
+      "loss": 0.7731,
+      "step": 2898
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 4.244444444444445e-06,
+      "loss": 0.7459,
+      "step": 2899
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 4.242424242424243e-06,
+      "loss": 0.4123,
+      "step": 2900
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 4.240404040404041e-06,
+      "loss": 0.475,
+      "step": 2901
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 4.238383838383839e-06,
+      "loss": 0.438,
+      "step": 2902
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 4.236363636363637e-06,
+      "loss": 0.5928,
+      "step": 2903
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 4.234343434343435e-06,
+      "loss": 0.8769,
+      "step": 2904
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 4.2323232323232325e-06,
+      "loss": 0.9683,
+      "step": 2905
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 4.2303030303030304e-06,
+      "loss": 0.9266,
+      "step": 2906
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 4.228282828282828e-06,
+      "loss": 0.9664,
+      "step": 2907
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 4.226262626262626e-06,
+      "loss": 1.1255,
+      "step": 2908
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 4.224242424242425e-06,
+      "loss": 1.2856,
+      "step": 2909
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 4.222222222222223e-06,
+      "loss": 0.6585,
+      "step": 2910
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 4.220202020202021e-06,
+      "loss": 1.005,
+      "step": 2911
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 4.218181818181819e-06,
+      "loss": 1.4453,
+      "step": 2912
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 4.2161616161616165e-06,
+      "loss": 0.5394,
+      "step": 2913
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 4.214141414141414e-06,
+      "loss": 1.0795,
+      "step": 2914
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 4.212121212121212e-06,
+      "loss": 1.1064,
+      "step": 2915
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 4.21010101010101e-06,
+      "loss": 1.2652,
+      "step": 2916
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 4.208080808080808e-06,
+      "loss": 0.8113,
+      "step": 2917
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 4.206060606060606e-06,
+      "loss": 1.5112,
+      "step": 2918
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 4.204040404040405e-06,
+      "loss": 0.5917,
+      "step": 2919
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 4.2020202020202026e-06,
+      "loss": 0.7526,
+      "step": 2920
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 0.8112,
+      "step": 2921
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 4.197979797979798e-06,
+      "loss": 0.9579,
+      "step": 2922
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 4.195959595959596e-06,
+      "loss": 0.6267,
+      "step": 2923
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 4.193939393939394e-06,
+      "loss": 0.6672,
+      "step": 2924
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 4.191919191919192e-06,
+      "loss": 1.0861,
+      "step": 2925
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 4.18989898989899e-06,
+      "loss": 0.6355,
+      "step": 2926
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 4.187878787878788e-06,
+      "loss": 0.4475,
+      "step": 2927
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 4.185858585858586e-06,
+      "loss": 1.6321,
+      "step": 2928
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 4.1838383838383835e-06,
+      "loss": 0.7355,
+      "step": 2929
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 4.181818181818182e-06,
+      "loss": 0.6559,
+      "step": 2930
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 4.17979797979798e-06,
+      "loss": 0.403,
+      "step": 2931
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 4.177777777777778e-06,
+      "loss": 0.6569,
+      "step": 2932
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 4.175757575757576e-06,
+      "loss": 2.3174,
+      "step": 2933
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 4.173737373737375e-06,
+      "loss": 0.6037,
+      "step": 2934
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 4.1717171717171726e-06,
+      "loss": 0.6694,
+      "step": 2935
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 4.1696969696969705e-06,
+      "loss": 0.9071,
+      "step": 2936
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 4.167676767676768e-06,
+      "loss": 0.715,
+      "step": 2937
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 4.165656565656566e-06,
+      "loss": 0.7504,
+      "step": 2938
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 4.163636363636364e-06,
+      "loss": 0.5353,
+      "step": 2939
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 4.161616161616162e-06,
+      "loss": 1.2605,
+      "step": 2940
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 4.15959595959596e-06,
+      "loss": 0.7529,
+      "step": 2941
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 4.157575757575758e-06,
+      "loss": 0.5857,
+      "step": 2942
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 4.155555555555556e-06,
+      "loss": 0.7818,
+      "step": 2943
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 4.1535353535353536e-06,
+      "loss": 0.6948,
+      "step": 2944
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 4.151515151515152e-06,
+      "loss": 1.5391,
+      "step": 2945
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 4.14949494949495e-06,
+      "loss": 1.8515,
+      "step": 2946
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 4.147474747474748e-06,
+      "loss": 0.6452,
+      "step": 2947
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 4.145454545454546e-06,
+      "loss": 0.7999,
+      "step": 2948
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 4.143434343434344e-06,
+      "loss": 0.846,
+      "step": 2949
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 4.141414141414142e-06,
+      "loss": 0.8614,
+      "step": 2950
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 4.13939393939394e-06,
+      "loss": 1.399,
+      "step": 2951
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 4.1373737373737375e-06,
+      "loss": 1.1852,
+      "step": 2952
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 4.135353535353535e-06,
+      "loss": 0.4412,
+      "step": 2953
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 4.133333333333333e-06,
+      "loss": 0.9191,
+      "step": 2954
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 4.131313131313132e-06,
+      "loss": 0.4209,
+      "step": 2955
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 4.12929292929293e-06,
+      "loss": 0.326,
+      "step": 2956
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 4.127272727272728e-06,
+      "loss": 0.6608,
+      "step": 2957
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 4.125252525252526e-06,
+      "loss": 0.5656,
+      "step": 2958
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 4.1232323232323236e-06,
+      "loss": 0.5943,
+      "step": 2959
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 4.1212121212121215e-06,
+      "loss": 1.0092,
+      "step": 2960
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 4.119191919191919e-06,
+      "loss": 0.9088,
+      "step": 2961
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 4.117171717171717e-06,
+      "loss": 1.403,
+      "step": 2962
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 4.115151515151515e-06,
+      "loss": 1.0485,
+      "step": 2963
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 4.113131313131313e-06,
+      "loss": 1.1592,
+      "step": 2964
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 4.111111111111111e-06,
+      "loss": 0.7007,
+      "step": 2965
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 4.10909090909091e-06,
+      "loss": 0.5619,
+      "step": 2966
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 4.1070707070707075e-06,
+      "loss": 0.6494,
+      "step": 2967
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 4.105050505050505e-06,
+      "loss": 0.8343,
+      "step": 2968
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 4.103030303030303e-06,
+      "loss": 1.0686,
+      "step": 2969
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 4.101010101010101e-06,
+      "loss": 0.8192,
+      "step": 2970
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 4.098989898989899e-06,
+      "loss": 0.9747,
+      "step": 2971
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 4.096969696969697e-06,
+      "loss": 0.8644,
+      "step": 2972
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 4.094949494949495e-06,
+      "loss": 1.4844,
+      "step": 2973
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 4.092929292929294e-06,
+      "loss": 0.3326,
+      "step": 2974
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 4.0909090909090915e-06,
+      "loss": 0.5621,
+      "step": 2975
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 4.088888888888889e-06,
+      "loss": 1.2129,
+      "step": 2976
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 4.086868686868687e-06,
+      "loss": 0.6396,
+      "step": 2977
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 4.084848484848485e-06,
+      "loss": 0.4176,
+      "step": 2978
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 4.082828282828283e-06,
+      "loss": 0.5333,
+      "step": 2979
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 4.080808080808081e-06,
+      "loss": 1.0023,
+      "step": 2980
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 4.07878787878788e-06,
+      "loss": 1.165,
+      "step": 2981
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 4.0767676767676775e-06,
+      "loss": 0.6893,
+      "step": 2982
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 4.0747474747474754e-06,
+      "loss": 1.2751,
+      "step": 2983
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 4.072727272727273e-06,
+      "loss": 0.852,
+      "step": 2984
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 4.070707070707071e-06,
+      "loss": 0.8091,
+      "step": 2985
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 4.068686868686869e-06,
+      "loss": 0.6642,
+      "step": 2986
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 4.066666666666667e-06,
+      "loss": 0.5964,
+      "step": 2987
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 4.064646464646465e-06,
+      "loss": 0.8106,
+      "step": 2988
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 4.062626262626263e-06,
+      "loss": 0.6307,
+      "step": 2989
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 4.060606060606061e-06,
+      "loss": 0.2311,
+      "step": 2990
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 4.058585858585859e-06,
+      "loss": 1.0837,
+      "step": 2991
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 4.056565656565657e-06,
+      "loss": 0.9384,
+      "step": 2992
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 4.054545454545455e-06,
+      "loss": 0.824,
+      "step": 2993
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 4.052525252525253e-06,
+      "loss": 0.4433,
+      "step": 2994
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 4.050505050505051e-06,
+      "loss": 0.5421,
+      "step": 2995
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 4.048484848484849e-06,
+      "loss": 0.515,
+      "step": 2996
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 4.046464646464647e-06,
+      "loss": 0.9609,
+      "step": 2997
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 4.044444444444445e-06,
+      "loss": 0.5632,
+      "step": 2998
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 4.0424242424242425e-06,
+      "loss": 0.4953,
+      "step": 2999
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 4.04040404040404e-06,
+      "loss": 0.6489,
+      "step": 3000
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.74,
+      "eval_loss": 0.8801469206809998,
+      "eval_roc_auc": 0.95593459239844,
+      "eval_runtime": 92.7581,
+      "eval_samples_per_second": 2.156,
+      "eval_steps_per_second": 0.539,
+      "step": 3000
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 4.038383838383838e-06,
+      "loss": 0.3603,
+      "step": 3001
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 4.036363636363637e-06,
+      "loss": 0.8171,
+      "step": 3002
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 4.034343434343435e-06,
+      "loss": 0.8517,
+      "step": 3003
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 4.032323232323233e-06,
+      "loss": 0.666,
+      "step": 3004
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 4.030303030303031e-06,
+      "loss": 0.3167,
+      "step": 3005
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 4.0282828282828285e-06,
+      "loss": 0.6139,
+      "step": 3006
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 4.0262626262626264e-06,
+      "loss": 1.18,
+      "step": 3007
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 4.024242424242424e-06,
+      "loss": 0.3982,
+      "step": 3008
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 4.022222222222222e-06,
+      "loss": 0.4287,
+      "step": 3009
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 4.02020202020202e-06,
+      "loss": 0.9556,
+      "step": 3010
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 4.018181818181818e-06,
+      "loss": 0.9961,
+      "step": 3011
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 4.016161616161616e-06,
+      "loss": 0.5763,
+      "step": 3012
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 4.014141414141415e-06,
+      "loss": 1.5288,
+      "step": 3013
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 4.0121212121212125e-06,
+      "loss": 0.7004,
+      "step": 3014
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 4.01010101010101e-06,
+      "loss": 0.3401,
+      "step": 3015
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 4.008080808080808e-06,
+      "loss": 0.6034,
+      "step": 3016
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 4.006060606060607e-06,
+      "loss": 1.2351,
+      "step": 3017
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 4.004040404040405e-06,
+      "loss": 1.0158,
+      "step": 3018
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 4.002020202020203e-06,
+      "loss": 0.8902,
+      "step": 3019
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.5056,
+      "step": 3020
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 3.9979797979797986e-06,
+      "loss": 0.3876,
+      "step": 3021
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 3.9959595959595964e-06,
+      "loss": 0.6158,
+      "step": 3022
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 3.993939393939394e-06,
+      "loss": 0.4551,
+      "step": 3023
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 3.991919191919192e-06,
+      "loss": 0.5088,
+      "step": 3024
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 3.98989898989899e-06,
+      "loss": 0.5898,
+      "step": 3025
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 3.987878787878788e-06,
+      "loss": 0.7506,
+      "step": 3026
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 3.985858585858587e-06,
+      "loss": 1.0914,
+      "step": 3027
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 3.983838383838385e-06,
+      "loss": 0.6392,
+      "step": 3028
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 3.9818181818181825e-06,
+      "loss": 0.6385,
+      "step": 3029
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 3.97979797979798e-06,
+      "loss": 1.1522,
+      "step": 3030
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 3.977777777777778e-06,
+      "loss": 0.9641,
+      "step": 3031
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 3.975757575757576e-06,
+      "loss": 0.8162,
+      "step": 3032
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 3.973737373737374e-06,
+      "loss": 0.9109,
+      "step": 3033
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 3.971717171717172e-06,
+      "loss": 0.5782,
+      "step": 3034
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 3.96969696969697e-06,
+      "loss": 0.2033,
+      "step": 3035
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 3.967676767676768e-06,
+      "loss": 0.9792,
+      "step": 3036
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 3.965656565656566e-06,
+      "loss": 0.4523,
+      "step": 3037
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 3.963636363636364e-06,
+      "loss": 1.3578,
+      "step": 3038
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 3.961616161616162e-06,
+      "loss": 0.6433,
+      "step": 3039
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 3.95959595959596e-06,
+      "loss": 0.3644,
+      "step": 3040
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 3.957575757575758e-06,
+      "loss": 1.135,
+      "step": 3041
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 3.955555555555556e-06,
+      "loss": 0.823,
+      "step": 3042
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 3.953535353535354e-06,
+      "loss": 0.303,
+      "step": 3043
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 3.951515151515152e-06,
+      "loss": 1.162,
+      "step": 3044
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 3.9494949494949496e-06,
+      "loss": 1.7519,
+      "step": 3045
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 3.9474747474747474e-06,
+      "loss": 1.5264,
+      "step": 3046
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 3.945454545454545e-06,
+      "loss": 1.4264,
+      "step": 3047
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 3.943434343434343e-06,
+      "loss": 0.5725,
+      "step": 3048
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 3.941414141414142e-06,
+      "loss": 0.5025,
+      "step": 3049
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 3.93939393939394e-06,
+      "loss": 0.9415,
+      "step": 3050
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 3.937373737373738e-06,
+      "loss": 0.8571,
+      "step": 3051
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 3.935353535353536e-06,
+      "loss": 0.5476,
+      "step": 3052
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 3.9333333333333335e-06,
+      "loss": 0.7257,
+      "step": 3053
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 3.931313131313131e-06,
+      "loss": 0.5071,
+      "step": 3054
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 3.929292929292929e-06,
+      "loss": 0.4733,
+      "step": 3055
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 3.927272727272727e-06,
+      "loss": 0.7704,
+      "step": 3056
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 3.925252525252525e-06,
+      "loss": 0.3645,
+      "step": 3057
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 3.923232323232323e-06,
+      "loss": 1.0038,
+      "step": 3058
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 3.921212121212122e-06,
+      "loss": 0.5526,
+      "step": 3059
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 3.9191919191919196e-06,
+      "loss": 0.547,
+      "step": 3060
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 3.9171717171717175e-06,
+      "loss": 1.4945,
+      "step": 3061
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 3.915151515151515e-06,
+      "loss": 0.4833,
+      "step": 3062
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 3.913131313131314e-06,
+      "loss": 0.4274,
+      "step": 3063
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 3.911111111111112e-06,
+      "loss": 0.6901,
+      "step": 3064
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 3.90909090909091e-06,
+      "loss": 0.6968,
+      "step": 3065
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 3.907070707070708e-06,
+      "loss": 0.55,
+      "step": 3066
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 3.905050505050506e-06,
+      "loss": 1.3718,
+      "step": 3067
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 3.9030303030303035e-06,
+      "loss": 0.485,
+      "step": 3068
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 3.901010101010101e-06,
+      "loss": 1.4134,
+      "step": 3069
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 3.898989898989899e-06,
+      "loss": 0.5167,
+      "step": 3070
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 3.896969696969697e-06,
+      "loss": 0.9239,
+      "step": 3071
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 3.894949494949495e-06,
+      "loss": 0.9522,
+      "step": 3072
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 3.892929292929293e-06,
+      "loss": 0.702,
+      "step": 3073
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 3.890909090909092e-06,
+      "loss": 0.668,
+      "step": 3074
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 3.88888888888889e-06,
+      "loss": 0.5215,
+      "step": 3075
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 3.8868686868686875e-06,
+      "loss": 1.0593,
+      "step": 3076
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 3.884848484848485e-06,
+      "loss": 0.6111,
+      "step": 3077
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 3.882828282828283e-06,
+      "loss": 1.0809,
+      "step": 3078
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 3.880808080808081e-06,
+      "loss": 0.8752,
+      "step": 3079
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 3.878787878787879e-06,
+      "loss": 0.5605,
+      "step": 3080
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 3.876767676767677e-06,
+      "loss": 0.7624,
+      "step": 3081
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 3.874747474747475e-06,
+      "loss": 0.396,
+      "step": 3082
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 3.872727272727273e-06,
+      "loss": 0.6357,
+      "step": 3083
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 3.8707070707070706e-06,
+      "loss": 0.876,
+      "step": 3084
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 3.868686868686869e-06,
+      "loss": 1.0803,
+      "step": 3085
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 3.866666666666667e-06,
+      "loss": 1.7811,
+      "step": 3086
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 3.864646464646465e-06,
+      "loss": 0.2787,
+      "step": 3087
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 3.862626262626263e-06,
+      "loss": 0.6573,
+      "step": 3088
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 3.860606060606061e-06,
+      "loss": 0.6186,
+      "step": 3089
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 3.858585858585859e-06,
+      "loss": 1.4797,
+      "step": 3090
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 3.856565656565657e-06,
+      "loss": 1.2739,
+      "step": 3091
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 3.8545454545454545e-06,
+      "loss": 0.7925,
+      "step": 3092
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 3.852525252525252e-06,
+      "loss": 0.5578,
+      "step": 3093
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 3.85050505050505e-06,
+      "loss": 1.1563,
+      "step": 3094
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 3.848484848484848e-06,
+      "loss": 0.3552,
+      "step": 3095
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 3.846464646464647e-06,
+      "loss": 1.1381,
+      "step": 3096
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 3.844444444444445e-06,
+      "loss": 0.5336,
+      "step": 3097
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 3.842424242424243e-06,
+      "loss": 0.8805,
+      "step": 3098
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 3.840404040404041e-06,
+      "loss": 1.2589,
+      "step": 3099
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 3.8383838383838385e-06,
+      "loss": 1.3872,
+      "step": 3100
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 3.836363636363636e-06,
+      "loss": 0.4488,
+      "step": 3101
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 3.834343434343435e-06,
+      "loss": 0.8813,
+      "step": 3102
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 3.832323232323233e-06,
+      "loss": 2.1793,
+      "step": 3103
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 3.830303030303031e-06,
+      "loss": 0.4809,
+      "step": 3104
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 3.828282828282829e-06,
+      "loss": 0.8709,
+      "step": 3105
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 3.826262626262627e-06,
+      "loss": 0.501,
+      "step": 3106
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 3.8242424242424245e-06,
+      "loss": 0.7825,
+      "step": 3107
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 3.8222222222222224e-06,
+      "loss": 0.6158,
+      "step": 3108
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 3.82020202020202e-06,
+      "loss": 0.4566,
+      "step": 3109
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 3.818181818181819e-06,
+      "loss": 0.8008,
+      "step": 3110
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 3.816161616161617e-06,
+      "loss": 0.826,
+      "step": 3111
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 3.8141414141414144e-06,
+      "loss": 0.2345,
+      "step": 3112
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 3.8121212121212127e-06,
+      "loss": 0.3684,
+      "step": 3113
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 3.8101010101010106e-06,
+      "loss": 0.7266,
+      "step": 3114
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 3.8080808080808085e-06,
+      "loss": 0.8501,
+      "step": 3115
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 3.8060606060606064e-06,
+      "loss": 1.496,
+      "step": 3116
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 3.8040404040404043e-06,
+      "loss": 0.6509,
+      "step": 3117
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 3.8020202020202026e-06,
+      "loss": 0.7466,
+      "step": 3118
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.5013,
+      "step": 3119
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 3.7979797979797984e-06,
+      "loss": 0.6078,
+      "step": 3120
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 3.7959595959595962e-06,
+      "loss": 0.6164,
+      "step": 3121
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 3.793939393939394e-06,
+      "loss": 0.7004,
+      "step": 3122
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 3.791919191919192e-06,
+      "loss": 0.2268,
+      "step": 3123
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 3.7898989898989903e-06,
+      "loss": 0.5725,
+      "step": 3124
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 3.7878787878787882e-06,
+      "loss": 0.7361,
+      "step": 3125
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 3.785858585858586e-06,
+      "loss": 0.8704,
+      "step": 3126
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 3.783838383838384e-06,
+      "loss": 0.4579,
+      "step": 3127
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 3.781818181818182e-06,
+      "loss": 0.4745,
+      "step": 3128
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 3.77979797979798e-06,
+      "loss": 0.3197,
+      "step": 3129
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 3.777777777777778e-06,
+      "loss": 0.8377,
+      "step": 3130
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 3.775757575757576e-06,
+      "loss": 0.3994,
+      "step": 3131
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 3.773737373737374e-06,
+      "loss": 1.4767,
+      "step": 3132
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 3.7717171717171717e-06,
+      "loss": 0.795,
+      "step": 3133
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 3.76969696969697e-06,
+      "loss": 0.7527,
+      "step": 3134
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 3.767676767676768e-06,
+      "loss": 0.6088,
+      "step": 3135
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 3.765656565656566e-06,
