diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,165690 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 27610,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4125452352231605e-08,
+      "loss": 1.3055,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.825090470446321e-08,
+      "loss": 1.379,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.237635705669482e-08,
+      "loss": 0.8578,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.650180940892642e-08,
+      "loss": 1.4045,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2062726176115803e-07,
+      "loss": 1.3131,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4475271411338965e-07,
+      "loss": 0.8293,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6887816646562124e-07,
+      "loss": 1.3868,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9300361881785284e-07,
+      "loss": 1.3745,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1712907117008446e-07,
+      "loss": 0.8854,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4125452352231605e-07,
+      "loss": 1.3646,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.653799758745477e-07,
+      "loss": 1.3652,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.895054282267793e-07,
+      "loss": 1.2571,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1363088057901086e-07,
+      "loss": 1.3392,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.377563329312425e-07,
+      "loss": 1.3745,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.618817852834741e-07,
+      "loss": 1.3476,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.860072376357057e-07,
+      "loss": 1.3578,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.101326899879373e-07,
+      "loss": 1.3055,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.342581423401689e-07,
+      "loss": 1.3303,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.583835946924005e-07,
+      "loss": 1.3297,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.825090470446321e-07,
+      "loss": 1.35,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.066344993968638e-07,
+      "loss": 1.2326,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.307599517490953e-07,
+      "loss": 1.2598,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.548854041013269e-07,
+      "loss": 1.1889,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.790108564535586e-07,
+      "loss": 0.8147,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.031363088057902e-07,
+      "loss": 1.2777,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.272617611580217e-07,
+      "loss": 0.8219,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.513872135102534e-07,
+      "loss": 0.8472,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.75512665862485e-07,
+      "loss": 1.2081,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.996381182147166e-07,
+      "loss": 1.1534,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.237635705669482e-07,
+      "loss": 1.1812,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.478890229191799e-07,
+      "loss": 0.844,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.720144752714113e-07,
+      "loss": 1.1486,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.96139927623643e-07,
+      "loss": 1.1726,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.202653799758746e-07,
+      "loss": 1.1394,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.443908323281063e-07,
+      "loss": 1.1531,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.685162846803378e-07,
+      "loss": 1.1856,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.926417370325695e-07,
+      "loss": 0.8188,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.16767189384801e-07,
+      "loss": 1.1556,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.408926417370326e-07,
+      "loss": 0.8733,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.650180940892642e-07,
+      "loss": 0.8309,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.891435464414958e-07,
+      "loss": 1.168,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0132689987937276e-06,
+      "loss": 1.0721,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0373944511459591e-06,
+      "loss": 1.0915,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0615199034981907e-06,
+      "loss": 1.1587,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0856453558504223e-06,
+      "loss": 1.1088,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1097708082026538e-06,
+      "loss": 1.136,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1338962605548854e-06,
+      "loss": 1.1551,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1580217129071172e-06,
+      "loss": 1.0728,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1821471652593488e-06,
+      "loss": 1.0713,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2062726176115803e-06,
+      "loss": 0.8233,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2303980699638119e-06,
+      "loss": 0.8008,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2545235223160435e-06,
+      "loss": 1.1019,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.278648974668275e-06,
+      "loss": 0.8165,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3027744270205068e-06,
+      "loss": 0.8187,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3268998793727384e-06,
+      "loss": 1.0602,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.35102533172497e-06,
+      "loss": 1.0196,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3751507840772015e-06,
+      "loss": 0.7693,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3992762364294333e-06,
+      "loss": 0.805,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4234016887816649e-06,
+      "loss": 1.0946,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4475271411338964e-06,
+      "loss": 1.0239,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4716525934861282e-06,
+      "loss": 1.0772,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4957780458383598e-06,
+      "loss": 1.1011,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5199034981905911e-06,
+      "loss": 1.065,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5440289505428227e-06,
+      "loss": 0.7441,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5681544028950543e-06,
+      "loss": 1.0849,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.592279855247286e-06,
+      "loss": 1.0588,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6164053075995176e-06,
+      "loss": 1.0502,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6405307599517492e-06,
+      "loss": 1.0773,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6646562123039808e-06,
+      "loss": 1.0979,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6887816646562125e-06,
+      "loss": 1.0414,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.712907117008444e-06,
+      "loss": 0.9946,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7370325693606757e-06,
+      "loss": 1.0433,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7611580217129072e-06,
+      "loss": 1.0047,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.785283474065139e-06,
+      "loss": 1.0061,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8094089264173706e-06,
+      "loss": 0.9901,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.833534378769602e-06,
+      "loss": 1.0129,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8576598311218335e-06,
+      "loss": 1.055,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8817852834740653e-06,
+      "loss": 1.0031,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9059107358262969e-06,
+      "loss": 1.0281,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9300361881785284e-06,
+      "loss": 1.0158,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.95416164053076e-06,
+      "loss": 1.0256,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9782870928829916e-06,
+      "loss": 1.0424,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0024125452352233e-06,
+      "loss": 1.0328,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.026537997587455e-06,
+      "loss": 0.9915,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0506634499396865e-06,
+      "loss": 0.7781,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0747889022919183e-06,
+      "loss": 0.9638,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.09891435464415e-06,
+      "loss": 0.9436,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1230398069963814e-06,
+      "loss": 0.7256,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.147165259348613e-06,
+      "loss": 1.005,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1712907117008445e-06,
+      "loss": 0.8806,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.195416164053076e-06,
+      "loss": 1.0223,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2195416164053077e-06,
+      "loss": 1.0246,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2436670687575394e-06,
+      "loss": 0.98,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.267792521109771e-06,
+      "loss": 0.9615,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2919179734620026e-06,
+      "loss": 0.9704,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3160434258142344e-06,
+      "loss": 0.7567,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3401688781664657e-06,
+      "loss": 1.0298,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3642943305186975e-06,
+      "loss": 0.7088,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3884197828709293e-06,
+      "loss": 0.977,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4125452352231606e-06,
+      "loss": 0.7426,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4366706875753924e-06,
+      "loss": 0.7408,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4607961399276238e-06,
+      "loss": 0.9949,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4849215922798556e-06,
+      "loss": 0.9989,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.509047044632087e-06,
+      "loss": 0.7804,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5331724969843187e-06,
+      "loss": 0.967,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.55729794933655e-06,
+      "loss": 0.7332,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.581423401688782e-06,
+      "loss": 0.775,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6055488540410136e-06,
+      "loss": 0.9174,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.629674306393245e-06,
+      "loss": 0.7216,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6537997587454767e-06,
+      "loss": 0.8885,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.677925211097708e-06,
+      "loss": 0.965,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.70205066344994e-06,
+      "loss": 0.9455,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7261761158021717e-06,
+      "loss": 0.9843,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.750301568154403e-06,
+      "loss": 0.7422,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.774427020506635e-06,
+      "loss": 0.7028,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7985524728588666e-06,
+      "loss": 0.6772,
+      "step": 116
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.822677925211098e-06,
+      "loss": 0.9812,
+      "step": 117
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8468033775633297e-06,
+      "loss": 0.7641,
+      "step": 118
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8709288299155615e-06,
+      "loss": 0.9397,
+      "step": 119
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.895054282267793e-06,
+      "loss": 1.0375,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9191797346200246e-06,
+      "loss": 0.9322,
+      "step": 121
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9433051869722564e-06,
+      "loss": 0.9444,
+      "step": 122
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9674306393244878e-06,
+      "loss": 0.9686,
+      "step": 123
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9915560916767195e-06,
+      "loss": 0.9055,
+      "step": 124
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0156815440289505e-06,
+      "loss": 0.7226,
+      "step": 125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0398069963811823e-06,
+      "loss": 0.9501,
+      "step": 126
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0639324487334136e-06,
+      "loss": 0.9245,
+      "step": 127
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0880579010856454e-06,
+      "loss": 0.8966,
+      "step": 128
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.112183353437877e-06,
+      "loss": 0.9864,
+      "step": 129
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1363088057901085e-06,
+      "loss": 0.8771,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1604342581423403e-06,
+      "loss": 0.7058,
+      "step": 131
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.184559710494572e-06,
+      "loss": 0.943,
+      "step": 132
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2086851628468034e-06,
+      "loss": 0.9846,
+      "step": 133
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2328106151990352e-06,
+      "loss": 0.7221,
+      "step": 134
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2569360675512666e-06,
+      "loss": 0.9015,
+      "step": 135
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2810615199034984e-06,
+      "loss": 0.904,
+      "step": 136
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.30518697225573e-06,
+      "loss": 0.7505,
+      "step": 137
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3293124246079615e-06,
+      "loss": 0.9532,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3534378769601933e-06,
+      "loss": 0.8701,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.377563329312425e-06,
+      "loss": 0.9275,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4016887816646564e-06,
+      "loss": 0.6845,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.425814234016888e-06,
+      "loss": 0.9367,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.44993968636912e-06,
+      "loss": 0.9529,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4740651387213513e-06,
+      "loss": 0.7582,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.498190591073583e-06,
+      "loss": 0.9384,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5223160434258145e-06,
+      "loss": 0.9091,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5464414957780463e-06,
+      "loss": 0.929,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.570566948130278e-06,
+      "loss": 0.907,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5946924004825094e-06,
+      "loss": 0.7506,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.618817852834741e-06,
+      "loss": 0.9283,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.642943305186973e-06,
+      "loss": 0.7408,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.667068757539204e-06,
+      "loss": 0.9284,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6911942098914357e-06,
+      "loss": 0.9416,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.715319662243667e-06,
+      "loss": 0.9411,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.739445114595899e-06,
+      "loss": 0.9184,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7635705669481306e-06,
+      "loss": 0.8607,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.787696019300362e-06,
+      "loss": 0.6785,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8118214716525937e-06,
+      "loss": 0.9108,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.835946924004825e-06,
+      "loss": 0.706,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.860072376357057e-06,
+      "loss": 0.8988,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.884197828709289e-06,
+      "loss": 0.7342,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.90832328106152e-06,
+      "loss": 0.9321,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.932448733413752e-06,
+      "loss": 0.741,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.956574185765983e-06,
+      "loss": 0.909,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.980699638118215e-06,
+      "loss": 0.6973,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.004825090470447e-06,
+      "loss": 0.8741,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0289505428226785e-06,
+      "loss": 0.8483,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.05307599517491e-06,
+      "loss": 0.8711,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.077201447527141e-06,
+      "loss": 0.8304,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.101326899879373e-06,
+      "loss": 0.9319,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.125452352231605e-06,
+      "loss": 0.857,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1495778045838365e-06,
+      "loss": 0.7126,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.173703256936068e-06,
+      "loss": 0.8977,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1978287092883e-06,
+      "loss": 0.7341,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.221954161640531e-06,
+      "loss": 0.9176,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.246079613992763e-06,
+      "loss": 0.934,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2702050663449946e-06,
+      "loss": 0.7577,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.294330518697226e-06,
+      "loss": 0.9383,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.318455971049458e-06,
+      "loss": 0.697,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.342581423401689e-06,
+      "loss": 0.9918,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.366706875753921e-06,
+      "loss": 0.9703,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.390832328106152e-06,
+      "loss": 0.7463,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4149577804583836e-06,
+      "loss": 0.9126,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.439083232810615e-06,
+      "loss": 0.912,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.463208685162847e-06,
+      "loss": 0.8612,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.487334137515079e-06,
+      "loss": 0.9181,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.511459589867311e-06,
+      "loss": 0.9279,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.535585042219542e-06,
+      "loss": 0.9291,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.559710494571773e-06,
+      "loss": 0.9188,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.583835946924005e-06,
+      "loss": 0.8863,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.607961399276237e-06,
+      "loss": 0.8642,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.632086851628469e-06,
+      "loss": 0.8362,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6562123039807e-06,
+      "loss": 0.9184,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6803377563329314e-06,
+      "loss": 0.9298,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.704463208685163e-06,
+      "loss": 0.919,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.728588661037395e-06,
+      "loss": 0.8966,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.752714113389627e-06,
+      "loss": 0.8619,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7768395657418586e-06,
+      "loss": 0.9002,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8009650180940895e-06,
+      "loss": 0.8421,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.825090470446321e-06,
+      "loss": 0.8174,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.849215922798553e-06,
+      "loss": 0.6578,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.873341375150785e-06,
+      "loss": 0.931,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.897466827503017e-06,
+      "loss": 0.8923,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9215922798552475e-06,
+      "loss": 0.6687,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.945717732207479e-06,
+      "loss": 0.8682,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.969843184559711e-06,
+      "loss": 0.8592,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.993968636911943e-06,
+      "loss": 0.91,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.018094089264174e-06,
+      "loss": 0.9201,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.042219541616406e-06,
+      "loss": 0.883,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.066344993968637e-06,
+      "loss": 0.8924,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.090470446320868e-06,
+      "loss": 0.8062,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.1145958986731e-06,
+      "loss": 0.8401,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.138721351025332e-06,
+      "loss": 0.9088,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.162846803377564e-06,
+      "loss": 0.8771,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.1869722557297954e-06,
+      "loss": 0.9684,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.211097708082027e-06,
+      "loss": 0.6762,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.235223160434258e-06,
+      "loss": 0.9443,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.25934861278649e-06,
+      "loss": 0.7039,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.283474065138722e-06,
+      "loss": 0.9017,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.3075995174909535e-06,
+      "loss": 0.8491,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.331724969843185e-06,
+      "loss": 0.8652,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.355850422195416e-06,
+      "loss": 0.8939,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.379975874547648e-06,
+      "loss": 0.8768,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.40410132689988e-06,
+      "loss": 0.8599,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.4282267792521115e-06,
+      "loss": 0.8731,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.452352231604343e-06,
+      "loss": 0.8654,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.476477683956575e-06,
+      "loss": 0.8439,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.500603136308806e-06,
+      "loss": 0.8065,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.524728588661038e-06,
+      "loss": 0.9244,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.54885404101327e-06,
+      "loss": 0.9297,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.572979493365501e-06,
+      "loss": 0.8861,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.597104945717733e-06,
+      "loss": 0.9003,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.621230398069965e-06,
+      "loss": 0.712,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.645355850422196e-06,
+      "loss": 0.7536,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.669481302774428e-06,
+      "loss": 0.8076,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.6936067551266594e-06,
+      "loss": 0.8785,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.717732207478891e-06,
+      "loss": 0.8496,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.741857659831123e-06,
+      "loss": 0.7121,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.765983112183354e-06,
+      "loss": 0.925,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.790108564535586e-06,
+      "loss": 0.9201,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.8142340168878175e-06,
+      "loss": 0.6665,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.838359469240049e-06,
+      "loss": 0.689,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.862484921592281e-06,
+      "loss": 0.8738,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.886610373944513e-06,
+      "loss": 0.7023,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.910735826296744e-06,
+      "loss": 0.8067,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.9348612786489755e-06,
+      "loss": 0.6573,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.958986731001207e-06,
+      "loss": 0.6661,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.983112183353439e-06,
+      "loss": 0.9117,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.007237635705669e-06,
+      "loss": 0.8289,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.031363088057901e-06,
+      "loss": 0.672,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.055488540410133e-06,
+      "loss": 0.6892,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.0796139927623645e-06,
+      "loss": 0.8632,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.103739445114596e-06,
+      "loss": 0.7066,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.127864897466827e-06,
+      "loss": 0.656,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.151990349819059e-06,
+      "loss": 0.6213,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.176115802171291e-06,
+      "loss": 0.6714,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.2002412545235226e-06,
+      "loss": 0.7219,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.224366706875754e-06,
+      "loss": 0.8771,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.248492159227985e-06,
+      "loss": 0.8722,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.272617611580217e-06,
+      "loss": 0.6801,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.296743063932449e-06,
+      "loss": 0.9147,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.320868516284681e-06,
+      "loss": 0.8194,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.344993968636912e-06,
+      "loss": 0.6904,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.369119420989144e-06,