+      "loss": 0.2356,
+      "step": 3136
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 3.7636363636363637e-06,
+      "loss": 0.4475,
+      "step": 3137
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 3.7616161616161616e-06,
+      "loss": 1.2145,
+      "step": 3138
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 3.7595959595959595e-06,
+      "loss": 0.7352,
+      "step": 3139
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 3.757575757575758e-06,
+      "loss": 0.6493,
+      "step": 3140
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 3.7555555555555557e-06,
+      "loss": 0.7521,
+      "step": 3141
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 3.7535353535353536e-06,
+      "loss": 0.4749,
+      "step": 3142
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 3.7515151515151515e-06,
+      "loss": 0.8484,
+      "step": 3143
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 3.74949494949495e-06,
+      "loss": 0.864,
+      "step": 3144
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 3.747474747474748e-06,
+      "loss": 0.7516,
+      "step": 3145
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 3.745454545454546e-06,
+      "loss": 0.7114,
+      "step": 3146
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 3.743434343434344e-06,
+      "loss": 0.5769,
+      "step": 3147
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 3.7414141414141418e-06,
+      "loss": 0.4367,
+      "step": 3148
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 3.73939393939394e-06,
+      "loss": 0.592,
+      "step": 3149
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 3.737373737373738e-06,
+      "loss": 0.7094,
+      "step": 3150
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 3.735353535353536e-06,
+      "loss": 0.8303,
+      "step": 3151
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 3.7333333333333337e-06,
+      "loss": 0.592,
+      "step": 3152
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 3.7313131313131316e-06,
+      "loss": 0.5272,
+      "step": 3153
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 3.72929292929293e-06,
+      "loss": 0.4336,
+      "step": 3154
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 3.727272727272728e-06,
+      "loss": 1.6868,
+      "step": 3155
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 3.7252525252525257e-06,
+      "loss": 0.9638,
+      "step": 3156
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 3.7232323232323236e-06,
+      "loss": 0.5104,
+      "step": 3157
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 3.7212121212121215e-06,
+      "loss": 1.3323,
+      "step": 3158
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 3.7191919191919194e-06,
+      "loss": 0.8579,
+      "step": 3159
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 3.7171717171717177e-06,
+      "loss": 0.6129,
+      "step": 3160
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 3.7151515151515156e-06,
+      "loss": 0.5547,
+      "step": 3161
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 3.7131313131313135e-06,
+      "loss": 0.7023,
+      "step": 3162
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 3.7111111111111113e-06,
+      "loss": 0.4664,
+      "step": 3163
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 3.7090909090909092e-06,
+      "loss": 0.6765,
+      "step": 3164
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 3.7070707070707075e-06,
+      "loss": 0.4567,
+      "step": 3165
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 3.7050505050505054e-06,
+      "loss": 0.8241,
+      "step": 3166
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 3.7030303030303033e-06,
+      "loss": 0.2705,
+      "step": 3167
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 3.701010101010101e-06,
+      "loss": 0.5281,
+      "step": 3168
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 3.698989898989899e-06,
+      "loss": 0.8981,
+      "step": 3169
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 3.6969696969696974e-06,
+      "loss": 0.4112,
+      "step": 3170
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 3.6949494949494953e-06,
+      "loss": 0.734,
+      "step": 3171
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 3.692929292929293e-06,
+      "loss": 0.642,
+      "step": 3172
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 3.690909090909091e-06,
+      "loss": 1.5424,
+      "step": 3173
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 3.688888888888889e-06,
+      "loss": 0.9304,
+      "step": 3174
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 3.686868686868687e-06,
+      "loss": 0.3909,
+      "step": 3175
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 3.684848484848485e-06,
+      "loss": 0.5522,
+      "step": 3176
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 3.682828282828283e-06,
+      "loss": 0.9482,
+      "step": 3177
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 3.680808080808081e-06,
+      "loss": 0.4179,
+      "step": 3178
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 3.678787878787879e-06,
+      "loss": 0.8268,
+      "step": 3179
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 3.6767676767676767e-06,
+      "loss": 2.0221,
+      "step": 3180
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 3.674747474747475e-06,
+      "loss": 0.7144,
+      "step": 3181
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 3.672727272727273e-06,
+      "loss": 0.8897,
+      "step": 3182
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 3.670707070707071e-06,
+      "loss": 1.3963,
+      "step": 3183
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 3.6686868686868687e-06,
+      "loss": 2.2231,
+      "step": 3184
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 3.6666666666666666e-06,
+      "loss": 0.5766,
+      "step": 3185
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 3.664646464646465e-06,
+      "loss": 0.9618,
+      "step": 3186
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 3.662626262626263e-06,
+      "loss": 1.0085,
+      "step": 3187
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 3.660606060606061e-06,
+      "loss": 1.01,
+      "step": 3188
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 3.658585858585859e-06,
+      "loss": 0.2111,
+      "step": 3189
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 3.6565656565656573e-06,
+      "loss": 1.3851,
+      "step": 3190
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 3.654545454545455e-06,
+      "loss": 0.2749,
+      "step": 3191
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 3.652525252525253e-06,
+      "loss": 0.9974,
+      "step": 3192
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 3.650505050505051e-06,
+      "loss": 0.6794,
+      "step": 3193
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 3.648484848484849e-06,
+      "loss": 0.546,
+      "step": 3194
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 3.6464646464646467e-06,
+      "loss": 0.7786,
+      "step": 3195
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 3.644444444444445e-06,
+      "loss": 0.5393,
+      "step": 3196
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 3.642424242424243e-06,
+      "loss": 1.1669,
+      "step": 3197
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 3.640404040404041e-06,
+      "loss": 0.6961,
+      "step": 3198
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 3.6383838383838387e-06,
+      "loss": 0.7942,
+      "step": 3199
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 3.6363636363636366e-06,
+      "loss": 0.4556,
+      "step": 3200
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.78,
+      "eval_loss": 0.8396649956703186,
+      "eval_roc_auc": 0.9581413426382692,
+      "eval_runtime": 93.1958,
+      "eval_samples_per_second": 2.146,
+      "eval_steps_per_second": 0.537,
+      "step": 3200
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 3.634343434343435e-06,
+      "loss": 1.311,
+      "step": 3201
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 3.6323232323232328e-06,
+      "loss": 0.5855,
+      "step": 3202
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 3.6303030303030307e-06,
+      "loss": 0.7364,
+      "step": 3203
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 3.6282828282828286e-06,
+      "loss": 0.4475,
+      "step": 3204
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 3.6262626262626264e-06,
+      "loss": 0.694,
+      "step": 3205
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 3.6242424242424248e-06,
+      "loss": 1.9542,
+      "step": 3206
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 3.6222222222222226e-06,
+      "loss": 0.6599,
+      "step": 3207
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 3.6202020202020205e-06,
+      "loss": 0.4797,
+      "step": 3208
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 3.6181818181818184e-06,
+      "loss": 0.3419,
+      "step": 3209
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 3.6161616161616163e-06,
+      "loss": 0.8877,
+      "step": 3210
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 3.614141414141414e-06,
+      "loss": 0.8702,
+      "step": 3211
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 3.6121212121212125e-06,
+      "loss": 0.8095,
+      "step": 3212
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 3.6101010101010104e-06,
+      "loss": 0.534,
+      "step": 3213
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 3.6080808080808083e-06,
+      "loss": 0.8923,
+      "step": 3214
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 3.606060606060606e-06,
+      "loss": 1.4274,
+      "step": 3215
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 3.604040404040404e-06,
+      "loss": 0.9524,
+      "step": 3216
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 3.6020202020202024e-06,
+      "loss": 0.5435,
+      "step": 3217
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 0.4756,
+      "step": 3218
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 3.597979797979798e-06,
+      "loss": 0.4264,
+      "step": 3219
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 3.595959595959596e-06,
+      "loss": 1.2534,
+      "step": 3220
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 3.593939393939394e-06,
+      "loss": 0.6482,
+      "step": 3221
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 3.5919191919191922e-06,
+      "loss": 0.9833,
+      "step": 3222
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 3.58989898989899e-06,
+      "loss": 1.0569,
+      "step": 3223
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 3.587878787878788e-06,
+      "loss": 0.8642,
+      "step": 3224
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 3.585858585858586e-06,
+      "loss": 0.4163,
+      "step": 3225
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 3.5838383838383838e-06,
+      "loss": 1.1071,
+      "step": 3226
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 3.5818181818181817e-06,
+      "loss": 0.6465,
+      "step": 3227
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 3.57979797979798e-06,
+      "loss": 0.8656,
+      "step": 3228
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 3.577777777777778e-06,
+      "loss": 0.7338,
+      "step": 3229
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 3.575757575757576e-06,
+      "loss": 0.9209,
+      "step": 3230
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 3.573737373737374e-06,
+      "loss": 0.906,
+      "step": 3231
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 3.5717171717171724e-06,
+      "loss": 1.0328,
+      "step": 3232
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 3.5696969696969703e-06,
+      "loss": 0.643,
+      "step": 3233
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 3.567676767676768e-06,
+      "loss": 0.9685,
+      "step": 3234
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 3.565656565656566e-06,
+      "loss": 0.4199,
+      "step": 3235
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 3.563636363636364e-06,
+      "loss": 0.784,
+      "step": 3236
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 3.5616161616161622e-06,
+      "loss": 0.9196,
+      "step": 3237
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 3.55959595959596e-06,
+      "loss": 0.6312,
+      "step": 3238
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 3.557575757575758e-06,
+      "loss": 0.6664,
+      "step": 3239
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 3.555555555555556e-06,
+      "loss": 0.6953,
+      "step": 3240
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 3.553535353535354e-06,
+      "loss": 0.8262,
+      "step": 3241
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 3.551515151515152e-06,
+      "loss": 0.5,
+      "step": 3242
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 3.54949494949495e-06,
+      "loss": 0.5486,
+      "step": 3243
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 3.547474747474748e-06,
+      "loss": 0.9257,
+      "step": 3244
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 3.5454545454545458e-06,
+      "loss": 0.3889,
+      "step": 3245
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 3.5434343434343437e-06,
+      "loss": 1.7364,
+      "step": 3246
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 3.5414141414141416e-06,
+      "loss": 0.9573,
+      "step": 3247
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 3.53939393939394e-06,
+      "loss": 0.8852,
+      "step": 3248
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 3.5373737373737378e-06,
+      "loss": 0.7919,
+      "step": 3249
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 3.5353535353535356e-06,
+      "loss": 0.5042,
+      "step": 3250
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 3.5333333333333335e-06,
+      "loss": 0.7527,
+      "step": 3251
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 3.5313131313131314e-06,
+      "loss": 0.7532,
+      "step": 3252
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 3.5292929292929297e-06,
+      "loss": 0.4825,
+      "step": 3253
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 3.5272727272727276e-06,
+      "loss": 0.3832,
+      "step": 3254
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 3.5252525252525255e-06,
+      "loss": 0.3512,
+      "step": 3255
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 3.5232323232323234e-06,
+      "loss": 0.9748,
+      "step": 3256
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 3.5212121212121213e-06,
+      "loss": 1.1839,
+      "step": 3257
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 3.5191919191919196e-06,
+      "loss": 0.5186,
+      "step": 3258
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 3.5171717171717175e-06,
+      "loss": 0.487,
+      "step": 3259
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 3.5151515151515154e-06,
+      "loss": 0.513,
+      "step": 3260
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 3.5131313131313133e-06,
+      "loss": 0.2154,
+      "step": 3261
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 3.511111111111111e-06,
+      "loss": 0.6452,
+      "step": 3262
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 3.509090909090909e-06,
+      "loss": 0.7484,
+      "step": 3263
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 3.5070707070707073e-06,
+      "loss": 0.5485,
+      "step": 3264
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 3.5050505050505052e-06,
+      "loss": 0.2341,
+      "step": 3265
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 3.503030303030303e-06,
+      "loss": 0.5584,
+      "step": 3266
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 3.501010101010101e-06,
+      "loss": 0.7118,
+      "step": 3267
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 3.498989898989899e-06,
+      "loss": 0.518,
+      "step": 3268
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 3.496969696969697e-06,
+      "loss": 0.9716,
+      "step": 3269
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 3.494949494949495e-06,
+      "loss": 0.3277,
+      "step": 3270
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 3.492929292929293e-06,
+      "loss": 0.9658,
+      "step": 3271
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 3.4909090909090913e-06,
+      "loss": 0.6826,
+      "step": 3272
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 3.4888888888888896e-06,
+      "loss": 0.8305,
+      "step": 3273
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 3.4868686868686875e-06,
+      "loss": 0.42,
+      "step": 3274
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 3.4848484848484854e-06,
+      "loss": 0.451,
+      "step": 3275
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 3.4828282828282833e-06,
+      "loss": 0.5816,
+      "step": 3276
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 3.480808080808081e-06,
+      "loss": 0.4236,
+      "step": 3277
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 3.4787878787878795e-06,
+      "loss": 0.7329,
+      "step": 3278
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 3.4767676767676774e-06,
+      "loss": 0.6797,
+      "step": 3279
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 3.4747474747474752e-06,
+      "loss": 0.5225,
+      "step": 3280
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 3.472727272727273e-06,
+      "loss": 0.3818,
+      "step": 3281
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 3.470707070707071e-06,
+      "loss": 0.5088,
+      "step": 3282
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 3.468686868686869e-06,
+      "loss": 0.6725,
+      "step": 3283
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 3.4666666666666672e-06,
+      "loss": 1.0687,
+      "step": 3284
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 3.464646464646465e-06,
+      "loss": 0.5243,
+      "step": 3285
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 3.462626262626263e-06,
+      "loss": 0.5713,
+      "step": 3286
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 3.460606060606061e-06,
+      "loss": 0.643,
+      "step": 3287
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 3.4585858585858588e-06,
+      "loss": 0.486,
+      "step": 3288
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 3.456565656565657e-06,
+      "loss": 0.5217,
+      "step": 3289
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 3.454545454545455e-06,
+      "loss": 0.9512,
+      "step": 3290
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 3.452525252525253e-06,
+      "loss": 0.4897,
+      "step": 3291
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 3.4505050505050507e-06,
+      "loss": 0.4946,
+      "step": 3292
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 3.4484848484848486e-06,
+      "loss": 0.4843,
+      "step": 3293
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 3.446464646464647e-06,
+      "loss": 0.5169,
+      "step": 3294
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 3.444444444444445e-06,
+      "loss": 0.9156,
+      "step": 3295
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 3.4424242424242427e-06,
+      "loss": 1.0062,
+      "step": 3296
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 3.4404040404040406e-06,
+      "loss": 0.7988,
+      "step": 3297
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 3.4383838383838385e-06,
+      "loss": 0.5353,
+      "step": 3298
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 3.4363636363636364e-06,
+      "loss": 0.522,
+      "step": 3299
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 3.4343434343434347e-06,
+      "loss": 0.6102,
+      "step": 3300
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 3.4323232323232326e-06,
+      "loss": 0.6063,
+      "step": 3301
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 3.4303030303030305e-06,
+      "loss": 0.3536,
+      "step": 3302
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 3.4282828282828284e-06,
+      "loss": 0.5001,
+      "step": 3303
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 3.4262626262626262e-06,
+      "loss": 0.399,
+      "step": 3304
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 3.4242424242424246e-06,
+      "loss": 0.9977,
+      "step": 3305
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 3.4222222222222224e-06,
+      "loss": 0.2503,
+      "step": 3306
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 3.4202020202020203e-06,
+      "loss": 0.6293,
+      "step": 3307
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 3.4181818181818182e-06,
+      "loss": 0.4184,
+      "step": 3308
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 3.416161616161616e-06,
+      "loss": 0.8284,
+      "step": 3309
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 3.414141414141414e-06,
+      "loss": 0.716,
+      "step": 3310
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 3.4121212121212123e-06,
+      "loss": 0.3889,
+      "step": 3311
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 3.41010101010101e-06,
+      "loss": 0.7488,
+      "step": 3312
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 3.408080808080808e-06,
+      "loss": 0.3068,
+      "step": 3313
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 3.406060606060606e-06,
+      "loss": 0.3487,
+      "step": 3314
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 3.4040404040404047e-06,
+      "loss": 1.5129,
+      "step": 3315
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 3.4020202020202026e-06,
+      "loss": 0.7321,
+      "step": 3316
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 0.4979,
+      "step": 3317
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 3.3979797979797984e-06,
+      "loss": 0.8707,
+      "step": 3318
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 3.3959595959595963e-06,
+      "loss": 0.7073,
+      "step": 3319
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 3.3939393939393946e-06,
+      "loss": 0.76,
+      "step": 3320
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 3.3919191919191925e-06,
+      "loss": 0.6248,
+      "step": 3321
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 3.3898989898989903e-06,
+      "loss": 0.4751,
+      "step": 3322
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 3.3878787878787882e-06,
+      "loss": 0.8111,
+      "step": 3323
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 3.385858585858586e-06,
+      "loss": 0.5297,
+      "step": 3324
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 3.3838383838383844e-06,
+      "loss": 0.5932,
+      "step": 3325
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 3.3818181818181823e-06,
+      "loss": 0.6421,
+      "step": 3326
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 3.37979797979798e-06,
+      "loss": 1.595,
+      "step": 3327
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 3.377777777777778e-06,
+      "loss": 0.495,
+      "step": 3328
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 3.375757575757576e-06,
+      "loss": 0.8259,
+      "step": 3329
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 3.3737373737373743e-06,
+      "loss": 1.1553,
+      "step": 3330
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 3.371717171717172e-06,
+      "loss": 0.449,
+      "step": 3331
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 3.36969696969697e-06,
+      "loss": 0.4129,
+      "step": 3332
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 3.367676767676768e-06,
+      "loss": 0.7059,
+      "step": 3333
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 3.365656565656566e-06,
+      "loss": 0.6321,
+      "step": 3334
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 3.3636363636363637e-06,
+      "loss": 0.5893,
+      "step": 3335
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 3.361616161616162e-06,
+      "loss": 0.4938,
+      "step": 3336
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 3.35959595959596e-06,
+      "loss": 0.6069,
+      "step": 3337
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 3.357575757575758e-06,
+      "loss": 0.7351,
+      "step": 3338
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 3.3555555555555557e-06,
+      "loss": 0.3503,
+      "step": 3339
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 3.3535353535353536e-06,
+      "loss": 0.3436,
+      "step": 3340
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 3.351515151515152e-06,
+      "loss": 0.3035,
+      "step": 3341
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 3.34949494949495e-06,
+      "loss": 0.2772,
+      "step": 3342
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 3.3474747474747477e-06,
+      "loss": 0.7768,
+      "step": 3343
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 3.3454545454545456e-06,
+      "loss": 0.9039,
+      "step": 3344
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 3.3434343434343435e-06,
+      "loss": 0.5484,
+      "step": 3345
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 3.3414141414141413e-06,
+      "loss": 1.5704,
+      "step": 3346
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 3.3393939393939397e-06,
+      "loss": 0.2166,
+      "step": 3347
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 3.3373737373737375e-06,
+      "loss": 0.4683,
+      "step": 3348
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 3.3353535353535354e-06,
+      "loss": 0.6145,
+      "step": 3349
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.2049,
+      "step": 3350
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 3.331313131313131e-06,
+      "loss": 0.764,
+      "step": 3351
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 3.3292929292929295e-06,
+      "loss": 0.5639,
+      "step": 3352
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 3.3272727272727274e-06,
+      "loss": 0.521,
+      "step": 3353
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 3.3252525252525253e-06,
+      "loss": 0.4437,
+      "step": 3354
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 3.323232323232323e-06,
+      "loss": 0.8766,
+      "step": 3355
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 3.321212121212121e-06,
+      "loss": 1.3494,
+      "step": 3356
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 3.3191919191919194e-06,
+      "loss": 0.602,
+      "step": 3357
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 3.3171717171717177e-06,
+      "loss": 0.5864,
+      "step": 3358
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 3.3151515151515156e-06,
+      "loss": 0.857,
+      "step": 3359
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 3.3131313131313135e-06,
+      "loss": 1.1594,
+      "step": 3360
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 3.3111111111111118e-06,
+      "loss": 1.4548,
+      "step": 3361
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 3.3090909090909097e-06,
+      "loss": 0.5439,
+      "step": 3362
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 3.3070707070707076e-06,
+      "loss": 0.354,
+      "step": 3363
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 3.3050505050505054e-06,
+      "loss": 0.7887,
+      "step": 3364
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 3.3030303030303033e-06,
+      "loss": 0.3965,
+      "step": 3365
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 3.3010101010101016e-06,
+      "loss": 0.7158,
+      "step": 3366
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 3.2989898989898995e-06,
+      "loss": 0.4147,
+      "step": 3367
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 3.2969696969696974e-06,
+      "loss": 1.8254,
+      "step": 3368
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 3.2949494949494953e-06,
+      "loss": 1.3493,
+      "step": 3369
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 3.292929292929293e-06,
+      "loss": 0.6963,
+      "step": 3370
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 3.290909090909091e-06,
+      "loss": 1.4846,
+      "step": 3371
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 3.2888888888888894e-06,
+      "loss": 0.6795,
+      "step": 3372
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 3.2868686868686873e-06,
+      "loss": 0.7332,
+      "step": 3373