+      "loss": 0.8899,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.393244873341375e-06,
+      "loss": 0.8882,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.417370325693607e-06,
+      "loss": 0.8669,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.441495778045839e-06,
+      "loss": 0.9362,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.4656212303980705e-06,
+      "loss": 0.6796,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.489746682750302e-06,
+      "loss": 0.8417,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.513872135102533e-06,
+      "loss": 0.6993,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.537997587454765e-06,
+      "loss": 0.9396,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.562123039806997e-06,
+      "loss": 0.8483,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.5862484921592285e-06,
+      "loss": 0.7001,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.61037394451146e-06,
+      "loss": 0.8391,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.634499396863692e-06,
+      "loss": 0.7972,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.658624849215923e-06,
+      "loss": 0.8369,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.682750301568155e-06,
+      "loss": 0.8657,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.7068757539203866e-06,
+      "loss": 0.6859,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.731001206272618e-06,
+      "loss": 0.8867,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.75512665862485e-06,
+      "loss": 0.8327,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.779252110977081e-06,
+      "loss": 0.8931,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.803377563329313e-06,
+      "loss": 0.8716,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.827503015681545e-06,
+      "loss": 0.83,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.851628468033776e-06,
+      "loss": 0.8684,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.875753920386008e-06,
+      "loss": 0.7072,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.89987937273824e-06,
+      "loss": 0.8967,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.924004825090471e-06,
+      "loss": 0.8657,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.948130277442703e-06,
+      "loss": 0.8673,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.9722557297949344e-06,
+      "loss": 0.7069,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.996381182147166e-06,
+      "loss": 0.6204,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.020506634499398e-06,
+      "loss": 0.6521,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.044632086851629e-06,
+      "loss": 0.8672,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.068757539203861e-06,
+      "loss": 0.6905,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.0928829915560925e-06,
+      "loss": 0.8569,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.117008443908324e-06,
+      "loss": 0.8682,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.141133896260556e-06,
+      "loss": 0.8511,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.165259348612788e-06,
+      "loss": 0.7971,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.189384800965019e-06,
+      "loss": 0.8412,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.2135102533172506e-06,
+      "loss": 0.8658,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.237635705669482e-06,
+      "loss": 0.836,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.261761158021714e-06,
+      "loss": 0.8827,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.285886610373946e-06,
+      "loss": 0.6793,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.310012062726177e-06,
+      "loss": 0.8497,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.334137515078408e-06,
+      "loss": 0.7787,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.3582629674306395e-06,
+      "loss": 0.8404,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.382388419782871e-06,
+      "loss": 0.6452,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.406513872135102e-06,
+      "loss": 0.8478,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.430639324487334e-06,
+      "loss": 0.8624,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.454764776839566e-06,
+      "loss": 0.6996,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.478890229191798e-06,
+      "loss": 0.6923,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.503015681544029e-06,
+      "loss": 0.6442,
+      "step": 311
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.527141133896261e-06,
+      "loss": 0.8686,
+      "step": 312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.551266586248492e-06,
+      "loss": 0.6471,
+      "step": 313
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.575392038600724e-06,
+      "loss": 0.7992,
+      "step": 314
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.599517490952956e-06,
+      "loss": 0.8858,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.6236429433051874e-06,
+      "loss": 0.6827,
+      "step": 316
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.647768395657418e-06,
+      "loss": 0.6674,
+      "step": 317
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.67189384800965e-06,
+      "loss": 0.6876,
+      "step": 318
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.696019300361882e-06,
+      "loss": 0.6493,
+      "step": 319
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.720144752714114e-06,
+      "loss": 0.6496,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.744270205066345e-06,
+      "loss": 0.6511,
+      "step": 321
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.768395657418577e-06,
+      "loss": 0.7758,
+      "step": 322
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.792521109770809e-06,
+      "loss": 0.7274,
+      "step": 323
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.81664656212304e-06,
+      "loss": 0.7498,
+      "step": 324
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.840772014475273e-06,
+      "loss": 0.7108,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.864897466827504e-06,
+      "loss": 0.8373,
+      "step": 326
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.889022919179734e-06,
+      "loss": 0.6571,
+      "step": 327
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.913148371531966e-06,
+      "loss": 0.8669,
+      "step": 328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.937273823884198e-06,
+      "loss": 0.6277,
+      "step": 329
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.96139927623643e-06,
+      "loss": 0.8826,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.985524728588662e-06,
+      "loss": 0.8967,
+      "step": 331
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.009650180940893e-06,
+      "loss": 0.7915,
+      "step": 332
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.033775633293125e-06,
+      "loss": 0.8368,
+      "step": 333
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.057901085645357e-06,
+      "loss": 0.685,
+      "step": 334
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.082026537997589e-06,
+      "loss": 0.851,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.10615199034982e-06,
+      "loss": 0.8075,
+      "step": 336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.130277442702052e-06,
+      "loss": 0.7978,
+      "step": 337
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.154402895054282e-06,
+      "loss": 0.7976,
+      "step": 338
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.178528347406514e-06,
+      "loss": 0.6428,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.202653799758746e-06,
+      "loss": 0.877,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.226779252110978e-06,
+      "loss": 0.6469,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.25090470446321e-06,
+      "loss": 0.7505,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.275030156815441e-06,
+      "loss": 0.7591,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.299155609167673e-06,
+      "loss": 0.8112,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.323281061519905e-06,
+      "loss": 0.6889,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.347406513872137e-06,
+      "loss": 0.8441,
+      "step": 346
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.371531966224368e-06,
+      "loss": 0.8706,
+      "step": 347
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.3956574185766e-06,
+      "loss": 0.8549,
+      "step": 348
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.41978287092883e-06,
+      "loss": 0.6416,
+      "step": 349
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.443908323281062e-06,
+      "loss": 0.8275,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.468033775633294e-06,
+      "loss": 0.8602,
+      "step": 351
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.492159227985526e-06,
+      "loss": 0.6742,
+      "step": 352
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.516284680337757e-06,
+      "loss": 0.8361,
+      "step": 353
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.540410132689989e-06,
+      "loss": 0.7165,
+      "step": 354
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.564535585042221e-06,
+      "loss": 0.7915,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.588661037394453e-06,
+      "loss": 0.7821,
+      "step": 356
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.612786489746684e-06,
+      "loss": 0.7901,
+      "step": 357
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.636911942098916e-06,
+      "loss": 0.8723,
+      "step": 358
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.661037394451148e-06,
+      "loss": 0.8553,
+      "step": 359
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.685162846803378e-06,
+      "loss": 0.6636,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.70928829915561e-06,
+      "loss": 0.6891,
+      "step": 361
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.733413751507842e-06,
+      "loss": 0.8543,
+      "step": 362
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.757539203860072e-06,
+      "loss": 0.6716,
+      "step": 363
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.781664656212304e-06,
+      "loss": 0.8326,
+      "step": 364
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.805790108564535e-06,
+      "loss": 0.8171,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.829915560916767e-06,
+      "loss": 0.8746,
+      "step": 366
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.854041013268999e-06,
+      "loss": 0.8086,
+      "step": 367
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.87816646562123e-06,
+      "loss": 0.8564,
+      "step": 368
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.902291917973462e-06,
+      "loss": 0.6642,
+      "step": 369
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.926417370325694e-06,
+      "loss": 0.842,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.950542822677926e-06,
+      "loss": 0.8914,
+      "step": 371
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.974668275030158e-06,
+      "loss": 0.7905,
+      "step": 372
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.99879372738239e-06,
+      "loss": 0.6054,
+      "step": 373
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.022919179734621e-06,
+      "loss": 0.8372,
+      "step": 374
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.047044632086851e-06,
+      "loss": 0.6674,
+      "step": 375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.071170084439083e-06,
+      "loss": 0.8695,
+      "step": 376
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.095295536791315e-06,
+      "loss": 0.6821,
+      "step": 377
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.119420989143547e-06,
+      "loss": 0.844,
+      "step": 378
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.143546441495779e-06,
+      "loss": 0.8125,
+      "step": 379
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.16767189384801e-06,
+      "loss": 0.7615,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.191797346200242e-06,
+      "loss": 0.8573,
+      "step": 381
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.215922798552474e-06,
+      "loss": 0.7876,
+      "step": 382
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.240048250904706e-06,
+      "loss": 0.8332,
+      "step": 383
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.264173703256937e-06,
+      "loss": 0.8135,
+      "step": 384
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.28829915560917e-06,
+      "loss": 0.8646,
+      "step": 385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.3124246079614e-06,
+      "loss": 0.8727,
+      "step": 386
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.336550060313631e-06,
+      "loss": 0.8476,
+      "step": 387
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.360675512665863e-06,
+      "loss": 0.8216,
+      "step": 388
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.384800965018095e-06,
+      "loss": 0.8165,
+      "step": 389
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.408926417370326e-06,
+      "loss": 0.7923,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.433051869722558e-06,
+      "loss": 0.8764,
+      "step": 391
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.45717732207479e-06,
+      "loss": 0.7985,
+      "step": 392
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.481302774427022e-06,
+      "loss": 0.8179,
+      "step": 393
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.505428226779254e-06,
+      "loss": 0.8664,
+      "step": 394
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.529553679131485e-06,
+      "loss": 0.8762,
+      "step": 395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.553679131483717e-06,
+      "loss": 0.7506,
+      "step": 396
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.577804583835947e-06,
+      "loss": 0.8389,
+      "step": 397
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.601930036188179e-06,
+      "loss": 0.832,
+      "step": 398
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.62605548854041e-06,
+      "loss": 0.7742,
+      "step": 399
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.650180940892643e-06,
+      "loss": 0.7887,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.674306393244874e-06,
+      "loss": 0.6611,
+      "step": 401
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.698431845597106e-06,
+      "loss": 0.5767,
+      "step": 402
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.722557297949338e-06,
+      "loss": 0.8761,
+      "step": 403
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.74668275030157e-06,
+      "loss": 0.675,
+      "step": 404
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.770808202653801e-06,
+      "loss": 0.8582,
+      "step": 405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.794933655006033e-06,
+      "loss": 0.6879,
+      "step": 406
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.819059107358265e-06,
+      "loss": 0.9291,
+      "step": 407
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.843184559710495e-06,
+      "loss": 0.7653,
+      "step": 408
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.867310012062727e-06,
+      "loss": 0.8636,
+      "step": 409
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.891435464414959e-06,
+      "loss": 0.7518,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.91556091676719e-06,
+      "loss": 0.8413,
+      "step": 411
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.939686369119422e-06,
+      "loss": 0.85,
+      "step": 412
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.963811821471654e-06,
+      "loss": 0.6916,
+      "step": 413
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.987937273823886e-06,
+      "loss": 0.7868,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0012062726176118e-05,
+      "loss": 0.7757,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0036188178528348e-05,
+      "loss": 0.6743,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0060313630880581e-05,
+      "loss": 0.6547,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0084439083232811e-05,
+      "loss": 0.6646,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0108564535585043e-05,
+      "loss": 0.8273,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0132689987937275e-05,
+      "loss": 0.7736,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0156815440289507e-05,
+      "loss": 0.8106,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0180940892641737e-05,
+      "loss": 0.7745,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.020506634499397e-05,
+      "loss": 0.6106,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.02291917973462e-05,
+      "loss": 0.7845,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0253317249698434e-05,
+      "loss": 0.6013,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0277442702050664e-05,
+      "loss": 0.7797,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0301568154402897e-05,
+      "loss": 0.6573,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0325693606755127e-05,
+      "loss": 0.7643,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.034981905910736e-05,
+      "loss": 0.8065,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0373944511459591e-05,
+      "loss": 0.8027,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0398069963811823e-05,
+      "loss": 0.8223,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0422195416164054e-05,
+      "loss": 0.8688,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0446320868516286e-05,
+      "loss": 0.8394,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0470446320868516e-05,
+      "loss": 0.6942,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.049457177322075e-05,
+      "loss": 0.7984,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.051869722557298e-05,
+      "loss": 0.8382,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0542822677925213e-05,
+      "loss": 0.8541,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0566948130277443e-05,
+      "loss": 0.7603,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0591073582629677e-05,
+      "loss": 0.6384,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0615199034981907e-05,
+      "loss": 0.6431,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0639324487334139e-05,
+      "loss": 0.8294,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.066344993968637e-05,
+      "loss": 0.7742,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.06875753920386e-05,
+      "loss": 0.8612,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0711700844390832e-05,
+      "loss": 0.8237,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0735826296743064e-05,
+      "loss": 0.6103,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0759951749095296e-05,
+      "loss": 0.79,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0784077201447528e-05,
+      "loss": 0.7707,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.080820265379976e-05,
+      "loss": 0.8694,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.083232810615199e-05,
+      "loss": 0.7999,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0856453558504223e-05,
+      "loss": 0.9064,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0880579010856453e-05,
+      "loss": 0.6175,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0904704463208687e-05,
+      "loss": 0.7676,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0928829915560917e-05,
+      "loss": 0.7868,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.095295536791315e-05,
+      "loss": 0.8487,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.097708082026538e-05,
+      "loss": 0.8755,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1001206272617612e-05,
+      "loss": 0.8104,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1025331724969844e-05,
+      "loss": 0.641,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1049457177322076e-05,
+      "loss": 0.6526,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1073582629674306e-05,
+      "loss": 0.8491,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.109770808202654e-05,
+      "loss": 0.6467,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.112183353437877e-05,
+      "loss": 0.8284,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1145958986731003e-05,
+      "loss": 0.8215,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1170084439083233e-05,
+      "loss": 0.828,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1194209891435466e-05,
+      "loss": 0.761,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1218335343787696e-05,
+      "loss": 0.6377,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.124246079613993e-05,
+      "loss": 0.797,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.126658624849216e-05,
+      "loss": 0.7978,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1290711700844392e-05,
+      "loss": 0.6367,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1314837153196624e-05,
+      "loss": 0.7936,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1338962605548855e-05,
+      "loss": 0.6742,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1363088057901085e-05,
+      "loss": 0.9057,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1387213510253319e-05,
+      "loss": 0.6458,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1411338962605549e-05,
+      "loss": 0.809,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1435464414957782e-05,
+      "loss": 0.6515,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1459589867310012e-05,
+      "loss": 0.8545,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1483715319662246e-05,
+      "loss": 0.6503,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1507840772014476e-05,
+      "loss": 0.6643,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1531966224366708e-05,
+      "loss": 0.892,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.155609167671894e-05,
+      "loss": 0.6344,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1580217129071171e-05,
+      "loss": 0.8712,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1604342581423401e-05,
+      "loss": 0.8199,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1628468033775635e-05,
+      "loss": 0.656,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1652593486127865e-05,
+      "loss": 0.7934,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1676718938480099e-05,
+      "loss": 0.865,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1700844390832329e-05,
+      "loss": 0.7835,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1724969843184562e-05,
+      "loss": 0.7618,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1749095295536792e-05,
+      "loss": 0.8307,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1773220747889026e-05,
+      "loss": 0.7546,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1797346200241256e-05,
+      "loss": 0.7885,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1821471652593488e-05,
+      "loss": 0.8456,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.184559710494572e-05,
+      "loss": 0.6672,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1869722557297951e-05,
+      "loss": 0.7815,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1893848009650181e-05,
+      "loss": 0.8181,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1917973462002415e-05,
+      "loss": 0.7917,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1942098914354645e-05,
+      "loss": 0.8305,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1966224366706878e-05,
+      "loss": 0.8329,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1990349819059108e-05,
+      "loss": 0.7975,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2014475271411338e-05,
+      "loss": 0.6393,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2038600723763572e-05,
+      "loss": 0.7733,
+      "step": 499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2062726176115802e-05,
+      "loss": 0.8208,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2086851628468035e-05,
+      "loss": 0.8169,
+      "step": 501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2110977080820265e-05,
+      "loss": 0.82,
+      "step": 502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2135102533172497e-05,
+      "loss": 0.7933,
+      "step": 503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2159227985524729e-05,
+      "loss": 0.8875,
+      "step": 504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.218335343787696e-05,
+      "loss": 0.8394,
+      "step": 505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2207478890229193e-05,
+      "loss": 0.8045,
+      "step": 506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2231604342581424e-05,
+      "loss": 0.7965,
+      "step": 507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2255729794933654e-05,
+      "loss": 0.8007,
+      "step": 508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2279855247285888e-05,
+      "loss": 0.7712,
+      "step": 509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2303980699638118e-05,
+      "loss": 0.8088,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2328106151990351e-05,
+      "loss": 0.812,
+      "step": 511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2352231604342582e-05,
+      "loss": 0.8223,
+      "step": 512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2376357056694815e-05,
+      "loss": 0.7979,
+      "step": 513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2400482509047045e-05,
+      "loss": 0.7271,