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 3.284848484848485e-06,
+      "loss": 0.9128,
+      "step": 3374
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 3.282828282828283e-06,
+      "loss": 1.1983,
+      "step": 3375
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 3.280808080808081e-06,
+      "loss": 0.5619,
+      "step": 3376
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 3.2787878787878793e-06,
+      "loss": 0.8316,
+      "step": 3377
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 3.276767676767677e-06,
+      "loss": 0.2828,
+      "step": 3378
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 3.274747474747475e-06,
+      "loss": 0.6964,
+      "step": 3379
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 3.272727272727273e-06,
+      "loss": 0.5508,
+      "step": 3380
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 3.270707070707071e-06,
+      "loss": 0.7292,
+      "step": 3381
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 3.2686868686868687e-06,
+      "loss": 1.1331,
+      "step": 3382
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 3.266666666666667e-06,
+      "loss": 0.9608,
+      "step": 3383
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 3.264646464646465e-06,
+      "loss": 0.2052,
+      "step": 3384
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 3.262626262626263e-06,
+      "loss": 1.0979,
+      "step": 3385
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 3.2606060606060607e-06,
+      "loss": 0.8271,
+      "step": 3386
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 3.2585858585858586e-06,
+      "loss": 0.485,
+      "step": 3387
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 3.256565656565657e-06,
+      "loss": 0.5083,
+      "step": 3388
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 3.2545454545454548e-06,
+      "loss": 1.6217,
+      "step": 3389
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 3.2525252525252527e-06,
+      "loss": 0.5258,
+      "step": 3390
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 3.2505050505050505e-06,
+      "loss": 0.3903,
+      "step": 3391
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 3.2484848484848484e-06,
+      "loss": 0.8474,
+      "step": 3392
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 3.2464646464646467e-06,
+      "loss": 0.3899,
+      "step": 3393
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 3.2444444444444446e-06,
+      "loss": 0.5508,
+      "step": 3394
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 3.2424242424242425e-06,
+      "loss": 1.4358,
+      "step": 3395
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 3.2404040404040404e-06,
+      "loss": 0.9494,
+      "step": 3396
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 3.2383838383838383e-06,
+      "loss": 0.4852,
+      "step": 3397
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 3.236363636363636e-06,
+      "loss": 0.5731,
+      "step": 3398
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 3.2343434343434345e-06,
+      "loss": 0.4603,
+      "step": 3399
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 3.232323232323233e-06,
+      "loss": 0.6234,
+      "step": 3400
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.78,
+      "eval_loss": 0.8092541694641113,
+      "eval_roc_auc": 0.9619528853016733,
+      "eval_runtime": 92.9292,
+      "eval_samples_per_second": 2.152,
+      "eval_steps_per_second": 0.538,
+      "step": 3400
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 3.2303030303030307e-06,
+      "loss": 0.6898,
+      "step": 3401
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 3.228282828282829e-06,
+      "loss": 0.3291,
+      "step": 3402
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 3.226262626262627e-06,
+      "loss": 0.7748,
+      "step": 3403
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 3.2242424242424248e-06,
+      "loss": 0.6117,
+      "step": 3404
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 3.2222222222222227e-06,
+      "loss": 0.6382,
+      "step": 3405
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 3.2202020202020206e-06,
+      "loss": 0.6519,
+      "step": 3406
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 3.2181818181818184e-06,
+      "loss": 1.1374,
+      "step": 3407
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 3.2161616161616168e-06,
+      "loss": 0.5941,
+      "step": 3408
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 3.2141414141414146e-06,
+      "loss": 0.6524,
+      "step": 3409
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 3.2121212121212125e-06,
+      "loss": 0.5956,
+      "step": 3410
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 3.2101010101010104e-06,
+      "loss": 0.5926,
+      "step": 3411
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 3.2080808080808083e-06,
+      "loss": 0.7755,
+      "step": 3412
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 3.2060606060606066e-06,
+      "loss": 0.4037,
+      "step": 3413
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 3.2040404040404045e-06,
+      "loss": 0.3415,
+      "step": 3414
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 3.2020202020202024e-06,
+      "loss": 0.7995,
+      "step": 3415
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 0.5057,
+      "step": 3416
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 3.197979797979798e-06,
+      "loss": 0.642,
+      "step": 3417
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 3.195959595959596e-06,
+      "loss": 0.66,
+      "step": 3418
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 3.1939393939393944e-06,
+      "loss": 0.9972,
+      "step": 3419
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 3.1919191919191923e-06,
+      "loss": 0.376,
+      "step": 3420
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 3.18989898989899e-06,
+      "loss": 0.8332,
+      "step": 3421
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 3.187878787878788e-06,
+      "loss": 0.5761,
+      "step": 3422
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 3.185858585858586e-06,
+      "loss": 1.1405,
+      "step": 3423
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 3.1838383838383842e-06,
+      "loss": 1.1821,
+      "step": 3424
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 3.181818181818182e-06,
+      "loss": 0.3793,
+      "step": 3425
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 3.17979797979798e-06,
+      "loss": 0.3558,
+      "step": 3426
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 3.177777777777778e-06,
+      "loss": 0.3735,
+      "step": 3427
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 3.1757575757575758e-06,
+      "loss": 0.6904,
+      "step": 3428
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 3.173737373737374e-06,
+      "loss": 0.5626,
+      "step": 3429
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 3.171717171717172e-06,
+      "loss": 0.474,
+      "step": 3430
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 3.16969696969697e-06,
+      "loss": 1.2246,
+      "step": 3431
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 3.1676767676767678e-06,
+      "loss": 0.6826,
+      "step": 3432
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 3.1656565656565656e-06,
+      "loss": 0.4195,
+      "step": 3433
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 3.1636363636363635e-06,
+      "loss": 0.7319,
+      "step": 3434
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 3.161616161616162e-06,
+      "loss": 0.5065,
+      "step": 3435
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 3.1595959595959597e-06,
+      "loss": 0.8339,
+      "step": 3436
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 3.1575757575757576e-06,
+      "loss": 0.7255,
+      "step": 3437
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 3.1555555555555555e-06,
+      "loss": 0.6048,
+      "step": 3438
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 3.1535353535353534e-06,
+      "loss": 0.4987,
+      "step": 3439
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 3.1515151515151517e-06,
+      "loss": 0.6678,
+      "step": 3440
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 3.1494949494949496e-06,
+      "loss": 0.5078,
+      "step": 3441
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 3.1474747474747475e-06,
+      "loss": 0.8283,
+      "step": 3442
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 3.145454545454546e-06,
+      "loss": 0.2822,
+      "step": 3443
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 3.143434343434344e-06,
+      "loss": 0.9717,
+      "step": 3444
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 3.141414141414142e-06,
+      "loss": 0.4217,
+      "step": 3445
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 3.13939393939394e-06,
+      "loss": 0.6529,
+      "step": 3446
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 3.1373737373737378e-06,
+      "loss": 0.7408,
+      "step": 3447
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 3.1353535353535357e-06,
+      "loss": 0.465,
+      "step": 3448
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 3.133333333333334e-06,
+      "loss": 0.3428,
+      "step": 3449
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 3.131313131313132e-06,
+      "loss": 0.6222,
+      "step": 3450
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 3.1292929292929297e-06,
+      "loss": 0.5366,
+      "step": 3451
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 3.1272727272727276e-06,
+      "loss": 0.6008,
+      "step": 3452
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 3.1252525252525255e-06,
+      "loss": 0.6847,
+      "step": 3453
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 3.1232323232323234e-06,
+      "loss": 0.8894,
+      "step": 3454
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 3.1212121212121217e-06,
+      "loss": 0.5819,
+      "step": 3455
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 3.1191919191919196e-06,
+      "loss": 1.3308,
+      "step": 3456
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 3.1171717171717175e-06,
+      "loss": 0.7575,
+      "step": 3457
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 3.1151515151515154e-06,
+      "loss": 0.5176,
+      "step": 3458
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 3.1131313131313133e-06,
+      "loss": 1.3351,
+      "step": 3459
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 3.1111111111111116e-06,
+      "loss": 0.9897,
+      "step": 3460
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 3.1090909090909095e-06,
+      "loss": 0.585,
+      "step": 3461
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 3.1070707070707074e-06,
+      "loss": 1.0378,
+      "step": 3462
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 3.1050505050505052e-06,
+      "loss": 0.4632,
+      "step": 3463
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 3.103030303030303e-06,
+      "loss": 0.8587,
+      "step": 3464
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 3.1010101010101014e-06,
+      "loss": 0.7601,
+      "step": 3465
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 3.0989898989898993e-06,
+      "loss": 0.3937,
+      "step": 3466
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 3.0969696969696972e-06,
+      "loss": 0.9541,
+      "step": 3467
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 3.094949494949495e-06,
+      "loss": 0.7921,
+      "step": 3468
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 3.092929292929293e-06,
+      "loss": 0.3772,
+      "step": 3469
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 3.090909090909091e-06,
+      "loss": 0.2079,
+      "step": 3470
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 3.088888888888889e-06,
+      "loss": 0.9468,
+      "step": 3471
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 3.086868686868687e-06,
+      "loss": 0.3228,
+      "step": 3472
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 3.084848484848485e-06,
+      "loss": 0.7653,
+      "step": 3473
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 3.082828282828283e-06,
+      "loss": 0.6674,
+      "step": 3474
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 3.0808080808080807e-06,
+      "loss": 1.4514,
+      "step": 3475
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 3.078787878787879e-06,
+      "loss": 0.4607,
+      "step": 3476
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 3.076767676767677e-06,
+      "loss": 0.2586,
+      "step": 3477
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 3.074747474747475e-06,
+      "loss": 0.4074,
+      "step": 3478
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 3.0727272727272727e-06,
+      "loss": 0.3355,
+      "step": 3479
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 3.0707070707070706e-06,
+      "loss": 1.2539,
+      "step": 3480
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 3.068686868686869e-06,
+      "loss": 0.7855,
+      "step": 3481
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 3.066666666666667e-06,
+      "loss": 0.9537,
+      "step": 3482
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 3.0646464646464647e-06,
+      "loss": 0.7348,
+      "step": 3483
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 3.0626262626262626e-06,
+      "loss": 0.542,
+      "step": 3484
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 3.0606060606060605e-06,
+      "loss": 0.5407,
+      "step": 3485
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 3.058585858585859e-06,
+      "loss": 0.3838,
+      "step": 3486
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 3.056565656565657e-06,
+      "loss": 0.5716,
+      "step": 3487
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 3.054545454545455e-06,
+      "loss": 0.9128,
+      "step": 3488
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 3.052525252525253e-06,
+      "loss": 0.6932,
+      "step": 3489
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 3.0505050505050508e-06,
+      "loss": 0.4312,
+      "step": 3490
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 3.048484848484849e-06,
+      "loss": 0.3003,
+      "step": 3491
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 3.046464646464647e-06,
+      "loss": 1.0215,
+      "step": 3492
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 3.044444444444445e-06,
+      "loss": 0.5786,
+      "step": 3493
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 3.0424242424242427e-06,
+      "loss": 0.4094,
+      "step": 3494
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.0404040404040406e-06,
+      "loss": 0.5429,
+      "step": 3495
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.038383838383839e-06,
+      "loss": 0.4767,
+      "step": 3496
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.036363636363637e-06,
+      "loss": 0.705,
+      "step": 3497
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.0343434343434347e-06,
+      "loss": 0.3314,
+      "step": 3498
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.0323232323232326e-06,
+      "loss": 0.5397,
+      "step": 3499
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.0303030303030305e-06,
+      "loss": 0.5644,
+      "step": 3500
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.028282828282829e-06,
+      "loss": 1.087,
+      "step": 3501
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.0262626262626267e-06,
+      "loss": 0.3647,
+      "step": 3502
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.0242424242424246e-06,
+      "loss": 0.5947,
+      "step": 3503
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.0222222222222225e-06,
+      "loss": 1.2008,
+      "step": 3504
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 3.0202020202020203e-06,
+      "loss": 0.4396,
+      "step": 3505
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.0181818181818182e-06,
+      "loss": 0.4272,
+      "step": 3506
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.0161616161616165e-06,
+      "loss": 0.7361,
+      "step": 3507
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.0141414141414144e-06,
+      "loss": 0.8305,
+      "step": 3508
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.0121212121212123e-06,
+      "loss": 0.8621,
+      "step": 3509
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.0101010101010102e-06,
+      "loss": 1.0261,
+      "step": 3510
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.008080808080808e-06,
+      "loss": 0.8703,
+      "step": 3511
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.0060606060606064e-06,
+      "loss": 0.166,
+      "step": 3512
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.0040404040404043e-06,
+      "loss": 0.2457,
+      "step": 3513
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3.002020202020202e-06,
+      "loss": 0.4123,
+      "step": 3514
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 3e-06,
+      "loss": 1.2002,
+      "step": 3515
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 2.997979797979798e-06,
+      "loss": 0.7694,
+      "step": 3516
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 2.9959595959595963e-06,
+      "loss": 0.7002,
+      "step": 3517
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 2.993939393939394e-06,
+      "loss": 0.2083,
+      "step": 3518
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 2.991919191919192e-06,
+      "loss": 0.2762,
+      "step": 3519
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 2.98989898989899e-06,
+      "loss": 0.5192,
+      "step": 3520
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 2.987878787878788e-06,
+      "loss": 0.7283,
+      "step": 3521
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 2.9858585858585857e-06,
+      "loss": 0.2346,
+      "step": 3522
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 2.983838383838384e-06,
+      "loss": 0.2018,
+      "step": 3523
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 2.981818181818182e-06,
+      "loss": 0.6698,
+      "step": 3524
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 2.97979797979798e-06,
+      "loss": 0.4897,
+      "step": 3525
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 2.9777777777777777e-06,
+      "loss": 0.8744,
+      "step": 3526
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 2.9757575757575756e-06,
+      "loss": 0.3092,
+      "step": 3527
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 2.9737373737373743e-06,
+      "loss": 0.9334,
+      "step": 3528
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 2.971717171717172e-06,
+      "loss": 0.7362,
+      "step": 3529
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 2.96969696969697e-06,
+      "loss": 0.5564,
+      "step": 3530
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 2.967676767676768e-06,
+      "loss": 0.5574,
+      "step": 3531
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 2.9656565656565663e-06,
+      "loss": 1.2218,
+      "step": 3532
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 2.963636363636364e-06,
+      "loss": 0.3074,
+      "step": 3533
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 2.961616161616162e-06,
+      "loss": 0.8404,
+      "step": 3534
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 2.95959595959596e-06,
+      "loss": 1.4701,
+      "step": 3535
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 2.957575757575758e-06,
+      "loss": 1.0738,
+      "step": 3536
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 2.955555555555556e-06,
+      "loss": 0.4812,
+      "step": 3537
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 2.953535353535354e-06,
+      "loss": 0.8012,
+      "step": 3538
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 2.951515151515152e-06,
+      "loss": 0.7886,
+      "step": 3539
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 2.94949494949495e-06,
+      "loss": 0.378,
+      "step": 3540
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 2.9474747474747477e-06,
+      "loss": 0.6154,
+      "step": 3541
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 2.9454545454545456e-06,
+      "loss": 0.7536,
+      "step": 3542
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 2.943434343434344e-06,
+      "loss": 0.9649,
+      "step": 3543
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 2.941414141414142e-06,
+      "loss": 0.397,
+      "step": 3544
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 2.9393939393939397e-06,
+      "loss": 0.5851,
+      "step": 3545
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 2.9373737373737376e-06,
+      "loss": 1.3321,
+      "step": 3546
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 2.9353535353535355e-06,
+      "loss": 0.3926,
+      "step": 3547
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 2.9333333333333338e-06,
+      "loss": 0.5543,
+      "step": 3548
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 2.9313131313131317e-06,
+      "loss": 0.8428,
+      "step": 3549
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 2.9292929292929295e-06,
+      "loss": 0.3816,
+      "step": 3550
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 2.9272727272727274e-06,
+      "loss": 0.5727,
+      "step": 3551
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 2.9252525252525253e-06,
+      "loss": 0.2715,
+      "step": 3552
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 2.9232323232323236e-06,
+      "loss": 1.152,
+      "step": 3553
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 2.9212121212121215e-06,
+      "loss": 0.6979,
+      "step": 3554
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 2.9191919191919194e-06,
+      "loss": 0.5355,
+      "step": 3555
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 2.9171717171717173e-06,
+      "loss": 1.2737,
+      "step": 3556
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 2.915151515151515e-06,
+      "loss": 0.8386,
+      "step": 3557
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 2.913131313131313e-06,
+      "loss": 1.4232,
+      "step": 3558
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 2.9111111111111114e-06,
+      "loss": 0.4562,
+      "step": 3559
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 2.9090909090909093e-06,
+      "loss": 0.5188,
+      "step": 3560
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 2.907070707070707e-06,
+      "loss": 0.4614,
+      "step": 3561
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 2.905050505050505e-06,
+      "loss": 0.6016,
+      "step": 3562
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 2.903030303030303e-06,
+      "loss": 1.0225,
+      "step": 3563
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 2.9010101010101012e-06,
+      "loss": 0.3527,
+      "step": 3564
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 2.898989898989899e-06,
+      "loss": 0.7018,
+      "step": 3565
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 2.896969696969697e-06,
+      "loss": 1.3236,
+      "step": 3566
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 2.894949494949495e-06,
+      "loss": 0.8669,
+      "step": 3567
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 2.892929292929293e-06,
+      "loss": 0.2649,
+      "step": 3568
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 2.8909090909090907e-06,
+      "loss": 0.5386,
+      "step": 3569
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 2.888888888888889e-06,
+      "loss": 0.6213,
+      "step": 3570
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 2.8868686868686873e-06,
+      "loss": 0.9863,
+      "step": 3571
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 2.884848484848485e-06,
+      "loss": 0.4481,
+      "step": 3572
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 2.8828282828282835e-06,
+      "loss": 0.2504,
+      "step": 3573
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 2.8808080808080814e-06,
+      "loss": 0.8909,
+      "step": 3574
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 2.8787878787878793e-06,
+      "loss": 0.476,
+      "step": 3575
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 2.876767676767677e-06,
+      "loss": 0.5931,
+      "step": 3576
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 2.874747474747475e-06,
+      "loss": 0.4161,
+      "step": 3577
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 2.872727272727273e-06,
+      "loss": 1.599,
+      "step": 3578
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 2.8707070707070713e-06,
+      "loss": 0.8308,
+      "step": 3579
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 2.868686868686869e-06,
+      "loss": 0.9082,
+      "step": 3580
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 2.866666666666667e-06,
+      "loss": 0.9991,
+      "step": 3581
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 2.864646464646465e-06,
+      "loss": 0.3604,
+      "step": 3582
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 2.862626262626263e-06,
+      "loss": 1.0824,
+      "step": 3583
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 2.860606060606061e-06,
+      "loss": 1.2624,
+      "step": 3584
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 2.858585858585859e-06,
+      "loss": 0.2705,
+      "step": 3585
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 2.856565656565657e-06,
+      "loss": 0.4494,
+      "step": 3586
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 2.8545454545454548e-06,
+      "loss": 1.0209,
+      "step": 3587
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 2.8525252525252527e-06,
+      "loss": 0.8112,
+      "step": 3588
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 2.850505050505051e-06,
+      "loss": 0.4951,
+      "step": 3589
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 2.848484848484849e-06,
+      "loss": 1.3303,
+      "step": 3590
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 2.8464646464646468e-06,
+      "loss": 0.8324,
+      "step": 3591
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 2.8444444444444446e-06,
+      "loss": 0.3665,
+      "step": 3592
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 2.8424242424242425e-06,
+      "loss": 0.9034,
+      "step": 3593
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 2.8404040404040404e-06,
+      "loss": 0.6904,
+      "step": 3594
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 2.8383838383838387e-06,
+      "loss": 0.9416,
+      "step": 3595
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 2.8363636363636366e-06,
+      "loss": 0.5245,
+      "step": 3596
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 2.8343434343434345e-06,
+      "loss": 0.6023,
+      "step": 3597
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 2.8323232323232324e-06,
+      "loss": 0.6031,
+      "step": 3598
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 2.8303030303030303e-06,
+      "loss": 1.3709,
+      "step": 3599
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 2.8282828282828286e-06,
+      "loss": 0.8303,
+      "step": 3600
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.785,
+      "eval_loss": 0.7793353199958801,
+      "eval_roc_auc": 0.9567147725192919,
+      "eval_runtime": 93.2329,
+      "eval_samples_per_second": 2.145,
+      "eval_steps_per_second": 0.536,
+      "step": 3600
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 2.8262626262626265e-06,
+      "loss": 0.8691,
+      "step": 3601
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 2.8242424242424244e-06,
+      "loss": 0.4204,
+      "step": 3602
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 2.8222222222222223e-06,
+      "loss": 1.6602,
+      "step": 3603
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 2.82020202020202e-06,
+      "loss": 0.423,
+      "step": 3604
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 2.818181818181818e-06,
+      "loss": 0.2859,
+      "step": 3605
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 2.8161616161616163e-06,
+      "loss": 0.6017,
+      "step": 3606
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 2.8141414141414142e-06,
+      "loss": 0.3448,
+      "step": 3607
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 2.812121212121212e-06,
+      "loss": 0.7353,
+      "step": 3608
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 2.81010101010101e-06,
+      "loss": 0.3195,
+      "step": 3609
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 2.808080808080808e-06,
+      "loss": 0.503,
+      "step": 3610
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 2.806060606060606e-06,
+      "loss": 0.5281,
+      "step": 3611
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 2.804040404040404e-06,
+      "loss": 0.8049,
+      "step": 3612
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 2.802020202020202e-06,
+      "loss": 0.3267,
+      "step": 3613
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 1.1266,
+      "step": 3614
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 2.7979797979797986e-06,
+      "loss": 1.1279,
+      "step": 3615
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 2.7959595959595965e-06,
+      "loss": 0.4058,
+      "step": 3616
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 2.7939393939393944e-06,
+      "loss": 0.1955,
+      "step": 3617
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 2.7919191919191923e-06,
+      "loss": 0.3412,
+      "step": 3618
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 2.78989898989899e-06,
+      "loss": 1.1622,
+      "step": 3619
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 2.7878787878787885e-06,
+      "loss": 0.8658,
+      "step": 3620
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.7858585858585864e-06,
+      "loss": 0.7068,
+      "step": 3621
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.7838383838383842e-06,
+      "loss": 0.9178,
+      "step": 3622
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.781818181818182e-06,
+      "loss": 0.5934,
+      "step": 3623
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.77979797979798e-06,
+      "loss": 0.9291,
+      "step": 3624
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.7777777777777783e-06,
+      "loss": 0.3682,
+      "step": 3625
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.7757575757575762e-06,
+      "loss": 1.2031,
+      "step": 3626
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.773737373737374e-06,
+      "loss": 0.4855,
+      "step": 3627
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.771717171717172e-06,
+      "loss": 0.5493,
+      "step": 3628
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.76969696969697e-06,
+      "loss": 0.2443,
+      "step": 3629
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.7676767676767678e-06,
+      "loss": 0.4706,
+      "step": 3630
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.765656565656566e-06,
+      "loss": 1.2191,
+      "step": 3631
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.763636363636364e-06,
+      "loss": 0.8581,
+      "step": 3632
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.761616161616162e-06,
+      "loss": 0.5199,
+      "step": 3633
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.7595959595959597e-06,
+      "loss": 0.7535,
+      "step": 3634
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.7575757575757576e-06,
+      "loss": 1.1546,
+      "step": 3635
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.755555555555556e-06,
+      "loss": 0.6098,
+      "step": 3636
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.753535353535354e-06,
+      "loss": 0.9244,
+      "step": 3637
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.7515151515151517e-06,
+      "loss": 1.3451,
+      "step": 3638
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.7494949494949496e-06,
+      "loss": 0.4246,
+      "step": 3639
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.7474747474747475e-06,
+      "loss": 0.5924,
+      "step": 3640
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.7454545454545454e-06,
+      "loss": 1.5495,
+      "step": 3641
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.7434343434343437e-06,
+      "loss": 0.6627,
+      "step": 3642
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.7414141414141416e-06,
+      "loss": 0.5644,
+      "step": 3643
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.7393939393939395e-06,
+      "loss": 1.0132,
+      "step": 3644
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.7373737373737374e-06,
+      "loss": 0.6295,
+      "step": 3645
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.7353535353535353e-06,
+      "loss": 0.6051,
+      "step": 3646
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.7333333333333336e-06,
+      "loss": 0.3085,
+      "step": 3647
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7313131313131315e-06,
+      "loss": 0.3563,
+      "step": 3648
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.7292929292929293e-06,
+      "loss": 0.2519,
+      "step": 3649
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.7272727272727272e-06,
+      "loss": 1.1749,
+      "step": 3650
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.725252525252525e-06,
+      "loss": 0.8466,
+      "step": 3651
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.7232323232323234e-06,
+      "loss": 0.5042,
+      "step": 3652
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.7212121212121213e-06,
+      "loss": 0.5372,
+      "step": 3653
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.719191919191919e-06,
+      "loss": 0.3481,
+      "step": 3654
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.717171717171717e-06,
+      "loss": 0.5254,
+      "step": 3655
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.715151515151516e-06,
+      "loss": 0.7385,
+      "step": 3656
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7131313131313137e-06,
+      "loss": 1.0092,
+      "step": 3657
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.7111111111111116e-06,
+      "loss": 0.826,
+      "step": 3658
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.7090909090909095e-06,
+      "loss": 0.6298,
+      "step": 3659
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.7070707070707074e-06,
+      "loss": 0.6278,
+      "step": 3660
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.7050505050505057e-06,
+      "loss": 0.6093,
+      "step": 3661
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.7030303030303036e-06,
+      "loss": 0.335,
+      "step": 3662
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.7010101010101015e-06,
+      "loss": 1.0388,
+      "step": 3663
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6989898989898994e-06,
+      "loss": 0.9009,
+      "step": 3664
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6969696969696972e-06,
+      "loss": 0.6026,
+      "step": 3665
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.694949494949495e-06,
+      "loss": 1.0052,
+      "step": 3666
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6929292929292934e-06,
+      "loss": 0.4841,
+      "step": 3667
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6909090909090913e-06,
+      "loss": 0.7379,
+      "step": 3668
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.6888888888888892e-06,
+      "loss": 0.4157,
+      "step": 3669
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.686868686868687e-06,
+      "loss": 0.8727,
+      "step": 3670
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.684848484848485e-06,
+      "loss": 0.3266,
+      "step": 3671
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.6828282828282833e-06,
+      "loss": 0.4001,
+      "step": 3672
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 2.680808080808081e-06,
+      "loss": 0.372,
+      "step": 3673
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.678787878787879e-06,
+      "loss": 0.7121,
+      "step": 3674
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.676767676767677e-06,
+      "loss": 0.4774,
+      "step": 3675
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.674747474747475e-06,
+      "loss": 0.8992,
+      "step": 3676
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.6727272727272727e-06,
+      "loss": 0.6487,
+      "step": 3677
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.670707070707071e-06,
+      "loss": 0.616,
+      "step": 3678
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.668686868686869e-06,
+      "loss": 0.9227,
+      "step": 3679
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.666666666666667e-06,
+      "loss": 0.4262,
+      "step": 3680
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.6646464646464647e-06,
+      "loss": 0.4162,
+      "step": 3681
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.6626262626262626e-06,
+      "loss": 0.3364,
+      "step": 3682
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.660606060606061e-06,
+      "loss": 0.3625,
+      "step": 3683
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.658585858585859e-06,
+      "loss": 0.4257,
+      "step": 3684
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.6565656565656567e-06,
+      "loss": 1.0795,
+      "step": 3685
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.6545454545454546e-06,
+      "loss": 0.7391,
+      "step": 3686
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.6525252525252525e-06,
+      "loss": 0.4412,
+      "step": 3687
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.6505050505050508e-06,
+      "loss": 0.6266,
+      "step": 3688
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.6484848484848487e-06,
+      "loss": 0.7201,
+      "step": 3689
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.6464646464646466e-06,
+      "loss": 0.455,
+      "step": 3690
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2.6444444444444444e-06,
+      "loss": 0.312,
+      "step": 3691
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.6424242424242423e-06,
+      "loss": 0.3962,
+      "step": 3692
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.6404040404040402e-06,
+      "loss": 0.5605,
+      "step": 3693
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.6383838383838385e-06,
+      "loss": 0.5133,
+      "step": 3694
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.6363636363636364e-06,
+      "loss": 0.3888,
+      "step": 3695
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.6343434343434343e-06,
+      "loss": 0.2172,
+      "step": 3696
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.632323232323232e-06,
+      "loss": 0.364,
+      "step": 3697
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.63030303030303e-06,
+      "loss": 0.5265,
+      "step": 3698
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.628282828282829e-06,
+      "loss": 1.072,
+      "step": 3699
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.6262626262626267e-06,
+      "loss": 0.3592,
+      "step": 3700
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2.6242424242424246e-06,
+      "loss": 0.1924,
+      "step": 3701
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.6222222222222225e-06,
+      "loss": 0.9711,
+      "step": 3702
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.620202020202021e-06,
+      "loss": 0.6271,
+      "step": 3703
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.6181818181818187e-06,
+      "loss": 0.35,
+      "step": 3704
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.6161616161616166e-06,
+      "loss": 1.0908,
+      "step": 3705
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.6141414141414145e-06,
+      "loss": 0.2738,
+      "step": 3706
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.6121212121212123e-06,
+      "loss": 0.2218,
+      "step": 3707
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 2.6101010101010107e-06,
+      "loss": 0.8147,
+      "step": 3708
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.6080808080808085e-06,
+      "loss": 0.3755,
+      "step": 3709
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.6060606060606064e-06,
+      "loss": 0.8751,
+      "step": 3710
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.6040404040404043e-06,
+      "loss": 0.3409,
+      "step": 3711
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.602020202020202e-06,
+      "loss": 0.4684,
+      "step": 3712
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.6e-06,
+      "loss": 0.5038,
+      "step": 3713
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.5979797979797984e-06,
+      "loss": 0.396,
+      "step": 3714
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.5959595959595963e-06,
+      "loss": 0.3894,
+      "step": 3715
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.593939393939394e-06,
+      "loss": 0.4088,
+      "step": 3716
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.591919191919192e-06,
+      "loss": 0.46,
+      "step": 3717
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.58989898989899e-06,
+      "loss": 1.0413,
+      "step": 3718
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 2.5878787878787883e-06,
+      "loss": 0.6198,
+      "step": 3719
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.585858585858586e-06,
+      "loss": 0.641,
+      "step": 3720
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.583838383838384e-06,
+      "loss": 0.2592,
+      "step": 3721
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.581818181818182e-06,
+      "loss": 0.9484,
+      "step": 3722
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.57979797979798e-06,
+      "loss": 1.2304,
+      "step": 3723
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.577777777777778e-06,
+      "loss": 0.4145,
+      "step": 3724
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.575757575757576e-06,
+      "loss": 0.6601,
+      "step": 3725
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.573737373737374e-06,
+      "loss": 0.9319,
+      "step": 3726
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.571717171717172e-06,
+      "loss": 0.8274,
+      "step": 3727
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.5696969696969697e-06,
+      "loss": 0.6566,
+      "step": 3728
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.5676767676767676e-06,
+      "loss": 0.3293,
+      "step": 3729
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.565656565656566e-06,
+      "loss": 0.708,
+      "step": 3730
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.5636363636363638e-06,
+      "loss": 1.191,
+      "step": 3731
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.5616161616161617e-06,
+      "loss": 0.2792,
+      "step": 3732
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.5595959595959595e-06,
+      "loss": 0.5994,
+      "step": 3733
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.5575757575757574e-06,
+      "loss": 0.8684,
+      "step": 3734
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.5555555555555557e-06,
+      "loss": 0.8675,
+      "step": 3735
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.5535353535353536e-06,
+      "loss": 0.4032,
+      "step": 3736
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 2.5515151515151515e-06,
+      "loss": 0.2315,
+      "step": 3737
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.5494949494949494e-06,
+      "loss": 0.3207,
+      "step": 3738
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.5474747474747473e-06,
+      "loss": 0.7331,
+      "step": 3739
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.5454545454545456e-06,
+      "loss": 0.2914,
+      "step": 3740
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.5434343434343435e-06,
+      "loss": 0.3985,
+      "step": 3741
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.541414141414142e-06,
+      "loss": 0.251,
+      "step": 3742
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.5393939393939397e-06,
+      "loss": 0.5068,
+      "step": 3743
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 2.537373737373738e-06,
+      "loss": 0.3378,
+      "step": 3744
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.535353535353536e-06,
+      "loss": 0.2173,
+      "step": 3745
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.5333333333333338e-06,
+      "loss": 0.3845,
+      "step": 3746
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.5313131313131317e-06,
+      "loss": 1.0372,
+      "step": 3747
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 2.5292929292929296e-06,
+      "loss": 0.456,
+      "step": 3748
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 2.5272727272727274e-06,
+      "loss": 1.1358,
+      "step": 3749
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 2.5252525252525258e-06,
+      "loss": 1.5742,
+      "step": 3750
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 2.5232323232323236e-06,
+      "loss": 0.1792,
+      "step": 3751
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 2.5212121212121215e-06,
+      "loss": 0.7467,
+      "step": 3752
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 2.5191919191919194e-06,
+      "loss": 0.9133,
+      "step": 3753
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 2.5171717171717173e-06,
+      "loss": 0.1333,
+      "step": 3754
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 2.5151515151515156e-06,
+      "loss": 0.6683,
+      "step": 3755
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 2.5131313131313135e-06,
+      "loss": 1.9652,
+      "step": 3756
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 2.5111111111111114e-06,
+      "loss": 0.5227,
+      "step": 3757
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 2.5090909090909093e-06,
+      "loss": 0.7593,
+      "step": 3758
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 2.507070707070707e-06,
+      "loss": 0.404,
+      "step": 3759
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 2.5050505050505055e-06,
+      "loss": 1.2615,
+      "step": 3760
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 2.5030303030303034e-06,
+      "loss": 0.4767,
+      "step": 3761
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2.5010101010101013e-06,
+      "loss": 0.4134,
+      "step": 3762
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.498989898989899e-06,
+      "loss": 0.4395,
+      "step": 3763
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.496969696969697e-06,
+      "loss": 0.683,
+      "step": 3764
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.494949494949495e-06,
+      "loss": 0.2596,
+      "step": 3765
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.4929292929292932e-06,
+      "loss": 0.9733,
+      "step": 3766
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.490909090909091e-06,
+      "loss": 1.0351,
+      "step": 3767
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.488888888888889e-06,
+      "loss": 0.4821,
+      "step": 3768
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.486868686868687e-06,
+      "loss": 0.4218,
+      "step": 3769
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 2.4848484848484848e-06,
+      "loss": 0.3198,
+      "step": 3770
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.482828282828283e-06,
+      "loss": 0.4945,
+      "step": 3771
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.480808080808081e-06,
+      "loss": 0.7672,
+      "step": 3772
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.478787878787879e-06,
+      "loss": 0.3466,
+      "step": 3773
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.476767676767677e-06,
+      "loss": 0.3199,
+      "step": 3774
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.474747474747475e-06,
+      "loss": 0.5912,
+      "step": 3775
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.472727272727273e-06,
+      "loss": 0.6246,
+      "step": 3776
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.470707070707071e-06,
+      "loss": 0.726,
+      "step": 3777
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.468686868686869e-06,
+      "loss": 0.3101,
+      "step": 3778
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.466666666666667e-06,
+      "loss": 0.5479,
+      "step": 3779
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.464646464646465e-06,
+      "loss": 0.5378,
+      "step": 3780
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.462626262626263e-06,
+      "loss": 1.0535,
+      "step": 3781
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.4606060606060607e-06,
+      "loss": 0.6278,
+      "step": 3782
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.4585858585858586e-06,
+      "loss": 0.5069,
+      "step": 3783
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 2.456565656565657e-06,
+      "loss": 1.0597,
+      "step": 3784
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.454545454545455e-06,
+      "loss": 0.7783,
+      "step": 3785
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.4525252525252527e-06,
+      "loss": 0.8961,
+      "step": 3786
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.4505050505050506e-06,
+      "loss": 0.363,
+      "step": 3787
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 2.4484848484848485e-06,
+      "loss": 0.5885,
+      "step": 3788
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.4464646464646468e-06,
+      "loss": 0.7647,
+      "step": 3789
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.4444444444444447e-06,
+      "loss": 1.3194,
+      "step": 3790
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.4424242424242426e-06,
+      "loss": 0.2025,
+      "step": 3791
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.4404040404040404e-06,
+      "loss": 0.4053,
+      "step": 3792
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.4383838383838383e-06,
+      "loss": 0.3768,
+      "step": 3793
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.4363636363636366e-06,
+      "loss": 0.8332,
+      "step": 3794
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.4343434343434345e-06,
+      "loss": 1.0379,
+      "step": 3795
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.432323232323233e-06,
+      "loss": 0.6505,
+      "step": 3796
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.4303030303030307e-06,
+      "loss": 0.3587,
+      "step": 3797
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 2.4282828282828286e-06,
+      "loss": 0.4294,
+      "step": 3798
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.4262626262626265e-06,
+      "loss": 0.9064,
+      "step": 3799
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.4242424242424244e-06,
+      "loss": 0.426,
+      "step": 3800
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.82,
+      "eval_loss": 0.7321453094482422,
+      "eval_roc_auc": 0.9654516924811125,
+      "eval_runtime": 93.5199,
+      "eval_samples_per_second": 2.139,
+      "eval_steps_per_second": 0.535,
+      "step": 3800
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.4222222222222223e-06,
+      "loss": 0.3827,
+      "step": 3801
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 2.4202020202020206e-06,
+      "loss": 0.4901,
+      "step": 3802
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.4181818181818185e-06,
+      "loss": 0.2912,
+      "step": 3803
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.4161616161616164e-06,
+      "loss": 0.5105,
+      "step": 3804
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.4141414141414143e-06,
+      "loss": 1.1106,
+      "step": 3805
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.412121212121212e-06,
+      "loss": 0.6889,
+      "step": 3806
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.4101010101010105e-06,
+      "loss": 0.6305,
+      "step": 3807
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.4080808080808083e-06,
+      "loss": 0.4063,
+      "step": 3808
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.4060606060606062e-06,
+      "loss": 0.5017,
+      "step": 3809
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.404040404040404e-06,
+      "loss": 0.6665,
+      "step": 3810
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.402020202020202e-06,
+      "loss": 1.5779,
+      "step": 3811
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 0.3431,
+      "step": 3812
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.397979797979798e-06,
+      "loss": 1.0621,
+      "step": 3813
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.395959595959596e-06,
+      "loss": 0.4796,
+      "step": 3814
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.393939393939394e-06,
+      "loss": 0.5211,
+      "step": 3815
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.3919191919191923e-06,
+      "loss": 1.9562,
+      "step": 3816
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.38989898989899e-06,
+      "loss": 0.4815,
+      "step": 3817
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.387878787878788e-06,
+      "loss": 0.359,
+      "step": 3818
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.385858585858586e-06,
+      "loss": 0.9178,
+      "step": 3819
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.3838383838383843e-06,
+      "loss": 0.5138,
+      "step": 3820
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.381818181818182e-06,
+      "loss": 0.5769,
+      "step": 3821
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.37979797979798e-06,
+      "loss": 0.2449,
+      "step": 3822
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.377777777777778e-06,
+      "loss": 0.481,
+      "step": 3823
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.375757575757576e-06,
+      "loss": 1.5756,
+      "step": 3824
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.373737373737374e-06,
+      "loss": 0.4373,
+      "step": 3825
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.371717171717172e-06,
+      "loss": 0.9158,
+      "step": 3826
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.36969696969697e-06,
+      "loss": 1.0413,
+      "step": 3827
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.367676767676768e-06,
+      "loss": 0.6136,
+      "step": 3828
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.3656565656565657e-06,
+      "loss": 0.4319,
+      "step": 3829
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.363636363636364e-06,
+      "loss": 1.1271,
+      "step": 3830
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.361616161616162e-06,
+      "loss": 1.4792,
+      "step": 3831
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.3595959595959598e-06,
+      "loss": 0.6111,
+      "step": 3832
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.3575757575757577e-06,
+      "loss": 0.5009,
+      "step": 3833
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.3555555555555555e-06,
+      "loss": 0.5536,
+      "step": 3834
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.3535353535353534e-06,
+      "loss": 0.5361,
+      "step": 3835
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.3515151515151517e-06,
+      "loss": 0.7881,
+      "step": 3836
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.3494949494949496e-06,
+      "loss": 0.1466,
+      "step": 3837
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.347474747474748e-06,
+      "loss": 0.3065,
+      "step": 3838
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2.345454545454546e-06,
+      "loss": 0.9069,
+      "step": 3839
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2.3434343434343437e-06,
+      "loss": 1.0743,
+      "step": 3840
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2.3414141414141416e-06,
+      "loss": 0.4773,
+      "step": 3841
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 2.3393939393939395e-06,
+      "loss": 1.2687,
+      "step": 3842
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 2.337373737373738e-06,
+      "loss": 0.4763,
+      "step": 3843
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 2.3353535353535357e-06,
+      "loss": 0.7353,
+      "step": 3844
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 2.3333333333333336e-06,
+      "loss": 0.4099,
+      "step": 3845
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 2.3313131313131315e-06,
+      "loss": 0.1803,
+      "step": 3846
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 2.3292929292929294e-06,
+      "loss": 0.3844,
+      "step": 3847
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 2.3272727272727277e-06,