+      "step": 514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2424607961399277e-05,
+      "loss": 0.6456,
+      "step": 515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2448733413751509e-05,
+      "loss": 0.6004,
+      "step": 516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.247285886610374e-05,
+      "loss": 0.7942,
+      "step": 517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.249698431845597e-05,
+      "loss": 0.6684,
+      "step": 518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2521109770808204e-05,
+      "loss": 0.8989,
+      "step": 519
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2545235223160434e-05,
+      "loss": 0.8193,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2569360675512668e-05,
+      "loss": 0.6066,
+      "step": 521
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2593486127864898e-05,
+      "loss": 0.7723,
+      "step": 522
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2617611580217131e-05,
+      "loss": 0.8531,
+      "step": 523
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2641737032569361e-05,
+      "loss": 0.8066,
+      "step": 524
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2665862484921593e-05,
+      "loss": 0.7662,
+      "step": 525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2689987937273825e-05,
+      "loss": 0.6377,
+      "step": 526
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2714113389626057e-05,
+      "loss": 0.644,
+      "step": 527
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2738238841978288e-05,
+      "loss": 0.836,
+      "step": 528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.276236429433052e-05,
+      "loss": 0.6469,
+      "step": 529
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.278648974668275e-05,
+      "loss": 0.7608,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2810615199034984e-05,
+      "loss": 0.8169,
+      "step": 531
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2834740651387214e-05,
+      "loss": 0.7657,
+      "step": 532
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2858866103739447e-05,
+      "loss": 0.6209,
+      "step": 533
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2882991556091677e-05,
+      "loss": 0.8246,
+      "step": 534
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.290711700844391e-05,
+      "loss": 0.7375,
+      "step": 535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2931242460796141e-05,
+      "loss": 0.7668,
+      "step": 536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2955367913148373e-05,
+      "loss": 0.7869,
+      "step": 537
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2979493365500604e-05,
+      "loss": 0.7882,
+      "step": 538
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3003618817852836e-05,
+      "loss": 0.835,
+      "step": 539
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3027744270205066e-05,
+      "loss": 0.5977,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.30518697225573e-05,
+      "loss": 0.7254,
+      "step": 541
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.307599517490953e-05,
+      "loss": 0.8005,
+      "step": 542
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3100120627261763e-05,
+      "loss": 0.7229,
+      "step": 543
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3124246079613993e-05,
+      "loss": 0.7612,
+      "step": 544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3148371531966227e-05,
+      "loss": 0.8082,
+      "step": 545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3172496984318457e-05,
+      "loss": 0.8815,
+      "step": 546
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.319662243667069e-05,
+      "loss": 0.7933,
+      "step": 547
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.322074788902292e-05,
+      "loss": 0.6306,
+      "step": 548
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3244873341375152e-05,
+      "loss": 0.8084,
+      "step": 549
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3268998793727384e-05,
+      "loss": 0.6152,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3293124246079616e-05,
+      "loss": 0.5983,
+      "step": 551
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3317249698431846e-05,
+      "loss": 0.6605,
+      "step": 552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3341375150784078e-05,
+      "loss": 0.6198,
+      "step": 553
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.336550060313631e-05,
+      "loss": 0.7481,
+      "step": 554
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.338962605548854e-05,
+      "loss": 0.79,
+      "step": 555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3413751507840773e-05,
+      "loss": 0.8117,
+      "step": 556
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3437876960193003e-05,
+      "loss": 0.8068,
+      "step": 557
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3462002412545237e-05,
+      "loss": 0.7415,
+      "step": 558
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3486127864897467e-05,
+      "loss": 0.7958,
+      "step": 559
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.35102533172497e-05,
+      "loss": 0.8529,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.353437876960193e-05,
+      "loss": 0.6355,
+      "step": 561
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3558504221954162e-05,
+      "loss": 0.666,
+      "step": 562
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3582629674306394e-05,
+      "loss": 0.6521,
+      "step": 563
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3606755126658626e-05,
+      "loss": 0.7912,
+      "step": 564
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3630880579010857e-05,
+      "loss": 0.8121,
+      "step": 565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.365500603136309e-05,
+      "loss": 0.6734,
+      "step": 566
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.367913148371532e-05,
+      "loss": 0.773,
+      "step": 567
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3703256936067553e-05,
+      "loss": 0.782,
+      "step": 568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3727382388419783e-05,
+      "loss": 0.8046,
+      "step": 569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3751507840772016e-05,
+      "loss": 0.8232,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3775633293124246e-05,
+      "loss": 0.7921,
+      "step": 571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.379975874547648e-05,
+      "loss": 0.6379,
+      "step": 572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.382388419782871e-05,
+      "loss": 0.8146,
+      "step": 573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3848009650180942e-05,
+      "loss": 0.7809,
+      "step": 574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3872135102533174e-05,
+      "loss": 0.7664,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3896260554885405e-05,
+      "loss": 0.7602,
+      "step": 576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3920386007237635e-05,
+      "loss": 0.8063,
+      "step": 577
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3944511459589869e-05,
+      "loss": 0.7904,
+      "step": 578
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3968636911942099e-05,
+      "loss": 0.7769,
+      "step": 579
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3992762364294332e-05,
+      "loss": 0.8555,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4016887816646563e-05,
+      "loss": 0.8008,
+      "step": 581
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4041013268998796e-05,
+      "loss": 0.6153,
+      "step": 582
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4065138721351026e-05,
+      "loss": 0.8211,
+      "step": 583
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4089264173703258e-05,
+      "loss": 0.7983,
+      "step": 584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.411338962605549e-05,
+      "loss": 0.7832,
+      "step": 585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4137515078407721e-05,
+      "loss": 0.8039,
+      "step": 586
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4161640530759953e-05,
+      "loss": 0.7549,
+      "step": 587
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4185765983112185e-05,
+      "loss": 0.8208,
+      "step": 588
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4209891435464415e-05,
+      "loss": 0.8146,
+      "step": 589
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4234016887816649e-05,
+      "loss": 0.8132,
+      "step": 590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4258142340168879e-05,
+      "loss": 0.6343,
+      "step": 591
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4282267792521112e-05,
+      "loss": 0.6706,
+      "step": 592
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4306393244873342e-05,
+      "loss": 0.799,
+      "step": 593
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4330518697225576e-05,
+      "loss": 0.8845,
+      "step": 594
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4354644149577806e-05,
+      "loss": 0.7632,
+      "step": 595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4378769601930038e-05,
+      "loss": 0.7627,
+      "step": 596
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.440289505428227e-05,
+      "loss": 0.6638,
+      "step": 597
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4427020506634501e-05,
+      "loss": 0.787,
+      "step": 598
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4451145958986731e-05,
+      "loss": 0.8175,
+      "step": 599
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4475271411338965e-05,
+      "loss": 0.6553,
+      "step": 600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4499396863691195e-05,
+      "loss": 0.7687,
+      "step": 601
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4523522316043428e-05,
+      "loss": 0.773,
+      "step": 602
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4547647768395658e-05,
+      "loss": 0.816,
+      "step": 603
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4571773220747892e-05,
+      "loss": 0.829,
+      "step": 604
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4595898673100122e-05,
+      "loss": 0.7145,
+      "step": 605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4620024125452354e-05,
+      "loss": 0.7652,
+      "step": 606
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4644149577804585e-05,
+      "loss": 0.7446,
+      "step": 607
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4668275030156816e-05,
+      "loss": 0.8108,
+      "step": 608
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4692400482509049e-05,
+      "loss": 0.6878,
+      "step": 609
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4716525934861279e-05,
+      "loss": 0.8119,
+      "step": 610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4740651387213511e-05,
+      "loss": 0.8446,
+      "step": 611
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4764776839565743e-05,
+      "loss": 0.7785,
+      "step": 612
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4788902291917974e-05,
+      "loss": 0.6823,
+      "step": 613
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4813027744270205e-05,
+      "loss": 0.7815,
+      "step": 614
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4837153196622438e-05,
+      "loss": 0.7789,
+      "step": 615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4861278648974668e-05,
+      "loss": 0.6426,
+      "step": 616
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4885404101326902e-05,
+      "loss": 0.8204,
+      "step": 617
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4909529553679132e-05,
+      "loss": 0.8096,
+      "step": 618
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4933655006031365e-05,
+      "loss": 0.7929,
+      "step": 619
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4957780458383595e-05,
+      "loss": 0.8346,
+      "step": 620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4981905910735827e-05,
+      "loss": 0.6487,
+      "step": 621
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5006031363088059e-05,
+      "loss": 0.8023,
+      "step": 622
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.503015681544029e-05,
+      "loss": 0.835,
+      "step": 623
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5054282267792522e-05,
+      "loss": 0.8529,
+      "step": 624
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5078407720144754e-05,
+      "loss": 0.696,
+      "step": 625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5102533172496984e-05,
+      "loss": 0.7985,
+      "step": 626
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5126658624849218e-05,
+      "loss": 0.8069,
+      "step": 627
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5150784077201448e-05,
+      "loss": 0.7907,
+      "step": 628
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5174909529553681e-05,
+      "loss": 0.8392,
+      "step": 629
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5199034981905911e-05,
+      "loss": 0.8573,
+      "step": 630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5223160434258145e-05,
+      "loss": 0.7865,
+      "step": 631
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5247285886610375e-05,
+      "loss": 0.8252,
+      "step": 632
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5271411338962607e-05,
+      "loss": 0.7786,
+      "step": 633
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5295536791314837e-05,
+      "loss": 0.8063,
+      "step": 634
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.531966224366707e-05,
+      "loss": 0.6177,
+      "step": 635
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.53437876960193e-05,
+      "loss": 0.7412,
+      "step": 636
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5367913148371534e-05,
+      "loss": 0.7934,
+      "step": 637
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5392038600723764e-05,
+      "loss": 0.6223,
+      "step": 638
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5416164053075997e-05,
+      "loss": 0.8441,
+      "step": 639
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5440289505428227e-05,
+      "loss": 0.8032,
+      "step": 640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.546441495778046e-05,
+      "loss": 0.8037,
+      "step": 641
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.548854041013269e-05,
+      "loss": 0.7847,
+      "step": 642
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5512665862484924e-05,
+      "loss": 0.8238,
+      "step": 643
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5536791314837155e-05,
+      "loss": 0.8294,
+      "step": 644
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5560916767189388e-05,
+      "loss": 0.8341,
+      "step": 645
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5585042219541618e-05,
+      "loss": 0.7938,
+      "step": 646
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.560916767189385e-05,
+      "loss": 0.7728,
+      "step": 647
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.563329312424608e-05,
+      "loss": 0.7058,
+      "step": 648
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5657418576598312e-05,
+      "loss": 0.8194,
+      "step": 649
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5681544028950545e-05,
+      "loss": 0.8065,
+      "step": 650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5705669481302775e-05,
+      "loss": 0.7409,
+      "step": 651
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.572979493365501e-05,
+      "loss": 0.7868,
+      "step": 652
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.575392038600724e-05,
+      "loss": 0.8784,
+      "step": 653
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.577804583835947e-05,
+      "loss": 0.7146,
+      "step": 654
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5802171290711702e-05,
+      "loss": 0.8044,
+      "step": 655
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5826296743063932e-05,
+      "loss": 0.8066,
+      "step": 656
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5850422195416166e-05,
+      "loss": 0.8077,
+      "step": 657
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5874547647768396e-05,
+      "loss": 0.6248,
+      "step": 658
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.589867310012063e-05,
+      "loss": 0.6478,
+      "step": 659
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.592279855247286e-05,
+      "loss": 0.7778,
+      "step": 660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5946924004825093e-05,
+      "loss": 0.8215,
+      "step": 661
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5971049457177323e-05,
+      "loss": 0.774,
+      "step": 662
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5995174909529557e-05,
+      "loss": 0.8598,
+      "step": 663
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6019300361881787e-05,
+      "loss": 0.6702,
+      "step": 664
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6043425814234017e-05,
+      "loss": 0.7749,
+      "step": 665
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.606755126658625e-05,
+      "loss": 0.5755,
+      "step": 666
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.609167671893848e-05,
+      "loss": 0.5826,
+      "step": 667
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6115802171290714e-05,
+      "loss": 0.7978,
+      "step": 668
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6139927623642944e-05,
+      "loss": 0.8221,
+      "step": 669
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6164053075995177e-05,
+      "loss": 0.7248,
+      "step": 670
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6188178528347407e-05,
+      "loss": 0.8002,
+      "step": 671
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.621230398069964e-05,
+      "loss": 0.8149,
+      "step": 672
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.623642943305187e-05,
+      "loss": 0.8596,
+      "step": 673
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6260554885404105e-05,
+      "loss": 0.6726,
+      "step": 674
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6284680337756335e-05,
+      "loss": 0.8287,
+      "step": 675
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6308805790108565e-05,
+      "loss": 0.8417,
+      "step": 676
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6332931242460798e-05,
+      "loss": 0.7535,
+      "step": 677
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6357056694813028e-05,
+      "loss": 0.6638,
+      "step": 678
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.638118214716526e-05,
+      "loss": 0.6551,
+      "step": 679
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6405307599517492e-05,
+      "loss": 0.6274,
+      "step": 680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6429433051869722e-05,
+      "loss": 0.7472,
+      "step": 681
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6453558504221955e-05,
+      "loss": 0.6546,
+      "step": 682
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6477683956574185e-05,
+      "loss": 0.8127,
+      "step": 683
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.650180940892642e-05,
+      "loss": 0.8111,
+      "step": 684
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.652593486127865e-05,
+      "loss": 0.6483,
+      "step": 685
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6550060313630883e-05,
+      "loss": 0.8029,
+      "step": 686
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6574185765983113e-05,
+      "loss": 0.8531,
+      "step": 687
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6598311218335346e-05,
+      "loss": 0.6238,
+      "step": 688
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6622436670687576e-05,
+      "loss": 0.8452,
+      "step": 689
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.664656212303981e-05,
+      "loss": 0.8267,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.667068757539204e-05,
+      "loss": 0.7346,
+      "step": 691
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6694813027744273e-05,
+      "loss": 0.8141,
+      "step": 692
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6718938480096503e-05,
+      "loss": 0.7486,
+      "step": 693
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6743063932448737e-05,
+      "loss": 0.6398,
+      "step": 694
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6767189384800967e-05,
+      "loss": 0.8174,
+      "step": 695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.67913148371532e-05,
+      "loss": 0.8669,
+      "step": 696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.681544028950543e-05,
+      "loss": 0.794,
+      "step": 697
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.683956574185766e-05,
+      "loss": 0.837,
+      "step": 698
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6863691194209894e-05,
+      "loss": 0.6482,
+      "step": 699
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6887816646562124e-05,
+      "loss": 0.7801,
+      "step": 700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6911942098914354e-05,
+      "loss": 0.7273,
+      "step": 701
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6936067551266588e-05,
+      "loss": 0.7962,
+      "step": 702
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6960193003618818e-05,
+      "loss": 0.7727,
+      "step": 703
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.698431845597105e-05,
+      "loss": 0.7723,
+      "step": 704
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.700844390832328e-05,
+      "loss": 0.775,
+      "step": 705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7032569360675515e-05,
+      "loss": 0.6898,
+      "step": 706
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7056694813027745e-05,
+      "loss": 0.8271,
+      "step": 707
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7080820265379978e-05,
+      "loss": 0.804,
+      "step": 708
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.710494571773221e-05,
+      "loss": 0.7577,
+      "step": 709
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7129071170084442e-05,
+      "loss": 0.7986,
+      "step": 710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7153196622436672e-05,
+      "loss": 0.6648,
+      "step": 711
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7177322074788905e-05,
+      "loss": 0.636,
+      "step": 712
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7201447527141135e-05,
+      "loss": 0.7948,
+      "step": 713
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.722557297949337e-05,
+      "loss": 0.7614,
+      "step": 714
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.72496984318456e-05,
+      "loss": 0.6686,
+      "step": 715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7273823884197833e-05,
+      "loss": 0.742,
+      "step": 716
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7297949336550063e-05,
+      "loss": 0.7292,
+      "step": 717
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7322074788902296e-05,
+      "loss": 0.6533,
+      "step": 718
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7346200241254526e-05,
+      "loss": 0.6642,
+      "step": 719
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7370325693606756e-05,
+      "loss": 0.8288,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.739445114595899e-05,
+      "loss": 0.5822,
+      "step": 721
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.741857659831122e-05,
+      "loss": 0.6377,
+      "step": 722
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.744270205066345e-05,
+      "loss": 0.845,
+      "step": 723
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7466827503015683e-05,
+      "loss": 0.6917,
+      "step": 724
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7490952955367913e-05,
+      "loss": 0.8367,
+      "step": 725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7515078407720144e-05,
+      "loss": 0.7892,
+      "step": 726
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7539203860072377e-05,
+      "loss": 0.7463,
+      "step": 727
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7563329312424607e-05,
+      "loss": 0.806,
+      "step": 728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.758745476477684e-05,
+      "loss": 0.8486,
+      "step": 729
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.761158021712907e-05,
+      "loss": 0.7707,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7635705669481304e-05,
+      "loss": 0.648,
+      "step": 731
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7659831121833534e-05,
+      "loss": 0.8117,
+      "step": 732
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7683956574185768e-05,
+      "loss": 0.6437,
+      "step": 733
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7708082026537998e-05,
+      "loss": 0.6519,
+      "step": 734
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.773220747889023e-05,
+      "loss": 0.7481,
+      "step": 735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.775633293124246e-05,
+      "loss": 0.8263,
+      "step": 736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7780458383594695e-05,
+      "loss": 0.8542,
+      "step": 737
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7804583835946925e-05,
+      "loss": 0.7636,
+      "step": 738
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.782870928829916e-05,
+      "loss": 0.7992,
+      "step": 739
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.785283474065139e-05,