+      "loss": 0.4182,
+      "step": 3848
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 2.3252525252525256e-06,
+      "loss": 1.0519,
+      "step": 3849
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 2.3232323232323234e-06,
+      "loss": 0.7171,
+      "step": 3850
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 2.3212121212121213e-06,
+      "loss": 0.5792,
+      "step": 3851
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 2.3191919191919192e-06,
+      "loss": 0.5835,
+      "step": 3852
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 2.317171717171717e-06,
+      "loss": 0.4475,
+      "step": 3853
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 2.3151515151515154e-06,
+      "loss": 0.8411,
+      "step": 3854
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 2.3131313131313133e-06,
+      "loss": 0.3955,
+      "step": 3855
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 2.311111111111111e-06,
+      "loss": 0.4221,
+      "step": 3856
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 2.309090909090909e-06,
+      "loss": 0.6936,
+      "step": 3857
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 2.307070707070707e-06,
+      "loss": 0.3191,
+      "step": 3858
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 2.3050505050505053e-06,
+      "loss": 0.3869,
+      "step": 3859
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 2.303030303030303e-06,
+      "loss": 0.5269,
+      "step": 3860
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 2.3010101010101015e-06,
+      "loss": 0.3292,
+      "step": 3861
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 2.2989898989898994e-06,
+      "loss": 0.509,
+      "step": 3862
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 2.2969696969696973e-06,
+      "loss": 0.8789,
+      "step": 3863
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 2.294949494949495e-06,
+      "loss": 0.8712,
+      "step": 3864
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 2.292929292929293e-06,
+      "loss": 0.8949,
+      "step": 3865
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 2.2909090909090913e-06,
+      "loss": 0.5447,
+      "step": 3866
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 2.2888888888888892e-06,
+      "loss": 1.4459,
+      "step": 3867
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 2.286868686868687e-06,
+      "loss": 0.7459,
+      "step": 3868
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 2.284848484848485e-06,
+      "loss": 0.3726,
+      "step": 3869
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 2.282828282828283e-06,
+      "loss": 0.3106,
+      "step": 3870
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 2.2808080808080808e-06,
+      "loss": 0.1518,
+      "step": 3871
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 2.278787878787879e-06,
+      "loss": 0.2931,
+      "step": 3872
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 2.276767676767677e-06,
+      "loss": 0.4574,
+      "step": 3873
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 2.274747474747475e-06,
+      "loss": 0.581,
+      "step": 3874
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 2.2727272727272728e-06,
+      "loss": 0.3165,
+      "step": 3875
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 2.2707070707070706e-06,
+      "loss": 0.5246,
+      "step": 3876
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 2.268686868686869e-06,
+      "loss": 1.1057,
+      "step": 3877
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 2.266666666666667e-06,
+      "loss": 0.394,
+      "step": 3878
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 2.2646464646464647e-06,
+      "loss": 0.6346,
+      "step": 3879
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 2.262626262626263e-06,
+      "loss": 0.6873,
+      "step": 3880
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 2.260606060606061e-06,
+      "loss": 0.1861,
+      "step": 3881
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 2.258585858585859e-06,
+      "loss": 0.5107,
+      "step": 3882
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 2.2565656565656567e-06,
+      "loss": 0.6119,
+      "step": 3883
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 2.254545454545455e-06,
+      "loss": 0.5626,
+      "step": 3884
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 2.252525252525253e-06,
+      "loss": 0.7417,
+      "step": 3885
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 2.250505050505051e-06,
+      "loss": 0.6109,
+      "step": 3886
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 2.2484848484848487e-06,
+      "loss": 0.4653,
+      "step": 3887
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 2.2464646464646466e-06,
+      "loss": 0.357,
+      "step": 3888
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 2.2444444444444445e-06,
+      "loss": 0.9818,
+      "step": 3889
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 2.2424242424242428e-06,
+      "loss": 1.2896,
+      "step": 3890
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 2.2404040404040407e-06,
+      "loss": 0.6894,
+      "step": 3891
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 2.2383838383838385e-06,
+      "loss": 1.2664,
+      "step": 3892
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 2.2363636363636364e-06,
+      "loss": 0.1893,
+      "step": 3893
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 2.2343434343434343e-06,
+      "loss": 0.3538,
+      "step": 3894
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 2.2323232323232326e-06,
+      "loss": 0.4863,
+      "step": 3895
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 2.2303030303030305e-06,
+      "loss": 1.4254,
+      "step": 3896
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 2.2282828282828284e-06,
+      "loss": 0.2798,
+      "step": 3897
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 2.2262626262626263e-06,
+      "loss": 0.4982,
+      "step": 3898
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 2.224242424242424e-06,
+      "loss": 0.9345,
+      "step": 3899
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 0.3109,
+      "step": 3900
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 2.2202020202020204e-06,
+      "loss": 0.6519,
+      "step": 3901
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 2.2181818181818187e-06,
+      "loss": 0.6873,
+      "step": 3902
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 2.2161616161616166e-06,
+      "loss": 0.546,
+      "step": 3903
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 2.2141414141414145e-06,
+      "loss": 0.3672,
+      "step": 3904
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 2.2121212121212124e-06,
+      "loss": 0.3729,
+      "step": 3905
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 2.2101010101010102e-06,
+      "loss": 0.2637,
+      "step": 3906
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 2.208080808080808e-06,
+      "loss": 0.5435,
+      "step": 3907
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 2.2060606060606064e-06,
+      "loss": 1.0392,
+      "step": 3908
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 2.2040404040404043e-06,
+      "loss": 0.9863,
+      "step": 3909
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 2.2020202020202022e-06,
+      "loss": 0.1932,
+      "step": 3910
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 2.2e-06,
+      "loss": 0.6432,
+      "step": 3911
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 2.197979797979798e-06,
+      "loss": 0.3163,
+      "step": 3912
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 2.1959595959595963e-06,
+      "loss": 0.8869,
+      "step": 3913
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 2.193939393939394e-06,
+      "loss": 0.3645,
+      "step": 3914
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 2.191919191919192e-06,
+      "loss": 0.5489,
+      "step": 3915
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 2.18989898989899e-06,
+      "loss": 0.952,
+      "step": 3916
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 2.187878787878788e-06,
+      "loss": 0.3035,
+      "step": 3917
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 2.1858585858585858e-06,
+      "loss": 0.645,
+      "step": 3918
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 2.183838383838384e-06,
+      "loss": 0.1608,
+      "step": 3919
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 2.181818181818182e-06,
+      "loss": 0.5926,
+      "step": 3920
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 2.17979797979798e-06,
+      "loss": 0.368,
+      "step": 3921
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 2.1777777777777777e-06,
+      "loss": 0.4497,
+      "step": 3922
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 2.175757575757576e-06,
+      "loss": 0.2977,
+      "step": 3923
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 2.173737373737374e-06,
+      "loss": 1.1758,
+      "step": 3924
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 2.171717171717172e-06,
+      "loss": 0.4536,
+      "step": 3925
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 2.16969696969697e-06,
+      "loss": 0.8472,
+      "step": 3926
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 2.167676767676768e-06,
+      "loss": 0.3612,
+      "step": 3927
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 2.165656565656566e-06,
+      "loss": 1.4498,
+      "step": 3928
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 2.163636363636364e-06,
+      "loss": 1.2167,
+      "step": 3929
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 2.1616161616161617e-06,
+      "loss": 0.6112,
+      "step": 3930
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 2.15959595959596e-06,
+      "loss": 0.3291,
+      "step": 3931
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 2.157575757575758e-06,
+      "loss": 0.3297,
+      "step": 3932
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 2.1555555555555558e-06,
+      "loss": 0.5158,
+      "step": 3933
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 2.1535353535353537e-06,
+      "loss": 0.4665,
+      "step": 3934
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 2.1515151515151515e-06,
+      "loss": 0.6281,
+      "step": 3935
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 2.1494949494949494e-06,
+      "loss": 0.6499,
+      "step": 3936
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 2.1474747474747477e-06,
+      "loss": 0.4607,
+      "step": 3937
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 2.1454545454545456e-06,
+      "loss": 0.2698,
+      "step": 3938
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 2.1434343434343435e-06,
+      "loss": 0.3621,
+      "step": 3939
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 2.1414141414141414e-06,
+      "loss": 0.8565,
+      "step": 3940
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 2.1393939393939393e-06,
+      "loss": 0.1252,
+      "step": 3941
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 2.1373737373737376e-06,
+      "loss": 0.3469,
+      "step": 3942
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 2.1353535353535355e-06,
+      "loss": 1.6591,
+      "step": 3943
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 2.133333333333334e-06,
+      "loss": 0.9942,
+      "step": 3944
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 2.1313131313131317e-06,
+      "loss": 0.4978,
+      "step": 3945
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 2.1292929292929296e-06,
+      "loss": 0.3506,
+      "step": 3946
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 2.1272727272727275e-06,
+      "loss": 0.671,
+      "step": 3947
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 2.1252525252525254e-06,
+      "loss": 0.68,
+      "step": 3948
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 2.1232323232323237e-06,
+      "loss": 1.1883,
+      "step": 3949
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 2.1212121212121216e-06,
+      "loss": 0.704,
+      "step": 3950
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 2.1191919191919194e-06,
+      "loss": 0.3942,
+      "step": 3951
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 2.1171717171717173e-06,
+      "loss": 0.4659,
+      "step": 3952
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 2.1151515151515152e-06,
+      "loss": 0.3503,
+      "step": 3953
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 2.113131313131313e-06,
+      "loss": 0.5149,
+      "step": 3954
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 2.1111111111111114e-06,
+      "loss": 0.2606,
+      "step": 3955
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 2.1090909090909093e-06,
+      "loss": 0.4104,
+      "step": 3956
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 2.107070707070707e-06,
+      "loss": 1.5767,
+      "step": 3957
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 2.105050505050505e-06,
+      "loss": 1.2984,
+      "step": 3958
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 2.103030303030303e-06,
+      "loss": 0.7739,
+      "step": 3959
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 2.1010101010101013e-06,
+      "loss": 0.211,
+      "step": 3960
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 2.098989898989899e-06,
+      "loss": 0.3735,
+      "step": 3961
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 2.096969696969697e-06,
+      "loss": 0.466,
+      "step": 3962
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 2.094949494949495e-06,
+      "loss": 0.3511,
+      "step": 3963
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 2.092929292929293e-06,
+      "loss": 0.235,
+      "step": 3964
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 2.090909090909091e-06,
+      "loss": 0.2156,
+      "step": 3965
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.088888888888889e-06,
+      "loss": 0.3108,
+      "step": 3966
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.0868686868686873e-06,
+      "loss": 0.3139,
+      "step": 3967
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.0848484848484852e-06,
+      "loss": 1.0638,
+      "step": 3968
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2.082828282828283e-06,
+      "loss": 0.5501,
+      "step": 3969
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.080808080808081e-06,
+      "loss": 0.7453,
+      "step": 3970
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.078787878787879e-06,
+      "loss": 0.5166,
+      "step": 3971
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.0767676767676768e-06,
+      "loss": 0.7174,
+      "step": 3972
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.074747474747475e-06,
+      "loss": 0.4025,
+      "step": 3973
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.072727272727273e-06,
+      "loss": 0.747,
+      "step": 3974
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.070707070707071e-06,
+      "loss": 0.5345,
+      "step": 3975
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.0686868686868688e-06,
+      "loss": 0.5158,
+      "step": 3976
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.0666666666666666e-06,
+      "loss": 0.5002,
+      "step": 3977
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.064646464646465e-06,
+      "loss": 0.3442,
+      "step": 3978
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.062626262626263e-06,
+      "loss": 0.8641,
+      "step": 3979
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.0606060606060607e-06,
+      "loss": 0.8827,
+      "step": 3980
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.0585858585858586e-06,
+      "loss": 0.3387,
+      "step": 3981
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.0565656565656565e-06,
+      "loss": 0.6827,
+      "step": 3982
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.054545454545455e-06,
+      "loss": 0.4337,
+      "step": 3983
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 2.0525252525252527e-06,
+      "loss": 0.3698,
+      "step": 3984
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 2.0505050505050506e-06,
+      "loss": 0.3306,
+      "step": 3985
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 2.0484848484848485e-06,
+      "loss": 1.0829,
+      "step": 3986
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 2.046464646464647e-06,
+      "loss": 1.0158,
+      "step": 3987
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 2.0444444444444447e-06,
+      "loss": 0.2985,
+      "step": 3988
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 2.0424242424242426e-06,
+      "loss": 0.4634,
+      "step": 3989
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 2.0404040404040405e-06,
+      "loss": 0.6192,
+      "step": 3990
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 2.0383838383838388e-06,
+      "loss": 0.6372,
+      "step": 3991
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 2.0363636363636367e-06,
+      "loss": 0.7424,
+      "step": 3992
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 2.0343434343434345e-06,
+      "loss": 1.3622,
+      "step": 3993
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 2.0323232323232324e-06,
+      "loss": 0.5032,
+      "step": 3994
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 2.0303030303030303e-06,
+      "loss": 0.4674,
+      "step": 3995
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 2.0282828282828286e-06,
+      "loss": 0.2072,
+      "step": 3996
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 2.0262626262626265e-06,
+      "loss": 0.446,
+      "step": 3997
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 2.0242424242424244e-06,
+      "loss": 0.4921,
+      "step": 3998
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 2.0222222222222223e-06,
+      "loss": 0.4254,
+      "step": 3999
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 2.02020202020202e-06,
+      "loss": 0.4299,
+      "step": 4000
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.83,
+      "eval_loss": 0.7071203589439392,
+      "eval_roc_auc": 0.965075423654258,
+      "eval_runtime": 92.8699,
+      "eval_samples_per_second": 2.154,
+      "eval_steps_per_second": 0.538,
+      "step": 4000
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 2.0181818181818185e-06,
+      "loss": 0.3271,
+      "step": 4001
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 2.0161616161616164e-06,
+      "loss": 0.2542,
+      "step": 4002
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 2.0141414141414143e-06,
+      "loss": 0.3747,
+      "step": 4003
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 2.012121212121212e-06,
+      "loss": 0.6694,
+      "step": 4004
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 2.01010101010101e-06,
+      "loss": 0.4213,
+      "step": 4005
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 2.008080808080808e-06,
+      "loss": 0.6612,
+      "step": 4006
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 2.0060606060606062e-06,
+      "loss": 0.8191,
+      "step": 4007
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 2.004040404040404e-06,
+      "loss": 0.7951,
+      "step": 4008
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 2.0020202020202024e-06,
+      "loss": 0.1206,
+      "step": 4009
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.5091,
+      "step": 4010
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.9979797979797982e-06,
+      "loss": 0.8,
+      "step": 4011
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.995959595959596e-06,
+      "loss": 0.6136,
+      "step": 4012
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.993939393939394e-06,
+      "loss": 0.8407,
+      "step": 4013
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.9919191919191923e-06,
+      "loss": 0.5716,
+      "step": 4014
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.98989898989899e-06,
+      "loss": 0.8464,
+      "step": 4015
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.987878787878788e-06,
+      "loss": 0.6284,
+      "step": 4016
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.985858585858586e-06,
+      "loss": 0.4497,
+      "step": 4017
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.983838383838384e-06,
+      "loss": 0.1754,
+      "step": 4018
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.981818181818182e-06,
+      "loss": 0.4323,
+      "step": 4019
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.97979797979798e-06,
+      "loss": 1.4912,
+      "step": 4020
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.977777777777778e-06,
+      "loss": 0.6535,
+      "step": 4021
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.975757575757576e-06,
+      "loss": 0.4478,
+      "step": 4022
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.9737373737373737e-06,
+      "loss": 0.9072,
+      "step": 4023
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.9717171717171716e-06,
+      "loss": 0.5944,
+      "step": 4024
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.96969696969697e-06,
+      "loss": 1.1641,
+      "step": 4025
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.967676767676768e-06,
+      "loss": 0.3846,
+      "step": 4026
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.9656565656565657e-06,
+      "loss": 0.4067,
+      "step": 4027
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.9636363636363636e-06,
+      "loss": 0.3955,
+      "step": 4028
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.9616161616161615e-06,
+      "loss": 0.4641,
+      "step": 4029
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.9595959595959598e-06,
+      "loss": 0.3297,
+      "step": 4030
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.9575757575757577e-06,
+      "loss": 0.5759,
+      "step": 4031
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.955555555555556e-06,
+      "loss": 0.7613,
+      "step": 4032
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.953535353535354e-06,
+      "loss": 0.2712,
+      "step": 4033
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 1.9515151515151518e-06,
+      "loss": 0.4981,
+      "step": 4034
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.9494949494949496e-06,
+      "loss": 0.5947,
+      "step": 4035
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.9474747474747475e-06,
+      "loss": 0.3477,
+      "step": 4036
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.945454545454546e-06,
+      "loss": 0.4528,
+      "step": 4037
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 1.9434343434343437e-06,
+      "loss": 0.3558,
+      "step": 4038
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.9414141414141416e-06,
+      "loss": 0.4831,
+      "step": 4039
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.9393939393939395e-06,
+      "loss": 0.7067,
+      "step": 4040
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.9373737373737374e-06,
+      "loss": 1.4082,
+      "step": 4041
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.9353535353535353e-06,
+      "loss": 2.2224,
+      "step": 4042
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.9333333333333336e-06,
+      "loss": 0.5037,
+      "step": 4043
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.9313131313131315e-06,
+      "loss": 0.6529,
+      "step": 4044
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.9292929292929294e-06,
+      "loss": 0.3195,
+      "step": 4045
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.9272727272727273e-06,
+      "loss": 0.523,
+      "step": 4046
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.925252525252525e-06,
+      "loss": 0.1757,
+      "step": 4047
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.9232323232323235e-06,
+      "loss": 0.362,
+      "step": 4048
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.9212121212121213e-06,
+      "loss": 0.2774,
+      "step": 4049
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.9191919191919192e-06,
+      "loss": 0.674,
+      "step": 4050
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.9171717171717175e-06,
+      "loss": 1.0386,
+      "step": 4051
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.9151515151515154e-06,
+      "loss": 0.3244,
+      "step": 4052
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.9131313131313133e-06,
+      "loss": 0.437,
+      "step": 4053
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.9111111111111112e-06,
+      "loss": 0.476,
+      "step": 4054
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.9090909090909095e-06,
+      "loss": 1.2084,
+      "step": 4055
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 1.9070707070707072e-06,
+      "loss": 0.2488,
+      "step": 4056
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.9050505050505053e-06,
+      "loss": 0.6415,
+      "step": 4057
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.9030303030303032e-06,
+      "loss": 0.1748,
+      "step": 4058
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.9010101010101013e-06,
+      "loss": 0.8428,
+      "step": 4059
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.8989898989898992e-06,
+      "loss": 1.82,
+      "step": 4060
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.896969696969697e-06,
+      "loss": 0.1401,
+      "step": 4061
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 1.8949494949494952e-06,
+      "loss": 0.3353,
+      "step": 4062
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.892929292929293e-06,
+      "loss": 1.0,
+      "step": 4063
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.890909090909091e-06,
+      "loss": 0.3404,
+      "step": 4064
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.888888888888889e-06,
+      "loss": 0.6158,
+      "step": 4065
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 1.886868686868687e-06,
+      "loss": 0.735,
+      "step": 4066
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.884848484848485e-06,
+      "loss": 0.5054,
+      "step": 4067
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.882828282828283e-06,
+      "loss": 0.2767,
+      "step": 4068
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.8808080808080808e-06,
+      "loss": 0.3493,
+      "step": 4069
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 1.878787878787879e-06,
+      "loss": 0.2407,
+      "step": 4070
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.8767676767676768e-06,
+      "loss": 0.8098,
+      "step": 4071
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.874747474747475e-06,
+      "loss": 0.8562,
+      "step": 4072
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.872727272727273e-06,
+      "loss": 0.9772,
+      "step": 4073
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 1.8707070707070709e-06,
+      "loss": 0.4215,
+      "step": 4074
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.868686868686869e-06,
+      "loss": 1.034,
+      "step": 4075
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.8666666666666669e-06,
+      "loss": 0.556,
+      "step": 4076
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.864646464646465e-06,
+      "loss": 0.1731,
+      "step": 4077
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.8626262626262629e-06,
+      "loss": 0.3918,
+      "step": 4078
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.8606060606060607e-06,
+      "loss": 0.3048,
+      "step": 4079
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.8585858585858588e-06,
+      "loss": 0.6133,
+      "step": 4080
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.8565656565656567e-06,
+      "loss": 0.2649,
+      "step": 4081
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.8545454545454546e-06,
+      "loss": 0.5525,
+      "step": 4082
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.8525252525252527e-06,
+      "loss": 0.3217,
+      "step": 4083
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.8505050505050506e-06,
+      "loss": 1.2332,
+      "step": 4084
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.8484848484848487e-06,
+      "loss": 0.4478,
+      "step": 4085
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.8464646464646466e-06,
+      "loss": 1.3464,
+      "step": 4086
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.8444444444444445e-06,
+      "loss": 0.9984,
+      "step": 4087
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.8424242424242426e-06,
+      "loss": 0.7665,
+      "step": 4088
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.8404040404040405e-06,
+      "loss": 0.8058,
+      "step": 4089
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.8383838383838384e-06,
+      "loss": 0.5904,
+      "step": 4090
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.8363636363636365e-06,
+      "loss": 0.6098,
+      "step": 4091
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.8343434343434343e-06,
+      "loss": 0.5453,
+      "step": 4092
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.8323232323232324e-06,
+      "loss": 0.3359,
+      "step": 4093
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 1.8303030303030305e-06,
+      "loss": 0.5873,
+      "step": 4094
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.8282828282828286e-06,
+      "loss": 0.5359,
+      "step": 4095
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.8262626262626265e-06,
+      "loss": 0.4396,
+      "step": 4096
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.8242424242424244e-06,
+      "loss": 0.2405,
+      "step": 4097
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.8222222222222225e-06,
+      "loss": 0.8246,
+      "step": 4098
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.8202020202020204e-06,
+      "loss": 0.5813,
+      "step": 4099
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.8181818181818183e-06,
+      "loss": 0.439,
+      "step": 4100
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.8161616161616164e-06,
+      "loss": 0.262,
+      "step": 4101
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.8141414141414143e-06,
+      "loss": 1.0171,
+      "step": 4102
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.8121212121212124e-06,
+      "loss": 0.3222,
+      "step": 4103
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.8101010101010103e-06,
+      "loss": 0.3503,
+      "step": 4104
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.8080808080808082e-06,
+      "loss": 0.3513,
+      "step": 4105
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.8060606060606063e-06,
+      "loss": 0.3616,
+      "step": 4106
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.8040404040404041e-06,
+      "loss": 0.7648,
+      "step": 4107
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.802020202020202e-06,
+      "loss": 0.5504,
+      "step": 4108
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 0.1839,
+      "step": 4109
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.797979797979798e-06,
+      "loss": 0.4329,
+      "step": 4110
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.7959595959595961e-06,
+      "loss": 0.2181,
+      "step": 4111
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.793939393939394e-06,
+      "loss": 0.2583,
+      "step": 4112
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.7919191919191919e-06,
+      "loss": 0.3144,
+      "step": 4113
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.78989898989899e-06,
+      "loss": 0.8424,
+      "step": 4114
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.787878787878788e-06,
+      "loss": 0.408,
+      "step": 4115
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.7858585858585862e-06,
+      "loss": 1.0826,
+      "step": 4116
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.783838383838384e-06,
+      "loss": 0.3853,
+      "step": 4117
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.781818181818182e-06,
+      "loss": 0.2792,
+      "step": 4118
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.77979797979798e-06,
+      "loss": 1.9943,
+      "step": 4119
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.777777777777778e-06,
+      "loss": 0.3031,
+      "step": 4120
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.775757575757576e-06,
+      "loss": 1.2911,
+      "step": 4121
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.773737373737374e-06,
+      "loss": 0.5068,
+      "step": 4122
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.7717171717171718e-06,
+      "loss": 0.4599,
+      "step": 4123
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.76969696969697e-06,
+      "loss": 0.3993,
+      "step": 4124
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.7676767676767678e-06,
+      "loss": 0.4119,
+      "step": 4125
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.7656565656565657e-06,
+      "loss": 0.6385,
+      "step": 4126
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.7636363636363638e-06,
+      "loss": 0.9098,
+      "step": 4127
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.7616161616161617e-06,
+      "loss": 0.8655,
+      "step": 4128
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.7595959595959598e-06,
+      "loss": 0.6021,
+      "step": 4129
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.7575757575757577e-06,
+      "loss": 0.2496,
+      "step": 4130
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.7555555555555556e-06,
+      "loss": 0.3625,
+      "step": 4131
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.7535353535353537e-06,
+      "loss": 0.3969,
+      "step": 4132
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.7515151515151516e-06,
+      "loss": 0.3561,
+      "step": 4133
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 1.7494949494949494e-06,
+      "loss": 0.4409,
+      "step": 4134
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.7474747474747475e-06,
+      "loss": 0.4015,
+      "step": 4135
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.7454545454545456e-06,
+      "loss": 1.0043,
+      "step": 4136
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.7434343434343437e-06,
+      "loss": 0.7955,
+      "step": 4137
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 1.7414141414141416e-06,
+      "loss": 0.3012,
+      "step": 4138
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.7393939393939397e-06,
+      "loss": 1.1435,
+      "step": 4139
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.7373737373737376e-06,
+      "loss": 0.4724,
+      "step": 4140
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.7353535353535355e-06,
+      "loss": 0.384,
+      "step": 4141
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.7333333333333336e-06,
+      "loss": 0.2722,
+      "step": 4142
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.7313131313131315e-06,
+      "loss": 0.3456,
+      "step": 4143
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 1.7292929292929294e-06,
+      "loss": 1.2849,
+      "step": 4144
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.7272727272727275e-06,
+      "loss": 0.4233,
+      "step": 4145
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.7252525252525254e-06,
+      "loss": 0.4883,
+      "step": 4146
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.7232323232323235e-06,
+      "loss": 0.3402,
+      "step": 4147
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.7212121212121214e-06,
+      "loss": 0.1673,
+      "step": 4148
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.7191919191919192e-06,
+      "loss": 1.1282,
+      "step": 4149
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.7171717171717173e-06,
+      "loss": 1.2188,
+      "step": 4150
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.7151515151515152e-06,
+      "loss": 0.285,
+      "step": 4151
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.7131313131313131e-06,
+      "loss": 0.6873,
+      "step": 4152
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.7111111111111112e-06,
+      "loss": 0.4252,
+      "step": 4153
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.7090909090909091e-06,
+      "loss": 0.4706,
+      "step": 4154
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.707070707070707e-06,
+      "loss": 0.2543,
+      "step": 4155
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.705050505050505e-06,
+      "loss": 1.0925,
+      "step": 4156
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.703030303030303e-06,
+      "loss": 0.3878,
+      "step": 4157
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.7010101010101013e-06,
+      "loss": 0.2297,
+      "step": 4158
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.6989898989898992e-06,
+      "loss": 0.3603,
+      "step": 4159
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.6969696969696973e-06,
+      "loss": 0.5616,
+      "step": 4160
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.6949494949494952e-06,
+      "loss": 0.2734,
+      "step": 4161
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 1.692929292929293e-06,
+      "loss": 0.4095,
+      "step": 4162
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.6909090909090912e-06,
+      "loss": 0.1805,
+      "step": 4163
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.688888888888889e-06,
+      "loss": 0.2574,
+      "step": 4164
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.6868686868686871e-06,
+      "loss": 0.2142,
+      "step": 4165
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.684848484848485e-06,
+      "loss": 0.4672,
+      "step": 4166
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.682828282828283e-06,
+      "loss": 0.1923,
+      "step": 4167
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.680808080808081e-06,
+      "loss": 0.4602,
+      "step": 4168
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.678787878787879e-06,
+      "loss": 0.8239,
+      "step": 4169
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.6767676767676768e-06,
+      "loss": 1.4672,
+      "step": 4170
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.674747474747475e-06,
+      "loss": 0.382,
+      "step": 4171
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.6727272727272728e-06,
+      "loss": 1.0534,
+      "step": 4172
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.6707070707070707e-06,
+      "loss": 0.999,
+      "step": 4173
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.6686868686868688e-06,
+      "loss": 0.4899,
+      "step": 4174
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 1.6666666666666667e-06,
+      "loss": 0.8101,
+      "step": 4175
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 1.6646464646464648e-06,
+      "loss": 1.381,
+      "step": 4176
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.6626262626262626e-06,
+      "loss": 0.3957,
+      "step": 4177
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.6606060606060605e-06,
+      "loss": 0.2438,
+      "step": 4178
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.6585858585858588e-06,
+      "loss": 0.7549,
+      "step": 4179
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 1.6565656565656567e-06,
+      "loss": 0.3491,
+      "step": 4180
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.6545454545454548e-06,
+      "loss": 0.3952,
+      "step": 4181
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.6525252525252527e-06,
+      "loss": 0.8517,
+      "step": 4182
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.6505050505050508e-06,
+      "loss": 0.4469,
+      "step": 4183
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.6484848484848487e-06,
+      "loss": 0.9374,
+      "step": 4184
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.6464646464646466e-06,
+      "loss": 0.8896,
+      "step": 4185
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.6444444444444447e-06,
+      "loss": 0.829,
+      "step": 4186
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.6424242424242426e-06,
+      "loss": 0.194,
+      "step": 4187
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.6404040404040405e-06,
+      "loss": 0.4846,
+      "step": 4188
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.6383838383838386e-06,
+      "loss": 0.3489,
+      "step": 4189
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.6363636363636365e-06,
+      "loss": 0.4007,
+      "step": 4190
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.6343434343434344e-06,
+      "loss": 1.9337,
+      "step": 4191
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.6323232323232325e-06,
+      "loss": 0.3295,
+      "step": 4192
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.6303030303030303e-06,
+      "loss": 0.4173,
+      "step": 4193
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.6282828282828284e-06,
+      "loss": 0.1555,
+      "step": 4194
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.6262626262626263e-06,
+      "loss": 0.326,
+      "step": 4195
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.6242424242424242e-06,
+      "loss": 0.3484,
+      "step": 4196
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.6222222222222223e-06,
+      "loss": 0.425,
+      "step": 4197
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.6202020202020202e-06,
+      "loss": 0.3216,
+      "step": 4198
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.618181818181818e-06,
+      "loss": 0.2779,
+      "step": 4199
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.6161616161616164e-06,
+      "loss": 1.3355,
+      "step": 4200
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": 0.805,
+      "eval_loss": 0.7661997079849243,
+      "eval_roc_auc": 0.9601987755481718,
+      "eval_runtime": 92.0735,
+      "eval_samples_per_second": 2.172,
+      "eval_steps_per_second": 0.543,
+      "step": 4200
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.6141414141414145e-06,
+      "loss": 0.4658,
+      "step": 4201
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.6121212121212124e-06,
+      "loss": 0.2605,
+      "step": 4202
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.6101010101010103e-06,
+      "loss": 0.67,
+      "step": 4203
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.6080808080808084e-06,
+      "loss": 0.2005,
+      "step": 4204
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.6060606060606063e-06,
+      "loss": 0.6869,
+      "step": 4205
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.6040404040404042e-06,
+      "loss": 0.3709,
+      "step": 4206
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.6020202020202023e-06,
+      "loss": 0.2049,
+      "step": 4207
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 0.5148,
+      "step": 4208
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.597979797979798e-06,
+      "loss": 0.2564,
+      "step": 4209
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.5959595959595961e-06,
+      "loss": 0.4181,
+      "step": 4210
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.593939393939394e-06,
+      "loss": 0.3027,
+      "step": 4211
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 1.5919191919191921e-06,
+      "loss": 0.9417,
+      "step": 4212
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.58989898989899e-06,
+      "loss": 0.6602,
+      "step": 4213
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.5878787878787879e-06,
+      "loss": 0.2758,
+      "step": 4214
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.585858585858586e-06,
+      "loss": 0.5746,
+      "step": 4215
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 1.5838383838383839e-06,
+      "loss": 1.1227,
+      "step": 4216
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.5818181818181818e-06,
+      "loss": 0.3564,
+      "step": 4217
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.5797979797979799e-06,
+      "loss": 0.547,
+      "step": 4218
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.5777777777777778e-06,
+      "loss": 0.3351,
+      "step": 4219
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 1.5757575757575759e-06,
+      "loss": 0.2875,
+      "step": 4220
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.5737373737373737e-06,
+      "loss": 0.3998,
+      "step": 4221
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.571717171717172e-06,
+      "loss": 0.4715,
+      "step": 4222
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.56969696969697e-06,
+      "loss": 0.769,
+      "step": 4223
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 1.5676767676767678e-06,
+      "loss": 0.4425,
+      "step": 4224
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 1.565656565656566e-06,
+      "loss": 0.6105,
+      "step": 4225
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.5636363636363638e-06,
+      "loss": 0.4296,
+      "step": 4226
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.5616161616161617e-06,
+      "loss": 0.6667,
+      "step": 4227
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.5595959595959598e-06,
+      "loss": 1.0703,
+      "step": 4228
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.5575757575757577e-06,
+      "loss": 0.987,
+      "step": 4229
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.5555555555555558e-06,
+      "loss": 1.1754,
+      "step": 4230
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.5535353535353537e-06,
+      "loss": 0.5758,
+      "step": 4231
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.5515151515151516e-06,
+      "loss": 0.2649,
+      "step": 4232
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.5494949494949497e-06,
+      "loss": 1.3573,
+      "step": 4233
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 1.5474747474747476e-06,
+      "loss": 1.1595,
+      "step": 4234
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.5454545454545454e-06,
+      "loss": 0.2882,
+      "step": 4235
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.5434343434343435e-06,
+      "loss": 0.2803,
+      "step": 4236
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.5414141414141414e-06,
+      "loss": 0.1971,
+      "step": 4237
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 1.5393939393939395e-06,
+      "loss": 0.4834,
+      "step": 4238
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.5373737373737374e-06,
+      "loss": 0.5386,
+      "step": 4239
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.5353535353535353e-06,
+      "loss": 0.1053,
+      "step": 4240
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.5333333333333334e-06,
+      "loss": 0.4525,
+      "step": 4241
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.5313131313131313e-06,
+      "loss": 0.6464,
+      "step": 4242
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.5292929292929296e-06,
+      "loss": 0.4807,
+      "step": 4243
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 1.5272727272727275e-06,
+      "loss": 0.2041,
+      "step": 4244
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.5252525252525254e-06,
+      "loss": 0.7065,
+      "step": 4245
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.5232323232323235e-06,
+      "loss": 0.4901,
+      "step": 4246
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.5212121212121214e-06,
+      "loss": 0.3939,
+      "step": 4247
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 1.5191919191919195e-06,
+      "loss": 0.807,
+      "step": 4248
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.5171717171717174e-06,
+      "loss": 0.259,
+      "step": 4249
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.5151515151515152e-06,
+      "loss": 0.5818,
+      "step": 4250
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.5131313131313133e-06,
+      "loss": 0.6142,
+      "step": 4251
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 1.5111111111111112e-06,
+      "loss": 0.6695,
+      "step": 4252
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.5090909090909091e-06,
+      "loss": 0.2513,
+      "step": 4253
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.5070707070707072e-06,
+      "loss": 0.2739,
+      "step": 4254
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.5050505050505051e-06,
+      "loss": 0.2826,
+      "step": 4255
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 1.5030303030303032e-06,
+      "loss": 0.8297,
+      "step": 4256
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.501010101010101e-06,
+      "loss": 1.2726,
+      "step": 4257
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.498989898989899e-06,
+      "loss": 0.4159,
+      "step": 4258
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.496969696969697e-06,
+      "loss": 0.8157,
+      "step": 4259
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.494949494949495e-06,
+      "loss": 0.4521,
+      "step": 4260
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.4929292929292929e-06,
+      "loss": 1.477,
+      "step": 4261
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.490909090909091e-06,
+      "loss": 0.8197,
+      "step": 4262
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.4888888888888888e-06,
+      "loss": 0.2647,
+      "step": 4263
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.4868686868686872e-06,
+      "loss": 0.3454,
+      "step": 4264
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.484848484848485e-06,
+      "loss": 0.6049,
+      "step": 4265
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.4828282828282831e-06,
+      "loss": 0.4692,
+      "step": 4266
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.480808080808081e-06,
+      "loss": 0.3245,
+      "step": 4267
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.478787878787879e-06,
+      "loss": 1.2367,
+      "step": 4268
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.476767676767677e-06,
+      "loss": 0.1882,
+      "step": 4269
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.474747474747475e-06,
+      "loss": 0.194,
+      "step": 4270
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.4727272727272728e-06,
+      "loss": 1.154,
+      "step": 4271
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.470707070707071e-06,
+      "loss": 0.2895,
+      "step": 4272
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.4686868686868688e-06,
+      "loss": 0.4002,
+      "step": 4273
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.4666666666666669e-06,
+      "loss": 0.9285,
+      "step": 4274
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.4646464646464648e-06,
+      "loss": 0.5598,
+      "step": 4275
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.4626262626262627e-06,
+      "loss": 0.6301,
+      "step": 4276
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.4606060606060608e-06,
+      "loss": 0.4659,
+      "step": 4277
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.4585858585858586e-06,
+      "loss": 0.3423,
+      "step": 4278
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.4565656565656565e-06,
+      "loss": 0.3356,
+      "step": 4279
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.4545454545454546e-06,
+      "loss": 0.2721,
+      "step": 4280
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.4525252525252525e-06,
+      "loss": 0.2903,
+      "step": 4281
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.4505050505050506e-06,
+      "loss": 1.2791,
+      "step": 4282
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.4484848484848485e-06,
+      "loss": 0.2826,
+      "step": 4283
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.4464646464646464e-06,
+      "loss": 0.4326,
+      "step": 4284
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.4444444444444445e-06,
+      "loss": 0.4279,
+      "step": 4285
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.4424242424242426e-06,
+      "loss": 0.4163,
+      "step": 4286
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.4404040404040407e-06,
+      "loss": 0.4561,
+      "step": 4287
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 1.4383838383838386e-06,
+      "loss": 1.5497,
+      "step": 4288
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.4363636363636365e-06,
+      "loss": 0.5052,
+      "step": 4289
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.4343434343434346e-06,
+      "loss": 0.3579,
+      "step": 4290
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.4323232323232325e-06,
+      "loss": 0.2345,
+      "step": 4291
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 1.4303030303030306e-06,
+      "loss": 0.2723,
+      "step": 4292
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 1.4282828282828284e-06,
+      "loss": 0.183,
+      "step": 4293
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.4262626262626263e-06,
+      "loss": 0.5082,
+      "step": 4294
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.4242424242424244e-06,
+      "loss": 0.3485,
+      "step": 4295
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.4222222222222223e-06,
+      "loss": 0.3589,
+      "step": 4296
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.4202020202020202e-06,
+      "loss": 0.3127,
+      "step": 4297
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.4181818181818183e-06,
+      "loss": 0.5982,
+      "step": 4298
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.4161616161616162e-06,
+      "loss": 1.4265,
+      "step": 4299
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.4141414141414143e-06,
+      "loss": 0.2343,
+      "step": 4300
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.4121212121212122e-06,
+      "loss": 0.6012,
+      "step": 4301
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.41010101010101e-06,
+      "loss": 0.3406,
+      "step": 4302
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.4080808080808082e-06,
+      "loss": 0.3982,
+      "step": 4303
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.406060606060606e-06,
+      "loss": 0.5096,
+      "step": 4304
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.404040404040404e-06,
+      "loss": 0.8418,
+      "step": 4305
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 1.402020202020202e-06,
+      "loss": 0.2134,
+      "step": 4306
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 0.3341,
+      "step": 4307
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.3979797979797982e-06,
+      "loss": 0.3486,
+      "step": 4308
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.3959595959595961e-06,
+      "loss": 1.0163,
+      "step": 4309
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.3939393939393942e-06,
+      "loss": 0.8699,
+      "step": 4310
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.3919191919191921e-06,
+      "loss": 0.4112,
+      "step": 4311
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.38989898989899e-06,
+      "loss": 0.6779,
+      "step": 4312
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.3878787878787881e-06,
+      "loss": 0.2285,
+      "step": 4313
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.385858585858586e-06,
+      "loss": 0.4003,
+      "step": 4314
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.3838383838383839e-06,
+      "loss": 0.424,
+      "step": 4315
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 1.381818181818182e-06,
+      "loss": 0.3198,
+      "step": 4316
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.3797979797979799e-06,
+      "loss": 0.3551,
+      "step": 4317
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.377777777777778e-06,
+      "loss": 0.4091,
+      "step": 4318
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.3757575757575759e-06,
+      "loss": 0.4799,
+      "step": 4319
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 1.3737373737373738e-06,
+      "loss": 0.8733,
+      "step": 4320
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.3717171717171718e-06,
+      "loss": 0.5905,
+      "step": 4321
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.3696969696969697e-06,
+      "loss": 0.374,
+      "step": 4322
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.3676767676767676e-06,
+      "loss": 0.3292,
+      "step": 4323
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.3656565656565657e-06,
+      "loss": 0.4289,
+      "step": 4324
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.3636363636363636e-06,
+      "loss": 0.5227,
+      "step": 4325
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.3616161616161617e-06,
+      "loss": 0.3589,
+      "step": 4326