+      "loss": 0.7532,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7876960193003622e-05,
+      "loss": 0.7056,
+      "step": 741
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7901085645355852e-05,
+      "loss": 0.8601,
+      "step": 742
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7925211097708085e-05,
+      "loss": 0.8196,
+      "step": 743
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7949336550060316e-05,
+      "loss": 0.6466,
+      "step": 744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7973462002412546e-05,
+      "loss": 0.8253,
+      "step": 745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.799758745476478e-05,
+      "loss": 0.8175,
+      "step": 746
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.802171290711701e-05,
+      "loss": 0.7454,
+      "step": 747
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8045838359469243e-05,
+      "loss": 0.8698,
+      "step": 748
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8069963811821473e-05,
+      "loss": 0.7722,
+      "step": 749
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8094089264173703e-05,
+      "loss": 0.7956,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8118214716525936e-05,
+      "loss": 0.6601,
+      "step": 751
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8142340168878166e-05,
+      "loss": 0.8281,
+      "step": 752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.81664656212304e-05,
+      "loss": 0.7701,
+      "step": 753
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.819059107358263e-05,
+      "loss": 0.7953,
+      "step": 754
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8214716525934863e-05,
+      "loss": 0.642,
+      "step": 755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8238841978287094e-05,
+      "loss": 0.6697,
+      "step": 756
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8262967430639327e-05,
+      "loss": 0.7718,
+      "step": 757
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8287092882991557e-05,
+      "loss": 0.6828,
+      "step": 758
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.831121833534379e-05,
+      "loss": 0.763,
+      "step": 759
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.833534378769602e-05,
+      "loss": 0.7875,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8359469240048254e-05,
+      "loss": 0.8539,
+      "step": 761
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8383594692400484e-05,
+      "loss": 0.5879,
+      "step": 762
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8407720144752718e-05,
+      "loss": 0.8964,
+      "step": 763
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8431845597104948e-05,
+      "loss": 0.8118,
+      "step": 764
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.845597104945718e-05,
+      "loss": 0.8247,
+      "step": 765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.848009650180941e-05,
+      "loss": 0.7486,
+      "step": 766
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.850422195416164e-05,
+      "loss": 0.6163,
+      "step": 767
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8528347406513875e-05,
+      "loss": 0.8325,
+      "step": 768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8552472858866105e-05,
+      "loss": 0.7901,
+      "step": 769
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.857659831121834e-05,
+      "loss": 0.8147,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.860072376357057e-05,
+      "loss": 0.7819,
+      "step": 771
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.86248492159228e-05,
+      "loss": 0.8451,
+      "step": 772
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8648974668275032e-05,
+      "loss": 0.808,
+      "step": 773
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8673100120627262e-05,
+      "loss": 0.7397,
+      "step": 774
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8697225572979492e-05,
+      "loss": 0.8662,
+      "step": 775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8721351025331726e-05,
+      "loss": 0.6395,
+      "step": 776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8745476477683956e-05,
+      "loss": 0.6234,
+      "step": 777
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.876960193003619e-05,
+      "loss": 0.7767,
+      "step": 778
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.879372738238842e-05,
+      "loss": 0.837,
+      "step": 779
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8817852834740653e-05,
+      "loss": 0.7994,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8841978287092883e-05,
+      "loss": 0.7644,
+      "step": 781
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8866103739445116e-05,
+      "loss": 0.7595,
+      "step": 782
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8890229191797347e-05,
+      "loss": 0.7562,
+      "step": 783
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.891435464414958e-05,
+      "loss": 0.851,
+      "step": 784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.893848009650181e-05,
+      "loss": 0.8342,
+      "step": 785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8962605548854044e-05,
+      "loss": 0.6281,
+      "step": 786
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8986731001206274e-05,
+      "loss": 0.8391,
+      "step": 787
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9010856453558507e-05,
+      "loss": 0.7789,
+      "step": 788
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9034981905910737e-05,
+      "loss": 0.6853,
+      "step": 789
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.905910735826297e-05,
+      "loss": 0.8407,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.90832328106152e-05,
+      "loss": 0.7909,
+      "step": 791
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9107358262967434e-05,
+      "loss": 0.7999,
+      "step": 792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9131483715319664e-05,
+      "loss": 0.7604,
+      "step": 793
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9155609167671894e-05,
+      "loss": 0.8238,
+      "step": 794
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9179734620024128e-05,
+      "loss": 0.7756,
+      "step": 795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9203860072376358e-05,
+      "loss": 0.7762,
+      "step": 796
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9227985524728588e-05,
+      "loss": 0.7299,
+      "step": 797
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.925211097708082e-05,
+      "loss": 0.827,
+      "step": 798
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.927623642943305e-05,
+      "loss": 0.7537,
+      "step": 799
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9300361881785285e-05,
+      "loss": 0.7653,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9324487334137515e-05,
+      "loss": 0.6793,
+      "step": 801
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.934861278648975e-05,
+      "loss": 0.7528,
+      "step": 802
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.937273823884198e-05,
+      "loss": 0.7525,
+      "step": 803
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9396863691194212e-05,
+      "loss": 0.7585,
+      "step": 804
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9420989143546442e-05,
+      "loss": 0.6332,
+      "step": 805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9445114595898676e-05,
+      "loss": 0.7824,
+      "step": 806
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9469240048250906e-05,
+      "loss": 0.6006,
+      "step": 807
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.949336550060314e-05,
+      "loss": 0.7951,
+      "step": 808
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.951749095295537e-05,
+      "loss": 0.8321,
+      "step": 809
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9541616405307603e-05,
+      "loss": 0.6634,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9565741857659833e-05,
+      "loss": 0.7844,
+      "step": 811
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9589867310012066e-05,
+      "loss": 0.7541,
+      "step": 812
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9613992762364297e-05,
+      "loss": 0.8062,
+      "step": 813
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.963811821471653e-05,
+      "loss": 0.7247,
+      "step": 814
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.966224366706876e-05,
+      "loss": 0.7441,
+      "step": 815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.968636911942099e-05,
+      "loss": 0.7803,
+      "step": 816
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9710494571773224e-05,
+      "loss": 0.8055,
+      "step": 817
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9734620024125454e-05,
+      "loss": 0.8232,
+      "step": 818
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9758745476477684e-05,
+      "loss": 0.7238,
+      "step": 819
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9782870928829917e-05,
+      "loss": 0.7912,
+      "step": 820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9806996381182147e-05,
+      "loss": 0.7782,
+      "step": 821
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983112183353438e-05,
+      "loss": 0.6476,
+      "step": 822
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.985524728588661e-05,
+      "loss": 0.6774,
+      "step": 823
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9879372738238844e-05,
+      "loss": 0.7964,
+      "step": 824
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9903498190591075e-05,
+      "loss": 0.834,
+      "step": 825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9927623642943308e-05,
+      "loss": 0.7766,
+      "step": 826
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9951749095295538e-05,
+      "loss": 0.8128,
+      "step": 827
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.997587454764777e-05,
+      "loss": 0.8034,
+      "step": 828
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2e-05,
+      "loss": 0.7779,
+      "step": 829
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999931195607e-05,
+      "loss": 0.6413,
+      "step": 830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999972478243e-05,
+      "loss": 0.8694,
+      "step": 831
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999380760465e-05,
+      "loss": 0.8397,
+      "step": 832
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999998899129725e-05,
+      "loss": 0.8297,
+      "step": 833
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999998279890212e-05,
+      "loss": 0.747,
+      "step": 834
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999752304194e-05,
+      "loss": 0.676,
+      "step": 835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999996628584913e-05,
+      "loss": 0.6417,
+      "step": 836
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999995596519143e-05,
+      "loss": 0.7816,
+      "step": 837
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999994426844648e-05,
+      "loss": 0.8547,
+      "step": 838
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999993119561444e-05,
+      "loss": 0.7276,
+      "step": 839
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999167466955e-05,
+      "loss": 0.7314,
+      "step": 840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999009216898e-05,
+      "loss": 0.7886,
+      "step": 841
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999998837205976e-05,
+      "loss": 0.8124,
+      "step": 842
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999986514341912e-05,
+      "loss": 0.6001,
+      "step": 843
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999998451901547e-05,
+      "loss": 0.7748,
+      "step": 844
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999982386080445e-05,
+      "loss": 0.7555,
+      "step": 845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999998011553688e-05,
+      "loss": 0.7248,
+      "step": 846
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999977707384802e-05,
+      "loss": 0.6666,
+      "step": 847
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999975161624244e-05,
+      "loss": 0.7587,
+      "step": 848
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999997247825524e-05,
+      "loss": 0.7796,
+      "step": 849
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999969657277828e-05,
+      "loss": 0.8166,
+      "step": 850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996669869205e-05,
+      "loss": 0.7607,
+      "step": 851
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996360249794e-05,
+      "loss": 0.8257,
+      "step": 852
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999960368695545e-05,
+      "loss": 0.8839,
+      "step": 853
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999956997284908e-05,
+      "loss": 0.8052,
+      "step": 854
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999953488266074e-05,
+      "loss": 0.8581,
+      "step": 855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999949841639098e-05,
+      "loss": 0.6654,
+      "step": 856
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999946057404023e-05,
+      "loss": 0.7852,
+      "step": 857
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999942135560904e-05,
+      "loss": 0.6232,
+      "step": 858
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999938076109795e-05,
+      "loss": 0.9059,
+      "step": 859
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999933879050747e-05,
+      "loss": 0.8476,
+      "step": 860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999992954438383e-05,
+      "loss": 0.7742,
+      "step": 861
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999992507210909e-05,
+      "loss": 0.8109,
+      "step": 862
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999920462226594e-05,
+      "loss": 0.73,
+      "step": 863
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999991571473641e-05,
+      "loss": 0.6969,
+      "step": 864
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999910829638596e-05,
+      "loss": 0.7742,
+      "step": 865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999905806933223e-05,
+      "loss": 0.7732,
+      "step": 866
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999990064662036e-05,
+      "loss": 0.8514,
+      "step": 867
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999895348700077e-05,
+      "loss": 0.7188,
+      "step": 868
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999889913172446e-05,
+      "loss": 0.8399,
+      "step": 869
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999884340037542e-05,
+      "loss": 0.7975,
+      "step": 870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999878629295445e-05,
+      "loss": 0.7662,
+      "step": 871
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999872780946235e-05,
+      "loss": 0.7866,
+      "step": 872
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999866794989984e-05,
+      "loss": 0.8634,
+      "step": 873
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999986067142678e-05,
+      "loss": 0.8114,
+      "step": 874
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999854410256707e-05,
+      "loss": 0.8741,
+      "step": 875
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999984801147985e-05,
+      "loss": 0.7929,
+      "step": 876
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99998414750963e-05,
+      "loss": 0.7797,
+      "step": 877
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999983480110614e-05,
+      "loss": 0.7679,
+      "step": 878
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999982798950947e-05,
+      "loss": 0.8157,
+      "step": 879
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999821040306383e-05,
+      "loss": 0.6207,
+      "step": 880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999813953496966e-05,
+      "loss": 0.8079,
+      "step": 881
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999806729081327e-05,
+      "loss": 0.7763,
+      "step": 882
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999799367059558e-05,
+      "loss": 0.7474,
+      "step": 883
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999791867431764e-05,
+      "loss": 0.7757,
+      "step": 884
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999784230198047e-05,
+      "loss": 0.7344,
+      "step": 885
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999977645535851e-05,
+      "loss": 0.7976,
+      "step": 886
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999976854291327e-05,
+      "loss": 0.8006,
+      "step": 887
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999760492862422e-05,
+      "loss": 0.7646,
+      "step": 888
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999752305206083e-05,
+      "loss": 0.7861,
+      "step": 889
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999974397994437e-05,
+      "loss": 0.7972,
+      "step": 890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999973551707739e-05,
+      "loss": 0.7616,
+      "step": 891
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999972691660526e-05,
+      "loss": 0.7671,
+      "step": 892
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999718178528108e-05,
+      "loss": 0.672,
+      "step": 893
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999970930284604e-05,
+      "loss": 0.7981,
+      "step": 894
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999700289559193e-05,
+      "loss": 0.7586,
+      "step": 895
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999969113866768e-05,
+      "loss": 0.7999,
+      "step": 896
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999681850171627e-05,
+      "loss": 0.7966,
+      "step": 897
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999967242407117e-05,
+      "loss": 0.8252,
+      "step": 898
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999662860366432e-05,
+      "loss": 0.759,
+      "step": 899
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999653159057547e-05,
+      "loss": 0.7689,
+      "step": 900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999964332014465e-05,
+      "loss": 0.8067,
+      "step": 901
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999963334362787e-05,
+      "loss": 0.7323,
+      "step": 902
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999623229507352e-05,
+      "loss": 0.774,
+      "step": 903
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999961297778323e-05,
+      "loss": 0.7492,
+      "step": 904
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999602588455648e-05,
+      "loss": 0.7579,
+      "step": 905
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999959206152475e-05,
+      "loss": 0.7129,
+      "step": 906
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999581396990673e-05,
+      "loss": 0.8119,
+      "step": 907
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999570594853575e-05,
+      "loss": 0.8502,
+      "step": 908
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99995596551136e-05,
+      "loss": 0.7579,
+      "step": 909
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999548577770892e-05,
+      "loss": 0.8516,
+      "step": 910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999537362825613e-05,
+      "loss": 0.7889,
+      "step": 911
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999526010277916e-05,
+      "loss": 0.8302,
+      "step": 912
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999514520127953e-05,
+      "loss": 0.824,
+      "step": 913
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999950289237588e-05,
+      "loss": 0.7609,
+      "step": 914
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999491127021863e-05,
+      "loss": 0.7657,
+      "step": 915
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999479224066066e-05,
+      "loss": 0.7422,
+      "step": 916
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999946718350864e-05,
+      "loss": 0.653,
+      "step": 917
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999455005349766e-05,
+      "loss": 0.7959,
+      "step": 918
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999442689589602e-05,
+      "loss": 0.8258,
+      "step": 919
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999430236228323e-05,
+      "loss": 0.797,
+      "step": 920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999417645266098e-05,
+      "loss": 0.7737,
+      "step": 921
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999404916703096e-05,
+      "loss": 0.6541,
+      "step": 922
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99993920505395e-05,
+      "loss": 0.7902,
+      "step": 923
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999937904677548e-05,
+      "loss": 0.7982,
+      "step": 924
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999365905411218e-05,
+      "loss": 0.672,
+      "step": 925
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999352626446898e-05,
+      "loss": 0.7996,
+      "step": 926
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99993392098827e-05,
+      "loss": 0.7495,
+      "step": 927
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999325655718805e-05,
+      "loss": 0.7721,
+      "step": 928
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99993119639554e-05,
+      "loss": 0.814,
+      "step": 929
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999298134592683e-05,
+      "loss": 0.6382,
+      "step": 930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999284167630835e-05,
+      "loss": 0.8082,
+      "step": 931
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999270063070048e-05,
+      "loss": 0.8236,
+      "step": 932
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999255820910523e-05,
+      "loss": 0.7702,
+      "step": 933
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999924144115245e-05,
+      "loss": 0.7585,
+      "step": 934
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999226923796027e-05,
+      "loss": 0.78,
+      "step": 935
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999921226884146e-05,
+      "loss": 0.7641,
+      "step": 936
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999197476288942e-05,
+      "loss": 0.8592,
+      "step": 937
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999182546138683e-05,
+      "loss": 0.7348,
+      "step": 938
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999167478390883e-05,
+      "loss": 0.7811,
+      "step": 939
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999152273045755e-05,
+      "loss": 0.7541,
+      "step": 940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999136930103506e-05,
+      "loss": 0.7953,
+      "step": 941
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999121449564347e-05,
+      "loss": 0.7739,
+      "step": 942
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999910583142849e-05,
+      "loss": 0.6686,
+      "step": 943
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999090075696152e-05,
+      "loss": 0.6397,
+      "step": 944
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999074182367547e-05,
+      "loss": 0.7685,
+      "step": 945
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999058151442892e-05,
+      "loss": 0.6206,
+      "step": 946
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999041982922418e-05,
+      "loss": 0.734,
+      "step": 947
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999025676806333e-05,
+      "loss": 0.7988,
+      "step": 948
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999009233094876e-05,
+      "loss": 0.5993,
+      "step": 949
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998992651788262e-05,
+      "loss": 0.7838,
+      "step": 950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998975932886724e-05,
+      "loss": 0.7996,
+      "step": 951
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999895907639049e-05,
+      "loss": 0.743,
+      "step": 952
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998942082299795e-05,
+      "loss": 0.7725,
+      "step": 953
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999892495061487e-05,
+      "loss": 0.6749,
+      "step": 954
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998907681335954e-05,
+      "loss": 0.8258,
+      "step": 955
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999889027446328e-05,
+      "loss": 0.7765,
+      "step": 956
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999887272999709e-05,
+      "loss": 0.8248,
+      "step": 957
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998855047937626e-05,
+      "loss": 0.8275,
+      "step": 958
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998837228285134e-05,
+      "loss": 0.8317,
+      "step": 959
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999881927103985e-05,
+      "loss": 0.7938,
+      "step": 960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999880117620203e-05,
+      "loss": 0.6206,
+      "step": 961
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998782943771923e-05,
+      "loss": 0.6706,
+      "step": 962
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999876457374977e-05,
+      "loss": 0.8086,
+      "step": 963
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998746066135836e-05,
+      "loss": 0.6463,
+      "step": 964
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998727420930373e-05,
+      "loss": 0.7597,
+      "step": 965
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999870863813363e-05,
+      "loss": 0.8512,
+      "step": 966
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998689717745872e-05,
+      "loss": 0.7899,
+      "step": 967
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999867065976736e-05,
+      "loss": 0.6922,
+      "step": 968
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999865146419835e-05,
+      "loss": 0.6071,
+      "step": 969
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999863213103911e-05,
+      "loss": 0.8364,
+      "step": 970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998612660289912e-05,
+      "loss": 0.8159,
+      "step": 971
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998593051951014e-05,
+      "loss": 0.7885,
+      "step": 972
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998573306022692e-05,
+      "loss": 0.7623,
+      "step": 973
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998553422505215e-05,
+      "loss": 0.7611,
+      "step": 974
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998533401398856e-05,
+      "loss": 0.7406,
+      "step": 975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998513242703895e-05,
+      "loss": 0.6337,
+      "step": 976
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99984929464206e-05,
+      "loss": 0.7698,
+      "step": 977
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998472512549264e-05,
+      "loss": 0.801,
+      "step": 978
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998451941090157e-05,
+      "loss": 0.7292,
+      "step": 979
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998431232043568e-05,
+      "loss": 0.8032,
+      "step": 980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998410385409778e-05,
+      "loss": 0.6321,
+      "step": 981
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998389401189076e-05,
+      "loss": 0.6245,
+      "step": 982
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999836827938175e-05,
+      "loss": 0.7589,
+      "step": 983
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998347019988094e-05,
+      "loss": 0.6226,
+      "step": 984
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998325623008396e-05,
+      "loss": 0.8266,
+      "step": 985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999830408844295e-05,
+      "loss": 0.7452,
+      "step": 986
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998282416292057e-05,
+      "loss": 0.648,
+      "step": 987
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998260606556013e-05,
+      "loss": 0.7073,
+      "step": 988
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998238659235113e-05,
+      "loss": 0.8034,
+      "step": 989
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998216574329668e-05,
+      "loss": 0.7999,
+      "step": 990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998194351839978e-05,
+      "loss": 0.8197,
+      "step": 991
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999817199176635e-05,
+      "loss": 0.7861,
+      "step": 992
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998149494109085e-05,
+      "loss": 0.6601,
+      "step": 993
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99981268588685e-05,
+      "loss": 0.8125,
+      "step": 994
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999810408604491e-05,
+      "loss": 0.7137,
+      "step": 995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998081175638616e-05,
+      "loss": 0.7709,
+      "step": 996
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999805812764994e-05,
+      "loss": 0.8099,
+      "step": 997
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998034942079202e-05,
+      "loss": 0.7178,
+      "step": 998
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998011618926714e-05,
+      "loss": 0.6369,
+      "step": 999
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997988158192806e-05,
+      "loss": 0.704,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997964559877796e-05,
+      "loss": 0.6716,
+      "step": 1001
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997940823982006e-05,
+      "loss": 0.7978,
+      "step": 1002
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997916950505765e-05,
+      "loss": 0.7693,
+      "step": 1003
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997892939449404e-05,
+      "loss": 0.6759,
+      "step": 1004
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999786879081325e-05,
+      "loss": 0.7566,
+      "step": 1005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997844504597636e-05,
+      "loss": 0.7715,
+      "step": 1006
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997820080802896e-05,
+      "loss": 0.6167,
+      "step": 1007
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997795519429372e-05,
+      "loss": 0.6721,
+      "step": 1008
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999777082047739e-05,
+      "loss": 0.8152,
+      "step": 1009
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99977459839473e-05,
+      "loss": 0.8003,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999772100983944e-05,
+      "loss": 0.6399,
+      "step": 1011
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997695898154153e-05,
+      "loss": 0.6548,
+      "step": 1012
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999767064889179e-05,
+      "loss": 0.5842,
+      "step": 1013
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999764526205269e-05,
+      "loss": 0.7625,
+      "step": 1014
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997619737637207e-05,
+      "loss": 0.8363,
+      "step": 1015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997594075645694e-05,
+      "loss": 0.6406,
+      "step": 1016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99975682760785e-05,
+      "loss": 0.7794,
+      "step": 1017
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997542338935983e-05,
+      "loss": 0.7487,
+      "step": 1018
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997516264218494e-05,
+      "loss": 0.6482,
+      "step": 1019
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99974900519264e-05,
+      "loss": 0.8035,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999746370206006e-05,
+      "loss": 0.7466,
+      "step": 1021
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997437214619837e-05,
+      "loss": 0.7084,
+      "step": 1022
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999741058960609e-05,
+      "loss": 0.7919,
+      "step": 1023
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997383827019188e-05,
+      "loss": 0.6547,
+      "step": 1024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997356926859503e-05,
+      "loss": 0.8369,
+      "step": 1025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997329889127397e-05,
+      "loss": 0.8446,
+      "step": 1026
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999730271382325e-05,
+      "loss": 0.7525,
+      "step": 1027
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997275400947437e-05,
+      "loss": 0.6065,
+      "step": 1028
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997247950500327e-05,
+      "loss": 0.6467,
+      "step": 1029
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997220362482302e-05,
+      "loss": 0.7574,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999719263689374e-05,
+      "loss": 0.8096,
+      "step": 1031
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999716477373502e-05,
+      "loss": 0.6821,
+      "step": 1032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999713677300653e-05,
+      "loss": 0.6072,
+      "step": 1033
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997108634708656e-05,
+      "loss": 0.7866,
+      "step": 1034
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997080358841782e-05,
+      "loss": 0.6262,
+      "step": 1035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997051945406295e-05,
+      "loss": 0.8695,
+      "step": 1036
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997023394402594e-05,
+      "loss": 0.7314,
+      "step": 1037
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996994705831063e-05,
+      "loss": 0.7955,
+      "step": 1038
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99969658796921e-05,
+      "loss": 0.7867,
+      "step": 1039
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996936915986108e-05,
+      "loss": 0.6609,
+      "step": 1040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996907814713474e-05,
+      "loss": 0.7916,
+      "step": 1041
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999687857587461e-05,
+      "loss": 0.724,
+      "step": 1042
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999684919946991e-05,
+      "loss": 0.634,
+      "step": 1043
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999681968549978e-05,
+      "loss": 0.6231,
+      "step": 1044
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999679003396463e-05,
+      "loss": 0.8457,
+      "step": 1045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999676024486486e-05,
+      "loss": 0.7684,
+      "step": 1046
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999673031820089e-05,
+      "loss": 0.7921,
+      "step": 1047
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996700253973123e-05,
+      "loss": 0.8521,
+      "step": 1048
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999667005218198e-05,
+      "loss": 0.7651,
+      "step": 1049
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996639712827872e-05,
+      "loss": 0.827,
+      "step": 1050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996609235911218e-05,
+      "loss": 0.8237,
+      "step": 1051
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996578621432438e-05,
+      "loss": 0.7958,
+      "step": 1052
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999654786939195e-05,
+      "loss": 0.6348,
+      "step": 1053
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999651697979018e-05,
+      "loss": 0.7172,
+      "step": 1054
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996485952627554e-05,
+      "loss": 0.8218,
+      "step": 1055
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99964547879045e-05,
+      "loss": 0.7894,
+      "step": 1056
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996423485621438e-05,
+      "loss": 0.6367,
+      "step": 1057
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999639204577881e-05,
+      "loss": 0.7489,
+      "step": 1058
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996360468377044e-05,
+      "loss": 0.7509,
+      "step": 1059
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996328753416572e-05,
+      "loss": 0.7614,
+      "step": 1060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996296900897835e-05,
+      "loss": 0.6169,
+      "step": 1061
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996264910821266e-05,
+      "loss": 0.7885,
+      "step": 1062
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999623278318731e-05,
+      "loss": 0.7997,
+      "step": 1063
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999620051799641e-05,
+      "loss": 0.6815,
+      "step": 1064
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996168115249003e-05,
+      "loss": 0.776,
+      "step": 1065
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996135574945543e-05,
+      "loss": 0.6778,
+      "step": 1066
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999610289708647e-05,
+      "loss": 0.7861,
+      "step": 1067
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999607008167224e-05,
+      "loss": 0.7843,
+      "step": 1068
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996037128703306e-05,
+      "loss": 0.7923,
+      "step": 1069
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996004038180112e-05,
+      "loss": 0.7618,
+      "step": 1070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995970810103125e-05,
+      "loss": 0.7508,
+      "step": 1071
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999593744447279e-05,
+      "loss": 0.6715,
+      "step": 1072
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995903941289575e-05,
+      "loss": 0.8027,
+      "step": 1073
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999587030055394e-05,
+      "loss": 0.7568,
+      "step": 1074
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995836522266347e-05,
+      "loss": 0.7698,
+      "step": 1075
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999580260642726e-05,
+      "loss": 0.5885,
+      "step": 1076
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995768553037144e-05,
+      "loss": 0.7743,
+      "step": 1077
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999573436209647e-05,
+      "loss": 0.6643,
+      "step": 1078
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999570003360571e-05,
+      "loss": 0.7658,
+      "step": 1079
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995665567565336e-05,
+      "loss": 0.6226,
+      "step": 1080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995630963975818e-05,
+      "loss": 0.7813,
+      "step": 1081
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995596222837636e-05,
+      "loss": 0.8465,
+      "step": 1082
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995561344151264e-05,
+      "loss": 0.6459,
+      "step": 1083
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995526327917187e-05,
+      "loss": 0.7357,
+      "step": 1084
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995491174135887e-05,
+      "loss": 0.7683,
+      "step": 1085
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995455882807847e-05,
+      "loss": 0.7829,
+      "step": 1086
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995420453933545e-05,
+      "loss": 0.7557,
+      "step": 1087
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999538488751348e-05,
+      "loss": 0.7164,
+      "step": 1088
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995349183548135e-05,
+      "loss": 0.622,
+      "step": 1089
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995313342038e-05,
+      "loss": 0.6817,
+      "step": 1090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995277362983574e-05,
+      "loss": 0.6497,
+      "step": 1091
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995241246385347e-05,
+      "loss": 0.796,
+      "step": 1092
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999520499224382e-05,
+      "loss": 0.6198,
+      "step": 1093
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995168600559486e-05,
+      "loss": 0.813,
+      "step": 1094
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999513207133285e-05,
+      "loss": 0.7377,
+      "step": 1095
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995095404564417e-05,
+      "loss": 0.5862,
+      "step": 1096
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999505860025469e-05,
+      "loss": 0.7626,
+      "step": 1097
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995021658404173e-05,
+      "loss": 0.7713,
+      "step": 1098
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994984579013373e-05,
+      "loss": 0.8031,
+      "step": 1099
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994947362082806e-05,
+      "loss": 0.69,
+      "step": 1100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994910007612978e-05,
+      "loss": 0.7629,
+      "step": 1101
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999487251560441e-05,
+      "loss": 0.8422,
+      "step": 1102
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999483488605761e-05,
+      "loss": 0.7526,
+      "step": 1103
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994797118973098e-05,
+      "loss": 0.8118,
+      "step": 1104
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99947592143514e-05,
+      "loss": 0.7538,
+      "step": 1105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999472117219303e-05,
+      "loss": 0.8136,
+      "step": 1106
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994682992498517e-05,
+      "loss": 0.6964,
+      "step": 1107
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994644675268384e-05,
+      "loss": 0.7413,
+      "step": 1108
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994606220503157e-05,
+      "loss": 0.7721,
+      "step": 1109
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994567628203367e-05,
+      "loss": 0.7245,
+      "step": 1110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994528898369544e-05,
+      "loss": 0.6758,
+      "step": 1111
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999449003100222e-05,
+      "loss": 0.6422,
+      "step": 1112
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994451026101933e-05,
+      "loss": 0.8057,
+      "step": 1113
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999441188366922e-05,
+      "loss": 0.6279,
+      "step": 1114
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994372603704613e-05,
+      "loss": 0.645,
+      "step": 1115
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999433318620866e-05,
+      "loss": 0.7416,
+      "step": 1116
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99942936311819e-05,
+      "loss": 0.7356,
+      "step": 1117
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994253938624877e-05,
+      "loss": 0.7774,
+      "step": 1118
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999421410853814e-05,
+      "loss": 0.7702,
+      "step": 1119
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994174140922235e-05,
+      "loss": 0.8114,
+      "step": 1120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999413403577771e-05,
+      "loss": 0.6221,
+      "step": 1121
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999409379310512e-05,
+      "loss": 0.6477,
+      "step": 1122
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994053412905018e-05,
+      "loss": 0.7742,
+      "step": 1123
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999401289517796e-05,
+      "loss": 0.8012,
+      "step": 1124
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99939722399245e-05,
+      "loss": 0.7202,
+      "step": 1125
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993931447145206e-05,
+      "loss": 0.744,
+      "step": 1126
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993890516840634e-05,
+      "loss": 0.7913,
+      "step": 1127
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999384944901134e-05,
+      "loss": 0.6476,
+      "step": 1128
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99938082436579e-05,
+      "loss": 0.8132,
+      "step": 1129
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999376690078088e-05,
+      "loss": 0.6265,
+      "step": 1130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993725420380842e-05,
+      "loss": 0.7391,
+      "step": 1131
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993683802458365e-05,
+      "loss": 0.7175,
+      "step": 1132
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993642047014014e-05,
+      "loss": 0.8028,
+      "step": 1133
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999360015404837e-05,
+      "loss": 0.6577,
+      "step": 1134
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993558123562005e-05,
+      "loss": 0.7605,
+      "step": 1135
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993515955555498e-05,
+      "loss": 0.7508,
+      "step": 1136
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993473650029432e-05,
+      "loss": 0.8205,
+      "step": 1137
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999343120698439e-05,
+      "loss": 0.757,
+      "step": 1138
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993388626420947e-05,
+      "loss": 0.7539,
+      "step": 1139
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99933459083397e-05,
+      "loss": 0.7195,
+      "step": 1140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993303052741234e-05,
+      "loss": 0.7243,
+      "step": 1141
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993260059626133e-05,
+      "loss": 0.6188,
+      "step": 1142
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993216928994997e-05,
+      "loss": 0.8468,
+      "step": 1143
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993173660848414e-05,
+      "loss": 0.6251,
+      "step": 1144
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999313025518698e-05,
+      "loss": 0.7478,
+      "step": 1145
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993086712011293e-05,
+      "loss": 0.8766,
+      "step": 1146
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999304303132195e-05,
+      "loss": 0.6555,
+      "step": 1147
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992999213119555e-05,
+      "loss": 0.8351,
+      "step": 1148
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999295525740471e-05,
+      "loss": 0.8125,
+      "step": 1149
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999291116417802e-05,
+      "loss": 0.6272,
+      "step": 1150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992866933440096e-05,
+      "loss": 0.7867,
+      "step": 1151
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999282256519154e-05,
+      "loss": 0.7967,
+      "step": 1152
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992778059432964e-05,
+      "loss": 0.745,
+      "step": 1153
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992733416164983e-05,
+      "loss": 0.8013,
+      "step": 1154
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999268863538821e-05,
+      "loss": 0.7995,
+      "step": 1155
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999264371710326e-05,
+      "loss": 0.7439,
+      "step": 1156
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999259866131075e-05,
+      "loss": 0.5782,
+      "step": 1157
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992553468011304e-05,
+      "loss": 0.8012,
+      "step": 1158
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999250813720554e-05,
+      "loss": 0.7927,
+      "step": 1159
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992462668894087e-05,
+      "loss": 0.818,
+      "step": 1160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999241706307757e-05,
+      "loss": 0.7461,
+      "step": 1161
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999237131975661e-05,
+      "loss": 0.6671,
+      "step": 1162
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999232543893184e-05,
+      "loss": 0.6179,
+      "step": 1163
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992279420603894e-05,
+      "loss": 0.7985,
+      "step": 1164
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99922332647734e-05,
+      "loss": 0.6681,
+      "step": 1165
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992186971441003e-05,
+      "loss": 0.6657,
+      "step": 1166
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992140540607327e-05,
+      "loss": 0.7699,
+      "step": 1167
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999209397227302e-05,
+      "loss": 0.7873,
+      "step": 1168
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992047266438723e-05,
+      "loss": 0.7492,
+      "step": 1169
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992000423105067e-05,
+      "loss": 0.7481,
+      "step": 1170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991953442272713e-05,
+      "loss": 0.8399,
+      "step": 1171
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99919063239423e-05,
+      "loss": 0.7496,
+      "step": 1172
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999185906811447e-05,
+      "loss": 0.6022,
+      "step": 1173
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991811674789886e-05,
+      "loss": 0.6703,
+      "step": 1174
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999176414396919e-05,
+      "loss": 0.7665,
+      "step": 1175
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991716475653043e-05,
+      "loss": 0.7664,
+      "step": 1176
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991668669842097e-05,
+      "loss": 0.7049,
+      "step": 1177
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999162072653701e-05,
+      "loss": 0.6081,
+      "step": 1178
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999157264573844e-05,
+      "loss": 0.7396,
+      "step": 1179
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991524427447054e-05,
+      "loss": 0.5934,
+      "step": 1180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991476071663515e-05,
+      "loss": 0.5918,
+      "step": 1181
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999142757838848e-05,
+      "loss": 0.6709,
+      "step": 1182
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991378947622622e-05,
+      "loss": 0.6991,
+      "step": 1183
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991330179366618e-05,
+      "loss": 0.6445,
+      "step": 1184
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991281273621126e-05,
+      "loss": 0.7264,
+      "step": 1185
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991232230386826e-05,
+      "loss": 0.766,
+      "step": 1186
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999118304966439e-05,
+      "loss": 0.7534,
+      "step": 1187
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991133731454498e-05,
+      "loss": 0.8139,
+      "step": 1188
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991084275757826e-05,
+      "loss": 0.7371,
+      "step": 1189
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991034682575058e-05,
+      "loss": 0.6177,
+      "step": 1190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999098495190687e-05,
+      "loss": 0.744,
+      "step": 1191