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.3595959595959596e-06,
+      "loss": 0.4884,
+      "step": 4327
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.357575757575758e-06,
+      "loss": 0.1897,
+      "step": 4328
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.3555555555555558e-06,
+      "loss": 0.4273,
+      "step": 4329
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.3535353535353537e-06,
+      "loss": 0.3435,
+      "step": 4330
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.3515151515151518e-06,
+      "loss": 0.8341,
+      "step": 4331
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.3494949494949497e-06,
+      "loss": 0.4074,
+      "step": 4332
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.3474747474747476e-06,
+      "loss": 0.4,
+      "step": 4333
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.3454545454545457e-06,
+      "loss": 0.28,
+      "step": 4334
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.3434343434343436e-06,
+      "loss": 0.3509,
+      "step": 4335
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.3414141414141417e-06,
+      "loss": 0.6763,
+      "step": 4336
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.3393939393939395e-06,
+      "loss": 0.3266,
+      "step": 4337
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.3373737373737374e-06,
+      "loss": 0.4482,
+      "step": 4338
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.3353535353535355e-06,
+      "loss": 0.3574,
+      "step": 4339
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.3333333333333334e-06,
+      "loss": 0.3564,
+      "step": 4340
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.3313131313131313e-06,
+      "loss": 1.1437,
+      "step": 4341
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.3292929292929294e-06,
+      "loss": 0.5065,
+      "step": 4342
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.3272727272727273e-06,
+      "loss": 0.311,
+      "step": 4343
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 1.3252525252525254e-06,
+      "loss": 0.8591,
+      "step": 4344
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.3232323232323233e-06,
+      "loss": 0.5548,
+      "step": 4345
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.3212121212121212e-06,
+      "loss": 0.465,
+      "step": 4346
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.3191919191919193e-06,
+      "loss": 0.2923,
+      "step": 4347
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.3171717171717172e-06,
+      "loss": 0.469,
+      "step": 4348
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.315151515151515e-06,
+      "loss": 0.5045,
+      "step": 4349
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.3131313131313134e-06,
+      "loss": 0.6751,
+      "step": 4350
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.3111111111111112e-06,
+      "loss": 0.7557,
+      "step": 4351
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.3090909090909093e-06,
+      "loss": 0.313,
+      "step": 4352
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.3070707070707072e-06,
+      "loss": 0.736,
+      "step": 4353
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.3050505050505053e-06,
+      "loss": 0.6605,
+      "step": 4354
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.3030303030303032e-06,
+      "loss": 0.5683,
+      "step": 4355
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.301010101010101e-06,
+      "loss": 0.4927,
+      "step": 4356
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.2989898989898992e-06,
+      "loss": 0.974,
+      "step": 4357
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.296969696969697e-06,
+      "loss": 0.2694,
+      "step": 4358
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.294949494949495e-06,
+      "loss": 0.5017,
+      "step": 4359
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.292929292929293e-06,
+      "loss": 0.1394,
+      "step": 4360
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.290909090909091e-06,
+      "loss": 1.0107,
+      "step": 4361
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.288888888888889e-06,
+      "loss": 0.6657,
+      "step": 4362
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.286868686868687e-06,
+      "loss": 0.3816,
+      "step": 4363
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.2848484848484848e-06,
+      "loss": 0.3945,
+      "step": 4364
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.282828282828283e-06,
+      "loss": 0.4425,
+      "step": 4365
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.2808080808080808e-06,
+      "loss": 0.5359,
+      "step": 4366
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.2787878787878787e-06,
+      "loss": 0.6813,
+      "step": 4367
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.2767676767676768e-06,
+      "loss": 0.7002,
+      "step": 4368
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.2747474747474747e-06,
+      "loss": 0.4077,
+      "step": 4369
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.2727272727272728e-06,
+      "loss": 0.5532,
+      "step": 4370
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.270707070707071e-06,
+      "loss": 0.3992,
+      "step": 4371
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.268686868686869e-06,
+      "loss": 0.5277,
+      "step": 4372
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.2666666666666669e-06,
+      "loss": 0.8907,
+      "step": 4373
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.2646464646464648e-06,
+      "loss": 0.4853,
+      "step": 4374
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.2626262626262629e-06,
+      "loss": 0.2756,
+      "step": 4375
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.2606060606060608e-06,
+      "loss": 0.4421,
+      "step": 4376
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.2585858585858587e-06,
+      "loss": 0.7553,
+      "step": 4377
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.2565656565656568e-06,
+      "loss": 0.2623,
+      "step": 4378
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.2545454545454546e-06,
+      "loss": 0.5548,
+      "step": 4379
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.2525252525252527e-06,
+      "loss": 0.3347,
+      "step": 4380
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.2505050505050506e-06,
+      "loss": 0.3236,
+      "step": 4381
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.2484848484848485e-06,
+      "loss": 0.3847,
+      "step": 4382
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.2464646464646466e-06,
+      "loss": 0.3227,
+      "step": 4383
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 1.2444444444444445e-06,
+      "loss": 0.3769,
+      "step": 4384
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.2424242424242424e-06,
+      "loss": 0.2363,
+      "step": 4385
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.2404040404040405e-06,
+      "loss": 1.2331,
+      "step": 4386
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.2383838383838386e-06,
+      "loss": 0.3193,
+      "step": 4387
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 1.2363636363636365e-06,
+      "loss": 0.2533,
+      "step": 4388
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.2343434343434346e-06,
+      "loss": 0.2475,
+      "step": 4389
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.2323232323232325e-06,
+      "loss": 0.1112,
+      "step": 4390
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.2303030303030304e-06,
+      "loss": 0.4635,
+      "step": 4391
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.2282828282828285e-06,
+      "loss": 0.6009,
+      "step": 4392
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.2262626262626263e-06,
+      "loss": 0.207,
+      "step": 4393
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.2242424242424242e-06,
+      "loss": 0.4543,
+      "step": 4394
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.2222222222222223e-06,
+      "loss": 0.4471,
+      "step": 4395
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.2202020202020202e-06,
+      "loss": 0.3651,
+      "step": 4396
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.2181818181818183e-06,
+      "loss": 0.7234,
+      "step": 4397
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 1.2161616161616164e-06,
+      "loss": 0.6135,
+      "step": 4398
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.2141414141414143e-06,
+      "loss": 1.549,
+      "step": 4399
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.2121212121212122e-06,
+      "loss": 0.338,
+      "step": 4400
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.81,
+      "eval_loss": 0.7186983227729797,
+      "eval_roc_auc": 0.9628144812213908,
+      "eval_runtime": 92.5071,
+      "eval_samples_per_second": 2.162,
+      "eval_steps_per_second": 0.54,
+      "step": 4400
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.2101010101010103e-06,
+      "loss": 0.4591,
+      "step": 4401
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.2080808080808082e-06,
+      "loss": 0.4414,
+      "step": 4402
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.206060606060606e-06,
+      "loss": 0.1827,
+      "step": 4403
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.2040404040404042e-06,
+      "loss": 0.2809,
+      "step": 4404
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.202020202020202e-06,
+      "loss": 0.5791,
+      "step": 4405
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 0.2705,
+      "step": 4406
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 1.197979797979798e-06,
+      "loss": 0.7874,
+      "step": 4407
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 1.1959595959595961e-06,
+      "loss": 0.2787,
+      "step": 4408
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.193939393939394e-06,
+      "loss": 0.2162,
+      "step": 4409
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.1919191919191921e-06,
+      "loss": 0.617,
+      "step": 4410
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.18989898989899e-06,
+      "loss": 0.3853,
+      "step": 4411
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 1.187878787878788e-06,
+      "loss": 0.3301,
+      "step": 4412
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.185858585858586e-06,
+      "loss": 0.6457,
+      "step": 4413
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.183838383838384e-06,
+      "loss": 0.413,
+      "step": 4414
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.181818181818182e-06,
+      "loss": 0.4783,
+      "step": 4415
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 1.1797979797979799e-06,
+      "loss": 0.6358,
+      "step": 4416
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.1777777777777778e-06,
+      "loss": 0.3482,
+      "step": 4417
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.1757575757575759e-06,
+      "loss": 1.006,
+      "step": 4418
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.173737373737374e-06,
+      "loss": 1.0229,
+      "step": 4419
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 1.1717171717171719e-06,
+      "loss": 0.3198,
+      "step": 4420
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.1696969696969697e-06,
+      "loss": 0.3521,
+      "step": 4421
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.1676767676767678e-06,
+      "loss": 0.9403,
+      "step": 4422
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.1656565656565657e-06,
+      "loss": 0.5453,
+      "step": 4423
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.1636363636363638e-06,
+      "loss": 0.582,
+      "step": 4424
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.1616161616161617e-06,
+      "loss": 0.3337,
+      "step": 4425
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.1595959595959596e-06,
+      "loss": 0.4192,
+      "step": 4426
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.1575757575757577e-06,
+      "loss": 0.4459,
+      "step": 4427
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.1555555555555556e-06,
+      "loss": 0.2226,
+      "step": 4428
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.1535353535353535e-06,
+      "loss": 0.3166,
+      "step": 4429
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.1515151515151516e-06,
+      "loss": 0.2758,
+      "step": 4430
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.1494949494949497e-06,
+      "loss": 0.524,
+      "step": 4431
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.1474747474747476e-06,
+      "loss": 0.3656,
+      "step": 4432
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.1454545454545457e-06,
+      "loss": 0.473,
+      "step": 4433
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.1434343434343436e-06,
+      "loss": 0.4439,
+      "step": 4434
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.1414141414141414e-06,
+      "loss": 0.3825,
+      "step": 4435
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.1393939393939395e-06,
+      "loss": 0.8256,
+      "step": 4436
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.1373737373737374e-06,
+      "loss": 0.2057,
+      "step": 4437
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.1353535353535353e-06,
+      "loss": 0.8533,
+      "step": 4438
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.1333333333333334e-06,
+      "loss": 0.4802,
+      "step": 4439
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.1313131313131315e-06,
+      "loss": 0.3295,
+      "step": 4440
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.1292929292929294e-06,
+      "loss": 0.1774,
+      "step": 4441
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.1272727272727275e-06,
+      "loss": 0.5592,
+      "step": 4442
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.1252525252525254e-06,
+      "loss": 0.4255,
+      "step": 4443
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 1.1232323232323233e-06,
+      "loss": 0.3545,
+      "step": 4444
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.1212121212121214e-06,
+      "loss": 0.3502,
+      "step": 4445
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.1191919191919193e-06,
+      "loss": 0.2362,
+      "step": 4446
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.1171717171717172e-06,
+      "loss": 0.5401,
+      "step": 4447
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 1.1151515151515153e-06,
+      "loss": 0.3759,
+      "step": 4448
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.1131313131313131e-06,
+      "loss": 0.3525,
+      "step": 4449
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.111111111111111e-06,
+      "loss": 0.8474,
+      "step": 4450
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.1090909090909093e-06,
+      "loss": 0.7327,
+      "step": 4451
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.1070707070707072e-06,
+      "loss": 0.2642,
+      "step": 4452
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.1050505050505051e-06,
+      "loss": 0.1491,
+      "step": 4453
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.1030303030303032e-06,
+      "loss": 1.4203,
+      "step": 4454
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.1010101010101011e-06,
+      "loss": 0.3401,
+      "step": 4455
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 1.098989898989899e-06,
+      "loss": 0.3628,
+      "step": 4456
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.096969696969697e-06,
+      "loss": 0.5114,
+      "step": 4457
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.094949494949495e-06,
+      "loss": 0.33,
+      "step": 4458
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.0929292929292929e-06,
+      "loss": 1.2345,
+      "step": 4459
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.090909090909091e-06,
+      "loss": 0.8045,
+      "step": 4460
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.0888888888888889e-06,
+      "loss": 0.8681,
+      "step": 4461
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.086868686868687e-06,
+      "loss": 0.2954,
+      "step": 4462
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.084848484848485e-06,
+      "loss": 0.2842,
+      "step": 4463
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.082828282828283e-06,
+      "loss": 1.3224,
+      "step": 4464
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.0808080808080808e-06,
+      "loss": 1.7641,
+      "step": 4465
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.078787878787879e-06,
+      "loss": 0.5423,
+      "step": 4466
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.0767676767676768e-06,
+      "loss": 0.1515,
+      "step": 4467
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.0747474747474747e-06,
+      "loss": 0.3898,
+      "step": 4468
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.0727272727272728e-06,
+      "loss": 0.8136,
+      "step": 4469
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 1.0707070707070707e-06,
+      "loss": 1.3656,
+      "step": 4470
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.0686868686868688e-06,
+      "loss": 0.2912,
+      "step": 4471
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.066666666666667e-06,
+      "loss": 1.0814,
+      "step": 4472
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.0646464646464648e-06,
+      "loss": 0.3703,
+      "step": 4473
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 1.0626262626262627e-06,
+      "loss": 0.4098,
+      "step": 4474
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.0606060606060608e-06,
+      "loss": 0.4113,
+      "step": 4475
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.0585858585858587e-06,
+      "loss": 0.3803,
+      "step": 4476
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.0565656565656566e-06,
+      "loss": 0.4789,
+      "step": 4477
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.0545454545454547e-06,
+      "loss": 0.3142,
+      "step": 4478
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.0525252525252525e-06,
+      "loss": 0.33,
+      "step": 4479
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 1.0505050505050506e-06,
+      "loss": 0.2045,
+      "step": 4480
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.0484848484848485e-06,
+      "loss": 0.2448,
+      "step": 4481
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.0464646464646464e-06,
+      "loss": 0.1916,
+      "step": 4482
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.0444444444444445e-06,
+      "loss": 0.3917,
+      "step": 4483
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 1.0424242424242426e-06,
+      "loss": 0.7668,
+      "step": 4484
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.0404040404040405e-06,
+      "loss": 0.242,
+      "step": 4485
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.0383838383838384e-06,
+      "loss": 0.7566,
+      "step": 4486
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.0363636363636365e-06,
+      "loss": 0.4923,
+      "step": 4487
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 1.0343434343434344e-06,
+      "loss": 0.3101,
+      "step": 4488
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.0323232323232325e-06,
+      "loss": 0.4355,
+      "step": 4489
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.0303030303030304e-06,
+      "loss": 0.3674,
+      "step": 4490
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.0282828282828283e-06,
+      "loss": 0.561,
+      "step": 4491
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 1.0262626262626264e-06,
+      "loss": 0.4148,
+      "step": 4492
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 1.0242424242424242e-06,
+      "loss": 0.2711,
+      "step": 4493
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.0222222222222223e-06,
+      "loss": 0.2458,
+      "step": 4494
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.0202020202020202e-06,
+      "loss": 0.5202,
+      "step": 4495
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.0181818181818183e-06,
+      "loss": 0.7935,
+      "step": 4496
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.0161616161616162e-06,
+      "loss": 0.8125,
+      "step": 4497
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.0141414141414143e-06,
+      "loss": 0.1784,
+      "step": 4498
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.0121212121212122e-06,
+      "loss": 0.6368,
+      "step": 4499
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.01010101010101e-06,
+      "loss": 0.3294,
+      "step": 4500
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.0080808080808082e-06,
+      "loss": 0.3066,
+      "step": 4501
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 1.006060606060606e-06,
+      "loss": 0.4078,
+      "step": 4502
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.004040404040404e-06,
+      "loss": 0.2676,
+      "step": 4503
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.002020202020202e-06,
+      "loss": 0.3933,
+      "step": 4504
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.2403,
+      "step": 4505
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 9.97979797979798e-07,
+      "loss": 0.3644,
+      "step": 4506
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 9.959595959595962e-07,
+      "loss": 0.4716,
+      "step": 4507
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 9.93939393939394e-07,
+      "loss": 0.2818,
+      "step": 4508
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 9.91919191919192e-07,
+      "loss": 0.365,
+      "step": 4509
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 9.8989898989899e-07,
+      "loss": 0.5029,
+      "step": 4510
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 9.87878787878788e-07,
+      "loss": 0.3346,
+      "step": 4511
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 9.858585858585858e-07,
+      "loss": 0.2441,
+      "step": 4512
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 9.83838383838384e-07,
+      "loss": 0.5123,
+      "step": 4513
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 9.818181818181818e-07,
+      "loss": 0.4687,
+      "step": 4514
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 9.797979797979799e-07,
+      "loss": 0.158,
+      "step": 4515
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 9.77777777777778e-07,
+      "loss": 0.3451,
+      "step": 4516
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 9.757575757575759e-07,
+      "loss": 0.3628,
+      "step": 4517
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 9.737373737373738e-07,
+      "loss": 0.4197,
+      "step": 4518
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 9.717171717171719e-07,
+      "loss": 1.025,
+      "step": 4519
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 9.696969696969698e-07,
+      "loss": 0.8114,
+      "step": 4520
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 9.676767676767676e-07,
+      "loss": 0.68,
+      "step": 4521
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 9.656565656565657e-07,
+      "loss": 0.3594,
+      "step": 4522
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 9.636363636363636e-07,
+      "loss": 0.4729,
+      "step": 4523
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 9.616161616161617e-07,
+      "loss": 0.8928,
+      "step": 4524
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 9.595959595959596e-07,
+      "loss": 1.2684,
+      "step": 4525
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 9.575757575757577e-07,
+      "loss": 0.276,
+      "step": 4526
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 9.555555555555556e-07,
+      "loss": 0.2046,
+      "step": 4527
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 9.535353535353536e-07,
+      "loss": 0.2466,
+      "step": 4528
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 9.515151515151516e-07,
+      "loss": 0.3319,
+      "step": 4529
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 9.494949494949496e-07,
+      "loss": 1.098,
+      "step": 4530
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 9.474747474747476e-07,
+      "loss": 0.701,
+      "step": 4531
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 9.454545454545455e-07,
+      "loss": 0.2042,
+      "step": 4532
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 9.434343434343435e-07,
+      "loss": 0.8121,
+      "step": 4533
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 9.414141414141415e-07,
+      "loss": 1.1492,
+      "step": 4534
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 9.393939393939395e-07,
+      "loss": 0.5015,
+      "step": 4535
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 9.373737373737376e-07,
+      "loss": 0.4417,
+      "step": 4536
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 9.353535353535354e-07,
+      "loss": 0.4726,
+      "step": 4537
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 9.333333333333334e-07,
+      "loss": 0.419,
+      "step": 4538
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 9.313131313131314e-07,
+      "loss": 1.1293,
+      "step": 4539
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 9.292929292929294e-07,
+      "loss": 0.2764,
+      "step": 4540
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 9.272727272727273e-07,
+      "loss": 0.3614,
+      "step": 4541
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 9.252525252525253e-07,
+      "loss": 0.5973,
+      "step": 4542
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 9.232323232323233e-07,
+      "loss": 1.1537,
+      "step": 4543
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 9.212121212121213e-07,
+      "loss": 0.8558,
+      "step": 4544
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 9.191919191919192e-07,
+      "loss": 0.3865,
+      "step": 4545
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 9.171717171717172e-07,
+      "loss": 0.2585,
+      "step": 4546
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 9.151515151515153e-07,
+      "loss": 1.0628,
+      "step": 4547
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 9.131313131313133e-07,
+      "loss": 0.5561,
+      "step": 4548
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 9.111111111111113e-07,
+      "loss": 0.7181,
+      "step": 4549
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 9.090909090909091e-07,
+      "loss": 0.4416,
+      "step": 4550
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 9.070707070707071e-07,
+      "loss": 0.777,
+      "step": 4551
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 9.050505050505051e-07,
+      "loss": 0.4878,
+      "step": 4552
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 9.030303030303031e-07,
+      "loss": 0.6176,
+      "step": 4553
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 9.01010101010101e-07,
+      "loss": 0.3547,
+      "step": 4554
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 8.98989898989899e-07,
+      "loss": 0.3956,
+      "step": 4555
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 8.96969696969697e-07,
+      "loss": 0.2774,
+      "step": 4556
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 8.94949494949495e-07,
+      "loss": 0.3818,
+      "step": 4557
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 8.929292929292931e-07,
+      "loss": 0.3526,
+      "step": 4558
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 8.90909090909091e-07,
+      "loss": 0.3796,
+      "step": 4559
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 8.88888888888889e-07,
+      "loss": 0.7635,
+      "step": 4560
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 8.86868686868687e-07,
+      "loss": 0.3559,
+      "step": 4561
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 8.84848484848485e-07,
+      "loss": 0.1398,
+      "step": 4562
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 8.828282828282829e-07,
+      "loss": 0.2777,
+      "step": 4563
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 8.808080808080808e-07,
+      "loss": 0.9383,
+      "step": 4564
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 8.787878787878788e-07,
+      "loss": 0.3346,
+      "step": 4565
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 8.767676767676768e-07,