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999093508375395e-05,
+      "loss": 0.7605,
+      "step": 1192
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990885078116988e-05,
+      "loss": 0.7464,
+      "step": 1193
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990834934996666e-05,
+      "loss": 0.7766,
+      "step": 1194
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999078465439368e-05,
+      "loss": 0.7077,
+      "step": 1195
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990734236308712e-05,
+      "loss": 0.6371,
+      "step": 1196
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990683680742465e-05,
+      "loss": 0.7532,
+      "step": 1197
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990632987695635e-05,
+      "loss": 0.7687,
+      "step": 1198
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990582157168914e-05,
+      "loss": 0.7123,
+      "step": 1199
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990531189163005e-05,
+      "loss": 0.7807,
+      "step": 1200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990480083678608e-05,
+      "loss": 0.722,
+      "step": 1201
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999042884071643e-05,
+      "loss": 0.8138,
+      "step": 1202
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990377460277166e-05,
+      "loss": 0.7724,
+      "step": 1203
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990325942361534e-05,
+      "loss": 0.6522,
+      "step": 1204
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990274286970236e-05,
+      "loss": 0.635,
+      "step": 1205
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999022249410399e-05,
+      "loss": 0.7333,
+      "step": 1206
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99901705637635e-05,
+      "loss": 0.766,
+      "step": 1207
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999011849594949e-05,
+      "loss": 0.7548,
+      "step": 1208
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999006629066267e-05,
+      "loss": 0.6852,
+      "step": 1209
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990013947903755e-05,
+      "loss": 0.7988,
+      "step": 1210
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989961467673473e-05,
+      "loss": 0.766,
+      "step": 1211
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989908849972543e-05,
+      "loss": 0.7396,
+      "step": 1212
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998985609480169e-05,
+      "loss": 0.6861,
+      "step": 1213
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998980320216164e-05,
+      "loss": 0.8247,
+      "step": 1214
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998975017205312e-05,
+      "loss": 0.7241,
+      "step": 1215
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998969700447686e-05,
+      "loss": 0.7079,
+      "step": 1216
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989643699433587e-05,
+      "loss": 0.766,
+      "step": 1217
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989590256924045e-05,
+      "loss": 0.7472,
+      "step": 1218
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998953667694896e-05,
+      "loss": 0.7357,
+      "step": 1219
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989482959509073e-05,
+      "loss": 0.7767,
+      "step": 1220
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989429104605125e-05,
+      "loss": 0.7943,
+      "step": 1221
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989375112237853e-05,
+      "loss": 0.6242,
+      "step": 1222
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989320982408003e-05,
+      "loss": 0.7507,
+      "step": 1223
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989266715116316e-05,
+      "loss": 0.6459,
+      "step": 1224
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998921231036355e-05,
+      "loss": 0.7346,
+      "step": 1225
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989157768150437e-05,
+      "loss": 0.6611,
+      "step": 1226
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989103088477738e-05,
+      "loss": 0.7798,
+      "step": 1227
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989048271346204e-05,
+      "loss": 0.8077,
+      "step": 1228
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998899331675659e-05,
+      "loss": 0.7464,
+      "step": 1229
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988938224709647e-05,
+      "loss": 0.8024,
+      "step": 1230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998888299520614e-05,
+      "loss": 0.7363,
+      "step": 1231
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988827628246825e-05,
+      "loss": 0.7897,
+      "step": 1232
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988772123832464e-05,
+      "loss": 0.7624,
+      "step": 1233
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988716481963825e-05,
+      "loss": 0.7595,
+      "step": 1234
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988660702641668e-05,
+      "loss": 0.6119,
+      "step": 1235
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998860478586676e-05,
+      "loss": 0.626,
+      "step": 1236
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998854873163988e-05,
+      "loss": 0.6246,
+      "step": 1237
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988492539961788e-05,
+      "loss": 0.7875,
+      "step": 1238
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998843621083326e-05,
+      "loss": 0.7459,
+      "step": 1239
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998837974425508e-05,
+      "loss": 0.7877,
+      "step": 1240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988323140228015e-05,
+      "loss": 0.7734,
+      "step": 1241
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988266398752845e-05,
+      "loss": 0.6268,
+      "step": 1242
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988209519830355e-05,
+      "loss": 0.6956,
+      "step": 1243
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988152503461325e-05,
+      "loss": 0.7947,
+      "step": 1244
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988095349646543e-05,
+      "loss": 0.7972,
+      "step": 1245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988038058386787e-05,
+      "loss": 0.712,
+      "step": 1246
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987980629682854e-05,
+      "loss": 0.6774,
+      "step": 1247
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987923063535533e-05,
+      "loss": 0.7844,
+      "step": 1248
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987865359945614e-05,
+      "loss": 0.6403,
+      "step": 1249
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998780751891389e-05,
+      "loss": 0.7923,
+      "step": 1250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987749540441158e-05,
+      "loss": 0.7406,
+      "step": 1251
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987691424528214e-05,
+      "loss": 0.7151,
+      "step": 1252
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987633171175864e-05,
+      "loss": 0.8147,
+      "step": 1253
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987574780384906e-05,
+      "loss": 0.6513,
+      "step": 1254
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987516252156143e-05,
+      "loss": 0.7406,
+      "step": 1255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998745758649038e-05,
+      "loss": 0.7076,
+      "step": 1256
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987398783388422e-05,
+      "loss": 0.778,
+      "step": 1257
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987339842851086e-05,
+      "loss": 0.7378,
+      "step": 1258
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987280764879174e-05,
+      "loss": 0.7581,
+      "step": 1259
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987221549473507e-05,
+      "loss": 0.7646,
+      "step": 1260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987162196634894e-05,
+      "loss": 0.7419,
+      "step": 1261
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987102706364152e-05,
+      "loss": 0.669,
+      "step": 1262
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99870430786621e-05,
+      "loss": 0.7052,
+      "step": 1263
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986983313529564e-05,
+      "loss": 0.7623,
+      "step": 1264
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998692341096736e-05,
+      "loss": 0.759,
+      "step": 1265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986863370976316e-05,
+      "loss": 0.7623,
+      "step": 1266
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998680319355726e-05,
+      "loss": 0.7681,
+      "step": 1267
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998674287871101e-05,
+      "loss": 0.7899,
+      "step": 1268
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986682426438406e-05,
+      "loss": 0.8262,
+      "step": 1269
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986621836740277e-05,
+      "loss": 0.618,
+      "step": 1270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986561109617457e-05,
+      "loss": 0.7404,
+      "step": 1271
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986500245070778e-05,
+      "loss": 0.7224,
+      "step": 1272
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986439243101084e-05,
+      "loss": 0.7921,
+      "step": 1273
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986378103709212e-05,
+      "loss": 0.7593,
+      "step": 1274
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986316826896e-05,
+      "loss": 0.763,
+      "step": 1275
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986255412662294e-05,
+      "loss": 0.7885,
+      "step": 1276
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998619386100894e-05,
+      "loss": 0.8088,
+      "step": 1277
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998613217193678e-05,
+      "loss": 0.7473,
+      "step": 1278
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998607034544667e-05,
+      "loss": 0.6848,
+      "step": 1279
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998600838153946e-05,
+      "loss": 0.7033,
+      "step": 1280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985946280215996e-05,
+      "loss": 0.6619,
+      "step": 1281
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998588404147714e-05,
+      "loss": 0.6038,
+      "step": 1282
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998582166532374e-05,
+      "loss": 0.7626,
+      "step": 1283
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985759151756668e-05,
+      "loss": 0.6253,
+      "step": 1284
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998569650077677e-05,
+      "loss": 0.7496,
+      "step": 1285
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985633712384915e-05,
+      "loss": 0.832,
+      "step": 1286
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985570786581966e-05,
+      "loss": 0.6113,
+      "step": 1287
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985507723368786e-05,
+      "loss": 0.6059,
+      "step": 1288
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998544452274625e-05,
+      "loss": 0.7665,
+      "step": 1289
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998538118471522e-05,
+      "loss": 0.8271,
+      "step": 1290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985317709276574e-05,
+      "loss": 0.7646,
+      "step": 1291
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998525409643118e-05,
+      "loss": 0.7553,
+      "step": 1292
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985190346179918e-05,
+      "loss": 0.6251,
+      "step": 1293
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998512645852366e-05,
+      "loss": 0.6047,
+      "step": 1294
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985062433463292e-05,
+      "loss": 0.6002,
+      "step": 1295
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984998270999688e-05,
+      "loss": 0.6506,
+      "step": 1296
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984933971133736e-05,
+      "loss": 0.7772,
+      "step": 1297
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998486953386632e-05,
+      "loss": 0.7701,
+      "step": 1298
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984804959198327e-05,
+      "loss": 0.7316,
+      "step": 1299
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998474024713064e-05,
+      "loss": 0.7686,
+      "step": 1300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984675397664154e-05,
+      "loss": 0.7838,
+      "step": 1301
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984610410799764e-05,
+      "loss": 0.7772,
+      "step": 1302
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984545286538362e-05,
+      "loss": 0.7355,
+      "step": 1303
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984480024880844e-05,
+      "loss": 0.7647,
+      "step": 1304
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984414625828107e-05,
+      "loss": 0.8216,
+      "step": 1305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984349089381046e-05,
+      "loss": 0.8685,
+      "step": 1306
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984283415540575e-05,
+      "loss": 0.6829,
+      "step": 1307
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998421760430759e-05,
+      "loss": 0.762,
+      "step": 1308
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984151655682997e-05,
+      "loss": 0.7537,
+      "step": 1309
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984085569667705e-05,
+      "loss": 0.7761,
+      "step": 1310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984019346262623e-05,
+      "loss": 0.6958,
+      "step": 1311
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998395298546866e-05,
+      "loss": 0.6955,
+      "step": 1312
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983886487286733e-05,
+      "loss": 0.6008,
+      "step": 1313
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983819851717753e-05,
+      "loss": 0.7496,
+      "step": 1314
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998375307876264e-05,
+      "loss": 0.77,
+      "step": 1315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983686168422315e-05,
+      "loss": 0.6474,
+      "step": 1316
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983619120697694e-05,
+      "loss": 0.8052,
+      "step": 1317
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983551935589705e-05,
+      "loss": 0.7497,
+      "step": 1318
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983484613099265e-05,
+      "loss": 0.7622,
+      "step": 1319
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983417153227304e-05,
+      "loss": 0.7608,
+      "step": 1320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983349555974754e-05,
+      "loss": 0.7831,
+      "step": 1321
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998328182134254e-05,
+      "loss": 0.7518,
+      "step": 1322
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983213949331594e-05,
+      "loss": 0.8188,
+      "step": 1323
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983145939942858e-05,
+      "loss": 0.7934,
+      "step": 1324
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983077793177253e-05,
+      "loss": 0.7716,
+      "step": 1325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983009509035735e-05,
+      "loss": 0.8011,
+      "step": 1326
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998294108751923e-05,
+      "loss": 0.5992,
+      "step": 1327
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998287252862868e-05,
+      "loss": 0.7583,
+      "step": 1328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998280383236504e-05,
+      "loss": 0.7626,
+      "step": 1329
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982734998729244e-05,
+      "loss": 0.7943,
+      "step": 1330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998266602772224e-05,
+      "loss": 0.7403,
+      "step": 1331
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982596919344983e-05,
+      "loss": 0.6462,
+      "step": 1332
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998252767359842e-05,
+      "loss": 0.7563,
+      "step": 1333
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982458290483504e-05,
+      "loss": 0.7736,
+      "step": 1334
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982388770001193e-05,
+      "loss": 0.7819,
+      "step": 1335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998231911215244e-05,
+      "loss": 0.6819,
+      "step": 1336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982249316938205e-05,
+      "loss": 0.7483,
+      "step": 1337
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982179384359447e-05,
+      "loss": 0.8184,
+      "step": 1338
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998210931441713e-05,
+      "loss": 0.7855,
+      "step": 1339
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982039107112218e-05,
+      "loss": 0.7966,
+      "step": 1340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981968762445678e-05,
+      "loss": 0.6201,
+      "step": 1341
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981898280418477e-05,
+      "loss": 0.7591,
+      "step": 1342
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981827661031586e-05,
+      "loss": 0.797,
+      "step": 1343
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981756904285973e-05,
+      "loss": 0.8684,
+      "step": 1344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981686010182614e-05,
+      "loss": 0.716,
+      "step": 1345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981614978722485e-05,
+      "loss": 0.6695,
+      "step": 1346
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981543809906564e-05,
+      "loss": 0.8304,
+      "step": 1347
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981472503735832e-05,
+      "loss": 0.7913,
+      "step": 1348
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981401060211267e-05,
+      "loss": 0.8055,
+      "step": 1349
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981329479333848e-05,
+      "loss": 0.7305,
+      "step": 1350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998125776110457e-05,
+      "loss": 0.7033,
+      "step": 1351
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998118590552441e-05,
+      "loss": 0.7676,
+      "step": 1352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981113912594368e-05,
+      "loss": 0.7596,
+      "step": 1353
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981041782315425e-05,
+      "loss": 0.7407,
+      "step": 1354
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980969514688575e-05,
+      "loss": 0.8035,
+      "step": 1355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980897109714818e-05,
+      "loss": 0.7532,
+      "step": 1356
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980824567395146e-05,
+      "loss": 0.766,
+      "step": 1357
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980751887730556e-05,
+      "loss": 0.7465,
+      "step": 1358
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980679070722053e-05,
+      "loss": 0.7487,
+      "step": 1359
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980606116370634e-05,
+      "loss": 0.672,
+      "step": 1360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980533024677305e-05,
+      "loss": 0.8111,
+      "step": 1361
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980459795643076e-05,
+      "loss": 0.8647,
+      "step": 1362
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980386429268946e-05,
+      "loss": 0.7558,
+      "step": 1363
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980312925555932e-05,
+      "loss": 0.7736,
+      "step": 1364
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998023928450504e-05,
+      "loss": 0.7241,
+      "step": 1365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998016550611729e-05,
+      "loss": 0.6498,
+      "step": 1366
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998009159039369e-05,
+      "loss": 0.7122,
+      "step": 1367
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998001753733526e-05,
+      "loss": 0.5795,
+      "step": 1368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997994334694302e-05,
+      "loss": 0.6224,
+      "step": 1369
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979869019217992e-05,
+      "loss": 0.7782,
+      "step": 1370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979794554161197e-05,
+      "loss": 0.7613,
+      "step": 1371
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997971995177366e-05,
+      "loss": 0.7841,
+      "step": 1372
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979645212056407e-05,
+      "loss": 0.7539,
+      "step": 1373
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979570335010464e-05,
+      "loss": 0.7651,
+      "step": 1374
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997949532063687e-05,
+      "loss": 0.8104,
+      "step": 1375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997942016893665e-05,
+      "loss": 0.7133,
+      "step": 1376
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979344879910837e-05,
+      "loss": 0.747,
+      "step": 1377
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997926945356047e-05,
+      "loss": 0.7937,
+      "step": 1378
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979193889886588e-05,
+      "loss": 0.7342,
+      "step": 1379
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979118188890234e-05,
+      "loss": 0.7476,
+      "step": 1380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979042350572437e-05,
+      "loss": 0.7849,
+      "step": 1381
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978966374934255e-05,
+      "loss": 0.8449,
+      "step": 1382
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978890261976725e-05,
+      "loss": 0.7863,
+      "step": 1383
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978814011700897e-05,
+      "loss": 0.7999,
+      "step": 1384
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997873762410782e-05,
+      "loss": 0.7511,
+      "step": 1385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978661099198545e-05,
+      "loss": 0.6473,
+      "step": 1386
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978584436974128e-05,
+      "loss": 0.7754,
+      "step": 1387
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978507637435615e-05,
+      "loss": 0.6851,
+      "step": 1388
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978430700584075e-05,
+      "loss": 0.6065,
+      "step": 1389
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997835362642056e-05,
+      "loss": 0.7814,
+      "step": 1390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978276414946126e-05,
+      "loss": 0.7388,
+      "step": 1391
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978199066161847e-05,
+      "loss": 0.666,
+      "step": 1392
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978121580068778e-05,
+      "loss": 0.6464,
+      "step": 1393
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997804395666799e-05,
+      "loss": 0.6516,
+      "step": 1394
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997796619596055e-05,
+      "loss": 0.782,
+      "step": 1395
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977888297947527e-05,
+      "loss": 0.7759,
+      "step": 1396
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977810262629993e-05,
+      "loss": 0.7202,
+      "step": 1397
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977732090009022e-05,
+      "loss": 0.6246,
+      "step": 1398
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977653780085694e-05,
+      "loss": 0.7613,
+      "step": 1399
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997757533286108e-05,
+      "loss": 0.7237,
+      "step": 1400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977496748336264e-05,
+      "loss": 0.7791,
+      "step": 1401
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977418026512323e-05,
+      "loss": 0.8215,
+      "step": 1402
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977339167390345e-05,
+      "loss": 0.7858,
+      "step": 1403
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997726017097141e-05,
+      "loss": 0.7976,
+      "step": 1404
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977181037256613e-05,
+      "loss": 0.7046,
+      "step": 1405
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977101766247036e-05,
+      "loss": 0.7307,
+      "step": 1406
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997702235794377e-05,
+      "loss": 0.7113,
+      "step": 1407
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976942812347914e-05,
+      "loss": 0.6403,
+      "step": 1408
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976863129460557e-05,
+      "loss": 0.8109,
+      "step": 1409
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976783309282793e-05,
+      "loss": 0.7838,
+      "step": 1410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976703351815727e-05,
+      "loss": 0.6515,
+      "step": 1411
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997662325706046e-05,
+      "loss": 0.7554,
+      "step": 1412
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976543025018086e-05,
+      "loss": 0.7769,
+      "step": 1413
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976462655689716e-05,
+      "loss": 0.8035,
+      "step": 1414
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976382149076455e-05,
+      "loss": 0.7909,
+      "step": 1415
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976301505179404e-05,
+      "loss": 0.7674,
+      "step": 1416
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976220723999685e-05,
+      "loss": 0.7418,
+      "step": 1417
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99761398055384e-05,
+      "loss": 0.6187,
+      "step": 1418
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976058749796667e-05,
+      "loss": 0.645,
+      "step": 1419
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975977556775597e-05,
+      "loss": 0.7888,
+      "step": 1420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975896226476314e-05,
+      "loss": 0.6254,
+      "step": 1421
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997581475889993e-05,
+      "loss": 0.7426,
+      "step": 1422
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997573315404757e-05,
+      "loss": 0.7295,
+      "step": 1423
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975651411920357e-05,
+      "loss": 0.6064,
+      "step": 1424
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975569532519413e-05,
+      "loss": 0.7241,
+      "step": 1425
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997548751584587e-05,
+      "loss": 0.7557,
+      "step": 1426
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975405361900853e-05,
+      "loss": 0.7319,
+      "step": 1427
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975323070685494e-05,
+      "loss": 0.785,
+      "step": 1428
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997524064220092e-05,
+      "loss": 0.7486,
+      "step": 1429
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975158076448274e-05,
+      "loss": 0.7622,
+      "step": 1430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975075373428688e-05,
+      "loss": 0.6152,
+      "step": 1431
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99749925331433e-05,
+      "loss": 0.7065,
+      "step": 1432
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974909555593246e-05,
+      "loss": 0.8278,
+      "step": 1433
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974826440779674e-05,
+      "loss": 0.7519,
+      "step": 1434
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974743188703726e-05,
+      "loss": 0.6171,
+      "step": 1435
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974659799366546e-05,
+      "loss": 0.7331,
+      "step": 1436
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974576272769286e-05,
+      "loss": 0.6647,
+      "step": 1437
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974492608913088e-05,
+      "loss": 0.8227,
+      "step": 1438
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997440880779911e-05,
+      "loss": 0.779,
+      "step": 1439
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974324869428504e-05,
+      "loss": 0.7517,
+      "step": 1440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997424079380242e-05,
+      "loss": 0.7654,
+      "step": 1441
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974156580922018e-05,
+      "loss": 0.7004,
+      "step": 1442
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997407223078846e-05,
+      "loss": 0.7835,
+      "step": 1443
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973987743402905e-05,
+      "loss": 0.7014,
+      "step": 1444
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973903118766515e-05,
+      "loss": 0.6603,
+      "step": 1445
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997381835688045e-05,
+      "loss": 0.8193,
+      "step": 1446
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973733457745883e-05,
+      "loss": 0.8223,
+      "step": 1447
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997364842136398e-05,
+      "loss": 0.8037,
+      "step": 1448
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997356324773591e-05,
+      "loss": 0.7615,
+      "step": 1449
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997347793686285e-05,
+      "loss": 0.627,
+      "step": 1450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973392488745965e-05,
+      "loss": 0.7773,
+      "step": 1451
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997330690338644e-05,
+      "loss": 0.738,
+      "step": 1452
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973221180785445e-05,
+      "loss": 0.711,
+      "step": 1453
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973135320944164e-05,
+      "loss": 0.7348,
+      "step": 1454
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973049323863783e-05,
+      "loss": 0.763,
+      "step": 1455
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972963189545474e-05,
+      "loss": 0.86,
+      "step": 1456
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972876917990433e-05,
+      "loss": 0.792,
+      "step": 1457
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997279050919984e-05,
+      "loss": 0.7208,
+      "step": 1458
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972703963174887e-05,
+      "loss": 0.779,
+      "step": 1459
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972617279916765e-05,
+      "loss": 0.752,
+      "step": 1460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972530459426663e-05,
+      "loss": 0.799,
+      "step": 1461
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972443501705787e-05,
+      "loss": 0.7397,
+      "step": 1462
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997235640675532e-05,
+      "loss": 0.7801,
+      "step": 1463
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997226917457647e-05,
+      "loss": 0.7495,
+      "step": 1464
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997218180517043e-05,
+      "loss": 0.6454,
+      "step": 1465
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972094298538407e-05,
+      "loss": 0.6886,
+      "step": 1466
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972006654681605e-05,
+      "loss": 0.7107,
+      "step": 1467
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971918873601227e-05,
+      "loss": 0.8358,
+      "step": 1468
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997183095529849e-05,
+      "loss": 0.7829,
+      "step": 1469
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971742899774588e-05,
+      "loss": 0.6803,
+      "step": 1470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997165470703075e-05,
+      "loss": 0.7676,
+      "step": 1471
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971566377068176e-05,
+      "loss": 0.7652,
+      "step": 1472
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971477909888088e-05,
+      "loss": 0.6966,
+      "step": 1473
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971389305491706e-05,
+      "loss": 0.7471,
+      "step": 1474
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971300563880238e-05,
+      "loss": 0.6709,
+      "step": 1475
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971211685054918e-05,
+      "loss": 0.711,
+      "step": 1476
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971122669016962e-05,
+      "loss": 0.7057,
+      "step": 1477
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971033515767597e-05,
+      "loss": 0.7075,
+      "step": 1478
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997094422530805e-05,
+      "loss": 0.677,
+      "step": 1479
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997085479763955e-05,
+      "loss": 0.6705,
+      "step": 1480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970765232763325e-05,
+      "loss": 0.6031,
+      "step": 1481
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997067553068061e-05,
+      "loss": 0.8094,
+      "step": 1482
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997058569139264e-05,
+      "loss": 0.6341,
+      "step": 1483
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997049571490065e-05,
+      "loss": 0.7573,
+      "step": 1484
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970405601205874e-05,
+      "loss": 0.6971,
+      "step": 1485
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997031535030956e-05,
+      "loss": 0.7577,
+      "step": 1486
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970224962212944e-05,
+      "loss": 0.7548,
+      "step": 1487
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970134436917274e-05,
+      "loss": 0.6553,
+      "step": 1488
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970043774423793e-05,
+      "loss": 0.7541,
+      "step": 1489
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969952974733747e-05,
+      "loss": 0.7246,
+      "step": 1490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996986203784839e-05,
+      "loss": 0.6709,
+      "step": 1491
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996977096376897e-05,
+      "loss": 0.7666,
+      "step": 1492
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969679752496744e-05,
+      "loss": 0.767,
+      "step": 1493
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969588404032964e-05,
+      "loss": 0.8134,
+      "step": 1494
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969496918378883e-05,
+      "loss": 0.6544,
+      "step": 1495
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969405295535768e-05,
+      "loss": 0.7816,
+      "step": 1496
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969313535504876e-05,
+      "loss": 0.7861,
+      "step": 1497
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996922163828747e-05,
+      "loss": 0.7773,
+      "step": 1498
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969129603884812e-05,
+      "loss": 0.8325,
+      "step": 1499
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969037432298175e-05,
+      "loss": 0.7276,
+      "step": 1500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968945123528823e-05,
+      "loss": 0.732,
+      "step": 1501
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968852677578026e-05,
+      "loss": 0.8143,
+      "step": 1502
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968760094447055e-05,
+      "loss": 0.7219,
+      "step": 1503
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968667374137184e-05,
+      "loss": 0.755,
+      "step": 1504
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968574516649694e-05,
+      "loss": 0.7411,
+      "step": 1505
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968481521985858e-05,
+      "loss": 0.7828,
+      "step": 1506
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996838839014696e-05,
+      "loss": 0.6481,
+      "step": 1507
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996829512113428e-05,
+      "loss": 0.5973,
+      "step": 1508
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968201714949095e-05,
+      "loss": 0.6576,
+      "step": 1509
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968108171592698e-05,
+      "loss": 0.7489,
+      "step": 1510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968014491066373e-05,
+      "loss": 0.797,
+      "step": 1511
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996792067337141e-05,
+      "loss": 0.8197,
+      "step": 1512
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99678267185091e-05,
+      "loss": 0.7438,
+      "step": 1513
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967732626480736e-05,
+      "loss": 0.7846,
+      "step": 1514
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967638397287613e-05,
+      "loss": 0.7199,
+      "step": 1515
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967544030931026e-05,
+      "loss": 0.7082,
+      "step": 1516
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967449527412277e-05,
+      "loss": 0.683,
+      "step": 1517
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967354886732662e-05,
+      "loss": 0.7785,
+      "step": 1518
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996726010889349e-05,
+      "loss": 0.7923,
+      "step": 1519
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967165193896055e-05,
+      "loss": 0.6146,
+      "step": 1520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967070141741672e-05,
+      "loss": 0.7784,
+      "step": 1521
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966974952431645e-05,
+      "loss": 0.7638,
+      "step": 1522
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966879625967288e-05,
+      "loss": 0.7016,
+      "step": 1523
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966784162349907e-05,
+      "loss": 0.7195,
+      "step": 1524
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996668856158082e-05,
+      "loss": 0.6385,
+      "step": 1525
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996659282366134e-05,
+      "loss": 0.7828,
+      "step": 1526
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966496948592786e-05,
+      "loss": 0.7479,
+      "step": 1527
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966400936376474e-05,
+      "loss": 0.712,
+      "step": 1528
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996630478701373e-05,
+      "loss": 0.7384,
+      "step": 1529
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966208500505874e-05,
+      "loss": 0.606,
+      "step": 1530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966112076854235e-05,
+      "loss": 0.7251,
+      "step": 1531
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996601551606013e-05,
+      "loss": 0.6818,
+      "step": 1532
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99659188181249e-05,
+      "loss": 0.5624,
+      "step": 1533
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996582198304987e-05,
+      "loss": 0.7727,
+      "step": 1534
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965725010836373e-05,
+      "loss": 0.6193,
+      "step": 1535
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965627901485742e-05,
+      "loss": 0.7511,
+      "step": 1536
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965530654999315e-05,
+      "loss": 0.7478,
+      "step": 1537
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965433271378427e-05,
+      "loss": 0.7105,
+      "step": 1538
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965335750624425e-05,
+      "loss": 0.7549,
+      "step": 1539
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965238092738643e-05,
+      "loss": 0.7538,
+      "step": 1540
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996514029772243e-05,
+      "loss": 0.7811,
+      "step": 1541
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965042365577134e-05,
+      "loss": 0.7418,
+      "step": 1542
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964944296304095e-05,
+      "loss": 0.7176,
+      "step": 1543
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964846089904668e-05,
+      "loss": 0.7858,
+      "step": 1544
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964747746380202e-05,
+      "loss": 0.7976,
+      "step": 1545
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996464926573205e-05,
+      "loss": 0.7323,
+      "step": 1546
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964550647961572e-05,
+      "loss": 0.7859,
+      "step": 1547
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996445189307012e-05,
+      "loss": 0.758,
+      "step": 1548
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996435300105905e-05,
+      "loss": 0.7459,
+      "step": 1549
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964253971929733e-05,
+      "loss": 0.811,
+      "step": 1550
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964154805683528e-05,
+      "loss": 0.7398,
+      "step": 1551
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996405550232179e-05,
+      "loss": 0.786,
+      "step": 1552
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963956061845897e-05,
+      "loss": 0.7618,
+      "step": 1553
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963856484257213e-05,
+      "loss": 0.7465,
+      "step": 1554
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996375676955711e-05,
+      "loss": 0.6923,
+      "step": 1555
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963656917746957e-05,
+      "loss": 0.7723,
+      "step": 1556
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963556928828133e-05,
+      "loss": 0.6531,
+      "step": 1557
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963456802802007e-05,
+      "loss": 0.7465,
+      "step": 1558
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963356539669965e-05,
+      "loss": 0.806,
+      "step": 1559
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996325613943338e-05,
+      "loss": 0.7837,
+      "step": 1560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963155602093637e-05,
+      "loss": 0.7504,
+      "step": 1561
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963054927652116e-05,
+      "loss": 0.6432,
+      "step": 1562
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962954116110205e-05,
+      "loss": 0.7252,
+      "step": 1563
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962853167469293e-05,
+      "loss": 0.712,
+      "step": 1564
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962752081730768e-05,
+      "loss": 0.7111,
+      "step": 1565
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996265085889602e-05,
+      "loss": 0.7812,
+      "step": 1566
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962549498966443e-05,
+      "loss": 0.5972,
+      "step": 1567
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962448001943427e-05,
+      "loss": 0.6394,
+      "step": 1568
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962346367828377e-05,
+      "loss": 0.695,
+      "step": 1569
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962244596622683e-05,
+      "loss": 0.7463,
+      "step": 1570
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962142688327753e-05,
+      "loss": 0.7672,
+      "step": 1571
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962040642944986e-05,
+      "loss": 0.7492,
+      "step": 1572
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961938460475784e-05,
+      "loss": 0.7122,
+      "step": 1573
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961836140921557e-05,
+      "loss": 0.6109,
+      "step": 1574
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961733684283714e-05,
+      "loss": 0.7433,
+      "step": 1575
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996163109056366e-05,
+      "loss": 0.7143,
+      "step": 1576
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961528359762806e-05,
+      "loss": 0.6834,
+      "step": 1577
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961425491882573e-05,
+      "loss": 0.6267,
+      "step": 1578
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961322486924372e-05,
+      "loss": 0.6311,
+      "step": 1579
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996121934488962e-05,
+      "loss": 0.7061,
+      "step": 1580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961116065779737e-05,
+      "loss": 0.7654,
+      "step": 1581
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961012649596144e-05,
+      "loss": 0.7917,
+      "step": 1582
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960909096340266e-05,
+      "loss": 0.6451,
+      "step": 1583
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960805406013524e-05,
+      "loss": 0.8521,
+      "step": 1584
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996070157861735e-05,
+      "loss": 0.6979,
+      "step": 1585
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960597614153166e-05,
+      "loss": 0.7559,
+      "step": 1586
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960493512622412e-05,
+      "loss": 0.7627,
+      "step": 1587
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996038927402651e-05,
+      "loss": 0.8187,
+      "step": 1588
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960284898366904e-05,
+      "loss": 0.6992,
+      "step": 1589
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960180385645025e-05,
+      "loss": 0.6876,
+      "step": 1590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960075735862313e-05,
+      "loss": 0.6533,
+      "step": 1591
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959970949020203e-05,
+      "loss": 0.7935,
+      "step": 1592
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959866025120143e-05,
+      "loss": 0.7203,
+      "step": 1593
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959760964163576e-05,
+      "loss": 0.5988,
+      "step": 1594
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959655766151948e-05,
+      "loss": 0.7624,
+      "step": 1595
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959550431086703e-05,
+      "loss": 0.6416,
+      "step": 1596
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995944495896929e-05,
+      "loss": 0.7699,
+      "step": 1597
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995933934980117e-05,
+      "loss": 0.6797,
+      "step": 1598
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959233603583786e-05,
+      "loss": 0.7695,
+      "step": 1599
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959127720318598e-05,
+      "loss": 0.738,
+      "step": 1600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995902170000706e-05,
+      "loss": 0.723,
+      "step": 1601
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958915542650632e-05,
+      "loss": 0.6431,
+      "step": 1602
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958809248250775e-05,
+      "loss": 0.7481,
+      "step": 1603
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958702816808956e-05,
+      "loss": 0.8016,
+      "step": 1604
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958596248326633e-05,
+      "loss": 0.7938,
+      "step": 1605
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958489542805278e-05,
+      "loss": 0.6295,
+      "step": 1606
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995838270024635e-05,
+      "loss": 0.7799,
+      "step": 1607
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958275720651335e-05,
+      "loss": 0.7175,
+      "step": 1608
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995816860402169e-05,
+      "loss": 0.815,
+      "step": 1609
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958061350358896e-05,
+      "loss": 0.8203,
+      "step": 1610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957953959664425e-05,
+      "loss": 0.746,
+      "step": 1611
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957846431939765e-05,
+      "loss": 0.7622,
+      "step": 1612
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957738767186385e-05,
+      "loss": 0.749,
+      "step": 1613
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957630965405767e-05,
+      "loss": 0.7285,
+      "step": 1614
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99575230265994e-05,
+      "loss": 0.7948,
+      "step": 1615
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995741495076877e-05,
+      "loss": 0.679,
+      "step": 1616
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957306737915353e-05,
+      "loss": 0.7477,
+      "step": 1617
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995719838804065e-05,
+      "loss": 0.8458,
+      "step": 1618
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995708990114615e-05,
+      "loss": 0.7797,
+      "step": 1619
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956981277233342e-05,
+      "loss": 0.7599,
+      "step": 1620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995687251630372e-05,
+      "loss": 0.7338,
+      "step": 1621
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995676361835879e-05,
+      "loss": 0.7131,
+      "step": 1622
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956654583400037e-05,
+      "loss": 0.7186,
+      "step": 1623
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956545411428972e-05,
+      "loss": 0.6614,
+      "step": 1624
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956436102447093e-05,
+      "loss": 0.7593,
+      "step": 1625
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956326656455903e-05,
+      "loss": 0.7987,
+      "step": 1626
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995621707345691e-05,
+      "loss": 0.7215,
+      "step": 1627
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956107353451622e-05,
+      "loss": 0.7483,
+      "step": 1628
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955997496441553e-05,
+      "loss": 0.8145,
+      "step": 1629
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955887502428205e-05,
+      "loss": 0.7858,
+      "step": 1630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955777371413102e-05,
+      "loss": 0.6872,
+      "step": 1631
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955667103397748e-05,
+      "loss": 0.7834,
+      "step": 1632
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995555669838367e-05,
+      "loss": 0.8521,
+      "step": 1633
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955446156372387e-05,
+      "loss": 0.7369,
+      "step": 1634
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955335477365415e-05,
+      "loss": 0.7674,
+      "step": 1635
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955224661364278e-05,
+      "loss": 0.6274,
+      "step": 1636
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955113708370503e-05,
+      "loss": 0.5999,
+      "step": 1637
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955002618385615e-05,
+      "loss": 0.6223,
+      "step": 1638