+      "loss": 0.2609,
+      "step": 4566
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 8.747474747474747e-07,
+      "loss": 0.6132,
+      "step": 4567
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 8.727272727272728e-07,
+      "loss": 0.6554,
+      "step": 4568
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 8.707070707070708e-07,
+      "loss": 0.3781,
+      "step": 4569
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 8.686868686868688e-07,
+      "loss": 0.6242,
+      "step": 4570
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 8.666666666666668e-07,
+      "loss": 0.3984,
+      "step": 4571
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 8.646464646464647e-07,
+      "loss": 0.2641,
+      "step": 4572
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 8.626262626262627e-07,
+      "loss": 1.4136,
+      "step": 4573
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 8.606060606060607e-07,
+      "loss": 0.8319,
+      "step": 4574
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 8.585858585858587e-07,
+      "loss": 0.2358,
+      "step": 4575
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 8.565656565656566e-07,
+      "loss": 0.9366,
+      "step": 4576
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 8.545454545454546e-07,
+      "loss": 1.0232,
+      "step": 4577
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 8.525252525252525e-07,
+      "loss": 1.5422,
+      "step": 4578
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 8.505050505050506e-07,
+      "loss": 0.2296,
+      "step": 4579
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 8.484848484848486e-07,
+      "loss": 0.9208,
+      "step": 4580
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 8.464646464646465e-07,
+      "loss": 0.5351,
+      "step": 4581
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 8.444444444444445e-07,
+      "loss": 0.3157,
+      "step": 4582
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 8.424242424242425e-07,
+      "loss": 0.3136,
+      "step": 4583
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 8.404040404040405e-07,
+      "loss": 0.5321,
+      "step": 4584
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 8.383838383838384e-07,
+      "loss": 0.263,
+      "step": 4585
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 8.363636363636364e-07,
+      "loss": 0.269,
+      "step": 4586
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 8.343434343434344e-07,
+      "loss": 0.2722,
+      "step": 4587
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 8.323232323232324e-07,
+      "loss": 0.1446,
+      "step": 4588
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 8.303030303030303e-07,
+      "loss": 0.3423,
+      "step": 4589
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 8.282828282828284e-07,
+      "loss": 0.2732,
+      "step": 4590
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 8.262626262626264e-07,
+      "loss": 0.193,
+      "step": 4591
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 8.242424242424244e-07,
+      "loss": 0.2133,
+      "step": 4592
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 8.222222222222223e-07,
+      "loss": 0.6048,
+      "step": 4593
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 8.202020202020202e-07,
+      "loss": 0.3434,
+      "step": 4594
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 8.181818181818182e-07,
+      "loss": 0.3575,
+      "step": 4595
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 8.161616161616162e-07,
+      "loss": 0.5279,
+      "step": 4596
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 8.141414141414142e-07,
+      "loss": 0.127,
+      "step": 4597
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 8.121212121212121e-07,
+      "loss": 0.3994,
+      "step": 4598
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 8.101010101010101e-07,
+      "loss": 0.6721,
+      "step": 4599
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 8.080808080808082e-07,
+      "loss": 0.396,
+      "step": 4600
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.83,
+      "eval_loss": 0.7169148921966553,
+      "eval_roc_auc": 0.9607189615954717,
+      "eval_runtime": 92.8289,
+      "eval_samples_per_second": 2.155,
+      "eval_steps_per_second": 0.539,
+      "step": 4600
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 8.060606060606062e-07,
+      "loss": 0.6146,
+      "step": 4601
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 8.040404040404042e-07,
+      "loss": 0.1971,
+      "step": 4602
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 8.020202020202021e-07,
+      "loss": 0.299,
+      "step": 4603
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 0.1944,
+      "step": 4604
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 7.979797979797981e-07,
+      "loss": 0.4059,
+      "step": 4605
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 7.959595959595961e-07,
+      "loss": 0.3654,
+      "step": 4606
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 7.939393939393939e-07,
+      "loss": 0.3313,
+      "step": 4607
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 7.919191919191919e-07,
+      "loss": 0.6201,
+      "step": 4608
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 7.898989898989899e-07,
+      "loss": 0.1771,
+      "step": 4609
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 7.878787878787879e-07,
+      "loss": 0.3175,
+      "step": 4610
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 7.85858585858586e-07,
+      "loss": 0.2754,
+      "step": 4611
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 7.838383838383839e-07,
+      "loss": 0.4776,
+      "step": 4612
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 7.818181818181819e-07,
+      "loss": 0.5186,
+      "step": 4613
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 7.797979797979799e-07,
+      "loss": 0.3446,
+      "step": 4614
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 7.777777777777779e-07,
+      "loss": 0.3148,
+      "step": 4615
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 7.757575757575758e-07,
+      "loss": 0.2859,
+      "step": 4616
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 7.737373737373738e-07,
+      "loss": 0.4826,
+      "step": 4617
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 7.717171717171718e-07,
+      "loss": 1.1932,
+      "step": 4618
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 7.696969696969698e-07,
+      "loss": 0.2815,
+      "step": 4619
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 7.676767676767677e-07,
+      "loss": 0.5692,
+      "step": 4620
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 7.656565656565656e-07,
+      "loss": 0.3012,
+      "step": 4621
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 7.636363636363637e-07,
+      "loss": 0.2438,
+      "step": 4622
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 7.616161616161617e-07,
+      "loss": 0.2414,
+      "step": 4623
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 7.595959595959597e-07,
+      "loss": 0.7091,
+      "step": 4624
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 7.575757575757576e-07,
+      "loss": 0.5168,
+      "step": 4625
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 7.555555555555556e-07,
+      "loss": 0.4344,
+      "step": 4626
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 7.535353535353536e-07,
+      "loss": 0.98,
+      "step": 4627
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 7.515151515151516e-07,
+      "loss": 0.2847,
+      "step": 4628
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 7.494949494949495e-07,
+      "loss": 0.5656,
+      "step": 4629
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 7.474747474747475e-07,
+      "loss": 1.1117,
+      "step": 4630
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 7.454545454545455e-07,
+      "loss": 0.1675,
+      "step": 4631
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 7.434343434343436e-07,
+      "loss": 0.3008,
+      "step": 4632
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 7.414141414141416e-07,
+      "loss": 0.7984,
+      "step": 4633
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 7.393939393939395e-07,
+      "loss": 0.3012,
+      "step": 4634
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 7.373737373737375e-07,
+      "loss": 0.3627,
+      "step": 4635
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 7.353535353535354e-07,
+      "loss": 0.3622,
+      "step": 4636
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 7.333333333333334e-07,
+      "loss": 0.1921,
+      "step": 4637
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 7.313131313131313e-07,
+      "loss": 0.5695,
+      "step": 4638
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 7.292929292929293e-07,
+      "loss": 0.5245,
+      "step": 4639
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 7.272727272727273e-07,
+      "loss": 0.4387,
+      "step": 4640
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 7.252525252525253e-07,
+      "loss": 0.2909,
+      "step": 4641
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 7.232323232323232e-07,
+      "loss": 0.781,
+      "step": 4642
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 7.212121212121213e-07,
+      "loss": 0.8559,
+      "step": 4643
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 7.191919191919193e-07,
+      "loss": 0.5107,
+      "step": 4644
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 7.171717171717173e-07,
+      "loss": 0.2148,
+      "step": 4645
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 7.151515151515153e-07,
+      "loss": 0.4149,
+      "step": 4646
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 7.131313131313132e-07,
+      "loss": 0.1596,
+      "step": 4647
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 7.111111111111112e-07,
+      "loss": 0.5986,
+      "step": 4648
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 7.090909090909092e-07,
+      "loss": 0.4399,
+      "step": 4649
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 7.070707070707071e-07,
+      "loss": 0.3236,
+      "step": 4650
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 7.05050505050505e-07,
+      "loss": 0.455,
+      "step": 4651
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 7.03030303030303e-07,
+      "loss": 0.3278,
+      "step": 4652
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 7.01010101010101e-07,
+      "loss": 0.3412,
+      "step": 4653
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 6.989898989898991e-07,
+      "loss": 0.5789,
+      "step": 4654
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 6.969696969696971e-07,
+      "loss": 0.4545,
+      "step": 4655
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 6.94949494949495e-07,
+      "loss": 0.3797,
+      "step": 4656
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 6.92929292929293e-07,
+      "loss": 0.6235,
+      "step": 4657
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 6.90909090909091e-07,
+      "loss": 0.546,
+      "step": 4658
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 6.88888888888889e-07,
+      "loss": 0.6259,
+      "step": 4659
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 6.868686868686869e-07,
+      "loss": 1.4432,
+      "step": 4660
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 6.848484848484849e-07,
+      "loss": 1.3421,
+      "step": 4661
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 6.828282828282829e-07,
+      "loss": 0.1916,
+      "step": 4662
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 6.808080808080809e-07,
+      "loss": 0.2978,
+      "step": 4663
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 6.78787878787879e-07,
+      "loss": 0.3879,
+      "step": 4664
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 6.767676767676768e-07,
+      "loss": 0.4096,
+      "step": 4665
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 6.747474747474748e-07,
+      "loss": 0.4811,
+      "step": 4666
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 6.727272727272728e-07,
+      "loss": 0.4242,
+      "step": 4667
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 6.707070707070708e-07,
+      "loss": 0.3431,
+      "step": 4668
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 6.686868686868687e-07,
+      "loss": 0.2928,
+      "step": 4669
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 6.666666666666667e-07,
+      "loss": 0.4993,
+      "step": 4670
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 6.646464646464647e-07,
+      "loss": 0.5192,
+      "step": 4671
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 6.626262626262627e-07,
+      "loss": 0.5908,
+      "step": 4672
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 6.606060606060606e-07,
+      "loss": 0.6425,
+      "step": 4673
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 6.585858585858586e-07,
+      "loss": 0.1526,
+      "step": 4674
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 6.565656565656567e-07,
+      "loss": 0.6356,
+      "step": 4675
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 6.545454545454547e-07,
+      "loss": 0.4345,
+      "step": 4676
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 6.525252525252527e-07,
+      "loss": 0.3182,
+      "step": 4677
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 6.505050505050506e-07,
+      "loss": 0.3019,
+      "step": 4678
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 6.484848484848485e-07,
+      "loss": 0.2431,
+      "step": 4679
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 6.464646464646465e-07,
+      "loss": 0.3497,
+      "step": 4680
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 6.444444444444445e-07,
+      "loss": 0.2735,
+      "step": 4681
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 6.424242424242424e-07,
+      "loss": 0.144,
+      "step": 4682
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 6.404040404040404e-07,
+      "loss": 0.3967,
+      "step": 4683
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 6.383838383838384e-07,
+      "loss": 0.8158,
+      "step": 4684
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 6.363636363636364e-07,
+      "loss": 0.3846,
+      "step": 4685
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 6.343434343434345e-07,
+      "loss": 0.2244,
+      "step": 4686
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 6.323232323232324e-07,
+      "loss": 0.5899,
+      "step": 4687
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 6.303030303030304e-07,
+      "loss": 0.3412,
+      "step": 4688
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 6.282828282828284e-07,
+      "loss": 0.2263,
+      "step": 4689
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 6.262626262626264e-07,
+      "loss": 0.7081,
+      "step": 4690
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 6.242424242424243e-07,
+      "loss": 1.3778,
+      "step": 4691
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 6.222222222222223e-07,
+      "loss": 0.2564,
+      "step": 4692
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 6.202020202020202e-07,
+      "loss": 0.5493,
+      "step": 4693
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 6.181818181818182e-07,
+      "loss": 0.3615,
+      "step": 4694
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 6.161616161616162e-07,
+      "loss": 0.9259,
+      "step": 4695
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 6.141414141414142e-07,
+      "loss": 0.296,
+      "step": 4696
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 6.121212121212121e-07,
+      "loss": 0.4,
+      "step": 4697
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 6.101010101010101e-07,
+      "loss": 0.6358,
+      "step": 4698
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 6.080808080808082e-07,
+      "loss": 0.4168,
+      "step": 4699
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 6.060606060606061e-07,
+      "loss": 0.3457,
+      "step": 4700
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 6.040404040404041e-07,
+      "loss": 0.1963,
+      "step": 4701
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 6.020202020202021e-07,
+      "loss": 0.8838,
+      "step": 4702
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 0.5465,
+      "step": 4703
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 5.979797979797981e-07,
+      "loss": 0.3071,
+      "step": 4704
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 5.959595959595961e-07,
+      "loss": 0.4069,
+      "step": 4705
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 5.93939393939394e-07,
+      "loss": 0.4741,
+      "step": 4706
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 5.91919191919192e-07,
+      "loss": 0.6781,
+      "step": 4707
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 5.898989898989899e-07,
+      "loss": 0.4332,
+      "step": 4708
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 5.878787878787879e-07,
+      "loss": 0.8455,
+      "step": 4709
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 5.858585858585859e-07,
+      "loss": 0.7735,
+      "step": 4710
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 5.838383838383839e-07,
+      "loss": 0.9008,
+      "step": 4711
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 5.818181818181819e-07,
+      "loss": 0.4906,
+      "step": 4712
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 5.797979797979798e-07,
+      "loss": 0.1224,
+      "step": 4713
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 5.777777777777778e-07,
+      "loss": 0.5762,
+      "step": 4714
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 5.757575757575758e-07,
+      "loss": 0.2658,
+      "step": 4715
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 5.737373737373738e-07,
+      "loss": 0.785,
+      "step": 4716
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 5.717171717171718e-07,
+      "loss": 0.2717,
+      "step": 4717
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 5.696969696969698e-07,
+      "loss": 0.4288,
+      "step": 4718
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 5.676767676767677e-07,
+      "loss": 0.2608,
+      "step": 4719
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 5.656565656565658e-07,
+      "loss": 0.5308,
+      "step": 4720
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 5.636363636363638e-07,
+      "loss": 0.3853,
+      "step": 4721
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 5.616161616161616e-07,
+      "loss": 1.0474,
+      "step": 4722
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 5.595959595959596e-07,
+      "loss": 0.7256,
+      "step": 4723
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 5.575757575757576e-07,
+      "loss": 0.3514,
+      "step": 4724
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 5.555555555555555e-07,
+      "loss": 0.3906,
+      "step": 4725
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 5.535353535353536e-07,
+      "loss": 0.3344,
+      "step": 4726
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 5.515151515151516e-07,
+      "loss": 0.414,
+      "step": 4727
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 5.494949494949495e-07,
+      "loss": 0.4061,
+      "step": 4728
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 5.474747474747475e-07,
+      "loss": 0.332,
+      "step": 4729
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 5.454545454545455e-07,
+      "loss": 0.4582,
+      "step": 4730
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 5.434343434343435e-07,
+      "loss": 0.2856,
+      "step": 4731
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 5.414141414141415e-07,
+      "loss": 0.7637,
+      "step": 4732
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 5.393939393939395e-07,
+      "loss": 0.2674,
+      "step": 4733
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 5.373737373737374e-07,
+      "loss": 0.2672,
+      "step": 4734
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 5.353535353535354e-07,
+      "loss": 0.4251,
+      "step": 4735
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 5.333333333333335e-07,
+      "loss": 0.217,
+      "step": 4736
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 5.313131313131313e-07,
+      "loss": 0.7414,
+      "step": 4737
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 5.292929292929293e-07,
+      "loss": 0.3848,
+      "step": 4738
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 5.272727272727273e-07,
+      "loss": 0.3652,
+      "step": 4739
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 5.252525252525253e-07,
+      "loss": 0.4284,
+      "step": 4740
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 5.232323232323232e-07,
+      "loss": 0.6244,
+      "step": 4741
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 5.212121212121213e-07,
+      "loss": 0.5224,
+      "step": 4742
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 5.191919191919192e-07,
+      "loss": 0.4187,
+      "step": 4743
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 5.171717171717172e-07,
+      "loss": 0.4476,
+      "step": 4744
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 5.151515151515152e-07,
+      "loss": 0.6788,
+      "step": 4745
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 5.131313131313132e-07,
+      "loss": 0.2879,
+      "step": 4746
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 5.111111111111112e-07,
+      "loss": 0.1839,
+      "step": 4747
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 5.090909090909092e-07,
+      "loss": 0.2468,
+      "step": 4748
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 5.070707070707072e-07,
+      "loss": 0.3701,
+      "step": 4749
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 5.05050505050505e-07,
+      "loss": 0.6393,
+      "step": 4750
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 5.03030303030303e-07,
+      "loss": 0.5873,
+      "step": 4751
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 5.01010101010101e-07,
+      "loss": 0.3021,
+      "step": 4752
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 4.98989898989899e-07,
+      "loss": 0.7746,
+      "step": 4753
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 4.96969696969697e-07,
+      "loss": 0.4171,
+      "step": 4754
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 4.94949494949495e-07,
+      "loss": 0.3597,
+      "step": 4755
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 4.929292929292929e-07,
+      "loss": 0.4565,
+      "step": 4756
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 4.909090909090909e-07,
+      "loss": 0.3338,
+      "step": 4757
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 4.88888888888889e-07,
+      "loss": 0.7302,
+      "step": 4758
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 4.868686868686869e-07,
+      "loss": 0.4327,
+      "step": 4759
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 4.848484848484849e-07,
+      "loss": 0.7678,
+      "step": 4760
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 4.828282828282829e-07,
+      "loss": 0.3472,
+      "step": 4761
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 4.808080808080809e-07,
+      "loss": 0.3182,
+      "step": 4762
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 4.787878787878789e-07,
+      "loss": 1.122,
+      "step": 4763
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 4.767676767676768e-07,
+      "loss": 0.2847,
+      "step": 4764
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 4.747474747474748e-07,
+      "loss": 0.7227,
+      "step": 4765
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 4.7272727272727273e-07,
+      "loss": 0.4097,
+      "step": 4766
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 4.7070707070707073e-07,
+      "loss": 0.5734,
+      "step": 4767
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 4.686868686868688e-07,
+      "loss": 0.5291,
+      "step": 4768
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 4.666666666666667e-07,
+      "loss": 0.4362,
+      "step": 4769
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 4.646464646464647e-07,
+      "loss": 0.3506,
+      "step": 4770
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 4.6262626262626265e-07,
+      "loss": 0.4118,
+      "step": 4771
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 4.6060606060606064e-07,
+      "loss": 0.333,
+      "step": 4772
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 4.585858585858586e-07,
+      "loss": 0.4819,
+      "step": 4773
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 4.5656565656565663e-07,
+      "loss": 0.4394,
+      "step": 4774
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 4.5454545454545457e-07,
+      "loss": 0.3272,
+      "step": 4775
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 4.5252525252525257e-07,
+      "loss": 0.7761,
+      "step": 4776
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 4.505050505050505e-07,
+      "loss": 0.3399,
+      "step": 4777
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 4.484848484848485e-07,
+      "loss": 0.7369,
+      "step": 4778
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 4.4646464646464655e-07,
+      "loss": 0.4508,
+      "step": 4779
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 4.444444444444445e-07,
+      "loss": 0.6545,
+      "step": 4780
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 4.424242424242425e-07,
+      "loss": 0.5403,
+      "step": 4781
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 4.404040404040404e-07,
+      "loss": 0.2428,
+      "step": 4782
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 4.383838383838384e-07,
+      "loss": 0.3244,
+      "step": 4783
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 4.363636363636364e-07,
+      "loss": 0.3366,
+      "step": 4784
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 4.343434343434344e-07,
+      "loss": 1.6965,
+      "step": 4785
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 4.3232323232323235e-07,
+      "loss": 0.4108,
+      "step": 4786
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 4.3030303030303034e-07,
+      "loss": 0.7825,
+      "step": 4787
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 4.282828282828283e-07,
+      "loss": 0.1301,
+      "step": 4788
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 4.262626262626263e-07,
+      "loss": 0.4976,
+      "step": 4789
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 4.242424242424243e-07,
+      "loss": 0.3649,
+      "step": 4790
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 4.2222222222222226e-07,
+      "loss": 0.3113,
+      "step": 4791
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 4.2020202020202026e-07,
+      "loss": 0.5202,
+      "step": 4792
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 4.181818181818182e-07,
+      "loss": 0.4123,
+      "step": 4793
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 4.161616161616162e-07,
+      "loss": 0.6343,
+      "step": 4794
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 4.141414141414142e-07,
+      "loss": 0.2378,
+      "step": 4795
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 4.121212121212122e-07,
+      "loss": 0.4056,
+      "step": 4796
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 4.101010101010101e-07,
+      "loss": 0.2493,
+      "step": 4797
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 4.080808080808081e-07,
+      "loss": 0.5583,
+      "step": 4798
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 4.0606060606060605e-07,
+      "loss": 0.6179,
+      "step": 4799
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 4.040404040404041e-07,
+      "loss": 0.6795,
+      "step": 4800
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.815,
+      "eval_loss": 0.7037800550460815,
+      "eval_roc_auc": 0.9641373496197048,
+      "eval_runtime": 92.5831,
+      "eval_samples_per_second": 2.16,
+      "eval_steps_per_second": 0.54,
+      "step": 4800
+    }
+  ],
+  "max_steps": 5000,
+  "num_train_epochs": 25,
+  "total_flos": 5.229406089216e+18,
+  "trial_name": null,
+  "trial_params": null
+}