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954891391411146e-05,
+      "loss": 0.8126,
+      "step": 1639
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954780027448624e-05,
+      "loss": 0.6327,
+      "step": 1640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995466852649958e-05,
+      "loss": 0.6597,
+      "step": 1641
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954556888565554e-05,
+      "loss": 0.7702,
+      "step": 1642
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954445113648074e-05,
+      "loss": 0.7609,
+      "step": 1643
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954333201748682e-05,
+      "loss": 0.7421,
+      "step": 1644
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954221152868922e-05,
+      "loss": 0.6494,
+      "step": 1645
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995410896701033e-05,
+      "loss": 0.6113,
+      "step": 1646
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953996644174453e-05,
+      "loss": 0.775,
+      "step": 1647
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953884184362838e-05,
+      "loss": 0.7367,
+      "step": 1648
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953771587577027e-05,
+      "loss": 0.6083,
+      "step": 1649
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953658853818572e-05,
+      "loss": 0.6838,
+      "step": 1650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995354598308903e-05,
+      "loss": 0.8064,
+      "step": 1651
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953432975389947e-05,
+      "loss": 0.5885,
+      "step": 1652
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995331983072288e-05,
+      "loss": 0.8147,
+      "step": 1653
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953206549089386e-05,
+      "loss": 0.7584,
+      "step": 1654
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953093130491024e-05,
+      "loss": 0.7928,
+      "step": 1655
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952979574929355e-05,
+      "loss": 0.7648,
+      "step": 1656
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952865882405942e-05,
+      "loss": 0.7571,
+      "step": 1657
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995275205292235e-05,
+      "loss": 0.7009,
+      "step": 1658
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952638086480146e-05,
+      "loss": 0.7522,
+      "step": 1659
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995252398308089e-05,
+      "loss": 0.6456,
+      "step": 1660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952409742726163e-05,
+      "loss": 0.7725,
+      "step": 1661
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952295365417537e-05,
+      "loss": 0.7621,
+      "step": 1662
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995218085115658e-05,
+      "loss": 0.7503,
+      "step": 1663
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952066199944867e-05,
+      "loss": 0.732,
+      "step": 1664
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951951411783977e-05,
+      "loss": 0.7303,
+      "step": 1665
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951836486675495e-05,
+      "loss": 0.6514,
+      "step": 1666
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951721424621e-05,
+      "loss": 0.6344,
+      "step": 1667
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951606225622066e-05,
+      "loss": 0.7691,
+      "step": 1668
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995149088968029e-05,
+      "loss": 0.7897,
+      "step": 1669
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951375416797258e-05,
+      "loss": 0.7117,
+      "step": 1670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951259806974558e-05,
+      "loss": 0.7401,
+      "step": 1671
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951144060213773e-05,
+      "loss": 0.6493,
+      "step": 1672
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951028176516505e-05,
+      "loss": 0.6366,
+      "step": 1673
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950912155884344e-05,
+      "loss": 0.688,
+      "step": 1674
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950795998318888e-05,
+      "loss": 0.8048,
+      "step": 1675
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950679703821737e-05,
+      "loss": 0.702,
+      "step": 1676
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995056327239449e-05,
+      "loss": 0.7521,
+      "step": 1677
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995044670403875e-05,
+      "loss": 0.8279,
+      "step": 1678
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995032999875612e-05,
+      "loss": 0.688,
+      "step": 1679
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950213156548202e-05,
+      "loss": 0.7796,
+      "step": 1680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995009617741661e-05,
+      "loss": 0.6397,
+      "step": 1681
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949979061362953e-05,
+      "loss": 0.6119,
+      "step": 1682
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994986180838884e-05,
+      "loss": 0.662,
+      "step": 1683
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949744418495886e-05,
+      "loss": 0.7443,
+      "step": 1684
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994962689168571e-05,
+      "loss": 0.7732,
+      "step": 1685
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994950922795992e-05,
+      "loss": 0.5952,
+      "step": 1686
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949391427320144e-05,
+      "loss": 0.7262,
+      "step": 1687
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949273489767998e-05,
+      "loss": 0.6533,
+      "step": 1688
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949155415305105e-05,
+      "loss": 0.8006,
+      "step": 1689
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949037203933095e-05,
+      "loss": 0.77,
+      "step": 1690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994891885565359e-05,
+      "loss": 0.7189,
+      "step": 1691
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948800370468218e-05,
+      "loss": 0.6816,
+      "step": 1692
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948681748378612e-05,
+      "loss": 0.6678,
+      "step": 1693
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948562989386406e-05,
+      "loss": 0.7514,
+      "step": 1694
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948444093493227e-05,
+      "loss": 0.615,
+      "step": 1695
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994832506070072e-05,
+      "loss": 0.6728,
+      "step": 1696
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948205891010515e-05,
+      "loss": 0.7354,
+      "step": 1697
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948086584424258e-05,
+      "loss": 0.6109,
+      "step": 1698
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947967140943588e-05,
+      "loss": 0.7491,
+      "step": 1699
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994784756057015e-05,
+      "loss": 0.7684,
+      "step": 1700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947727843305585e-05,
+      "loss": 0.7118,
+      "step": 1701
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947607989151548e-05,
+      "loss": 0.7247,
+      "step": 1702
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947487998109682e-05,
+      "loss": 0.6092,
+      "step": 1703
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994736787018164e-05,
+      "loss": 0.7802,
+      "step": 1704
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994724760536908e-05,
+      "loss": 0.6546,
+      "step": 1705
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994712720367365e-05,
+      "loss": 0.8361,
+      "step": 1706
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994700666509701e-05,
+      "loss": 0.7922,
+      "step": 1707
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946885989640816e-05,
+      "loss": 0.6507,
+      "step": 1708
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994676517730673e-05,
+      "loss": 0.7787,
+      "step": 1709
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946644228096418e-05,
+      "loss": 0.6592,
+      "step": 1710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946523142011544e-05,
+      "loss": 0.8177,
+      "step": 1711
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946401919053765e-05,
+      "loss": 0.6779,
+      "step": 1712
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946280559224763e-05,
+      "loss": 0.7964,
+      "step": 1713
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946159062526197e-05,
+      "loss": 0.7175,
+      "step": 1714
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946037428959744e-05,
+      "loss": 0.7576,
+      "step": 1715
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945915658527072e-05,
+      "loss": 0.769,
+      "step": 1716
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945793751229867e-05,
+      "loss": 0.7189,
+      "step": 1717
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945671707069797e-05,
+      "loss": 0.7445,
+      "step": 1718
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994554952604855e-05,
+      "loss": 0.7523,
+      "step": 1719
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945427208167797e-05,
+      "loss": 0.73,
+      "step": 1720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994530475342923e-05,
+      "loss": 0.741,
+      "step": 1721
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945182161834532e-05,
+      "loss": 0.6299,
+      "step": 1722
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945059433385388e-05,
+      "loss": 0.827,
+      "step": 1723
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944936568083488e-05,
+      "loss": 0.7231,
+      "step": 1724
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994481356593052e-05,
+      "loss": 0.6194,
+      "step": 1725
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944690426928178e-05,
+      "loss": 0.7428,
+      "step": 1726
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944567151078164e-05,
+      "loss": 0.7612,
+      "step": 1727
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944443738382164e-05,
+      "loss": 0.7193,
+      "step": 1728
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994432018884188e-05,
+      "loss": 0.6147,
+      "step": 1729
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994419650245901e-05,
+      "loss": 0.7298,
+      "step": 1730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994407267923526e-05,
+      "loss": 0.7714,
+      "step": 1731
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943948719172333e-05,
+      "loss": 0.7944,
+      "step": 1732
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943824622271934e-05,
+      "loss": 0.6495,
+      "step": 1733
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943700388535775e-05,
+      "loss": 0.6121,
+      "step": 1734
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943576017965556e-05,
+      "loss": 0.6613,
+      "step": 1735
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943451510562996e-05,
+      "loss": 0.5855,
+      "step": 1736
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943326866329805e-05,
+      "loss": 0.6311,
+      "step": 1737
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943202085267698e-05,
+      "loss": 0.7569,
+      "step": 1738
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943077167378395e-05,
+      "loss": 0.7613,
+      "step": 1739
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942952112663616e-05,
+      "loss": 0.7676,
+      "step": 1740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942826921125077e-05,
+      "loss": 0.6761,
+      "step": 1741
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942701592764503e-05,
+      "loss": 0.7725,
+      "step": 1742
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942576127583618e-05,
+      "loss": 0.708,
+      "step": 1743
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942450525584148e-05,
+      "loss": 0.7343,
+      "step": 1744
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942324786767826e-05,
+      "loss": 0.7142,
+      "step": 1745
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942198911136375e-05,
+      "loss": 0.6654,
+      "step": 1746
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942072898691535e-05,
+      "loss": 0.8299,
+      "step": 1747
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994194674943503e-05,
+      "loss": 0.708,
+      "step": 1748
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941820463368612e-05,
+      "loss": 0.6123,
+      "step": 1749
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941694040494002e-05,
+      "loss": 0.7661,
+      "step": 1750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941567480812945e-05,
+      "loss": 0.8011,
+      "step": 1751
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941440784327188e-05,
+      "loss": 0.8145,
+      "step": 1752
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994131395103847e-05,
+      "loss": 0.7821,
+      "step": 1753
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941186980948534e-05,
+      "loss": 0.8077,
+      "step": 1754
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941059874059135e-05,
+      "loss": 0.8073,
+      "step": 1755
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994093263037201e-05,
+      "loss": 0.6746,
+      "step": 1756
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940805249888923e-05,
+      "loss": 0.6522,
+      "step": 1757
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994067773261162e-05,
+      "loss": 0.7907,
+      "step": 1758
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940550078541855e-05,
+      "loss": 0.786,
+      "step": 1759
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940422287681388e-05,
+      "loss": 0.7437,
+      "step": 1760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940294360031974e-05,
+      "loss": 0.6514,
+      "step": 1761
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994016629559538e-05,
+      "loss": 0.7142,
+      "step": 1762
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940038094373358e-05,
+      "loss": 0.7588,
+      "step": 1763
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993990975636768e-05,
+      "loss": 0.7374,
+      "step": 1764
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939781281580108e-05,
+      "loss": 0.7745,
+      "step": 1765
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939652670012416e-05,
+      "loss": 0.7352,
+      "step": 1766
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939523921666363e-05,
+      "loss": 0.7155,
+      "step": 1767
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939395036543733e-05,
+      "loss": 0.7987,
+      "step": 1768
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993926601464629e-05,
+      "loss": 0.7767,
+      "step": 1769
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939136855975815e-05,
+      "loss": 0.7362,
+      "step": 1770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939007560534084e-05,
+      "loss": 0.7173,
+      "step": 1771
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938878128322875e-05,
+      "loss": 0.7153,
+      "step": 1772
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993874855934397e-05,
+      "loss": 0.75,
+      "step": 1773
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938618853599155e-05,
+      "loss": 0.7707,
+      "step": 1774
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938489011090208e-05,
+      "loss": 0.7281,
+      "step": 1775
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938359031818918e-05,
+      "loss": 0.6127,
+      "step": 1776
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993822891578708e-05,
+      "loss": 0.731,
+      "step": 1777
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938098662996475e-05,
+      "loss": 0.7062,
+      "step": 1778
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937968273448904e-05,
+      "loss": 0.7503,
+      "step": 1779
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937837747146155e-05,
+      "loss": 0.6709,
+      "step": 1780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993770708409003e-05,
+      "loss": 0.8273,
+      "step": 1781
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937576284282317e-05,
+      "loss": 0.774,
+      "step": 1782
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937445347724826e-05,
+      "loss": 0.7128,
+      "step": 1783
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937314274419357e-05,
+      "loss": 0.6853,
+      "step": 1784
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993718306436771e-05,
+      "loss": 0.5523,
+      "step": 1785
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937051717571688e-05,
+      "loss": 0.6424,
+      "step": 1786
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993692023403311e-05,
+      "loss": 0.7992,
+      "step": 1787
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936788613753774e-05,
+      "loss": 0.7402,
+      "step": 1788
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936656856735498e-05,
+      "loss": 0.7329,
+      "step": 1789
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993652496298009e-05,
+      "loss": 0.7751,
+      "step": 1790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936392932489372e-05,
+      "loss": 0.7725,
+      "step": 1791
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936260765265152e-05,
+      "loss": 0.7808,
+      "step": 1792
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936128461309258e-05,
+      "loss": 0.8292,
+      "step": 1793
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99359960206235e-05,
+      "loss": 0.6222,
+      "step": 1794
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935863443209707e-05,
+      "loss": 0.775,
+      "step": 1795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935730729069703e-05,
+      "loss": 0.6946,
+      "step": 1796
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935597878205316e-05,
+      "loss": 0.8355,
+      "step": 1797
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935464890618373e-05,
+      "loss": 0.637,
+      "step": 1798
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99353317663107e-05,
+      "loss": 0.5784,
+      "step": 1799
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993519850528413e-05,
+      "loss": 0.7294,
+      "step": 1800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935065107540505e-05,
+      "loss": 0.7873,
+      "step": 1801
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934931573081648e-05,
+      "loss": 0.7973,
+      "step": 1802
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993479790190941e-05,
+      "loss": 0.6866,
+      "step": 1803
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934664094025616e-05,
+      "loss": 0.7604,
+      "step": 1804
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993453014943212e-05,
+      "loss": 0.7875,
+      "step": 1805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993439606813076e-05,
+      "loss": 0.7572,
+      "step": 1806
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934261850123374e-05,
+      "loss": 0.6944,
+      "step": 1807
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934127495411824e-05,
+      "loss": 0.7704,
+      "step": 1808
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933993003997948e-05,
+      "loss": 0.7652,
+      "step": 1809
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933858375883597e-05,
+      "loss": 0.5986,
+      "step": 1810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993372361107063e-05,
+      "loss": 0.6656,
+      "step": 1811
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933588709560895e-05,
+      "loss": 0.768,
+      "step": 1812
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933453671356254e-05,
+      "loss": 0.7404,
+      "step": 1813
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933318496458557e-05,
+      "loss": 0.7599,
+      "step": 1814
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933183184869675e-05,
+      "loss": 0.5977,
+      "step": 1815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993304773659146e-05,
+      "loss": 0.707,
+      "step": 1816
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932912151625782e-05,
+      "loss": 0.7237,
+      "step": 1817
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932776429974505e-05,
+      "loss": 0.7225,
+      "step": 1818
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932640571639496e-05,
+      "loss": 0.7635,
+      "step": 1819
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932504576622626e-05,
+      "loss": 0.7596,
+      "step": 1820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932368444925765e-05,
+      "loss": 0.7645,
+      "step": 1821
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932232176550787e-05,
+      "loss": 0.7285,
+      "step": 1822
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993209577149957e-05,
+      "loss": 0.6689,
+      "step": 1823
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931959229773983e-05,
+      "loss": 0.8335,
+      "step": 1824
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931822551375912e-05,
+      "loss": 0.6506,
+      "step": 1825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931685736307235e-05,
+      "loss": 0.7536,
+      "step": 1826
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931548784569836e-05,
+      "loss": 0.8442,
+      "step": 1827
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931411696165602e-05,
+      "loss": 0.7684,
+      "step": 1828
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931274471096414e-05,
+      "loss": 0.7371,
+      "step": 1829
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931137109364162e-05,
+      "loss": 0.7809,
+      "step": 1830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930999610970737e-05,
+      "loss": 0.7098,
+      "step": 1831
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930861975918035e-05,
+      "loss": 0.7733,
+      "step": 1832
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930724204207942e-05,
+      "loss": 0.7468,
+      "step": 1833
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993058629584236e-05,
+      "loss": 0.7883,
+      "step": 1834
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930448250823186e-05,
+      "loss": 0.6358,
+      "step": 1835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930310069152314e-05,
+      "loss": 0.74,
+      "step": 1836
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930171750831653e-05,
+      "loss": 0.7452,
+      "step": 1837
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930033295863104e-05,
+      "loss": 0.6922,
+      "step": 1838
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992989470424857e-05,
+      "loss": 0.7413,
+      "step": 1839
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992975597598996e-05,
+      "loss": 0.7426,
+      "step": 1840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929617111089184e-05,
+      "loss": 0.7308,
+      "step": 1841
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992947810954815e-05,
+      "loss": 0.7535,
+      "step": 1842
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992933897136877e-05,
+      "loss": 0.663,
+      "step": 1843
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929199696552966e-05,
+      "loss": 0.7253,
+      "step": 1844
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992906028510265e-05,
+      "loss": 0.7206,
+      "step": 1845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928920737019735e-05,
+      "loss": 0.789,
+      "step": 1846
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992878105230615e-05,
+      "loss": 0.6212,
+      "step": 1847
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992864123096381e-05,
+      "loss": 0.7471,
+      "step": 1848
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928501272994646e-05,
+      "loss": 0.6842,
+      "step": 1849
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928361178400582e-05,
+      "loss": 0.7316,
+      "step": 1850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992822094718354e-05,
+      "loss": 0.7985,
+      "step": 1851
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992808057934546e-05,
+      "loss": 0.7084,
+      "step": 1852
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927940074888262e-05,
+      "loss": 0.5729,
+      "step": 1853
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927799433813887e-05,
+      "loss": 0.7617,
+      "step": 1854
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927658656124268e-05,
+      "loss": 0.8126,
+      "step": 1855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927517741821343e-05,
+      "loss": 0.7289,
+      "step": 1856
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992737669090705e-05,
+      "loss": 0.7358,
+      "step": 1857
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927235503383333e-05,
+      "loss": 0.6964,
+      "step": 1858
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927094179252132e-05,
+      "loss": 0.8026,
+      "step": 1859
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926952718515392e-05,
+      "loss": 0.6441,
+      "step": 1860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926811121175057e-05,
+      "loss": 0.6962,
+      "step": 1861
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926669387233084e-05,
+      "loss": 0.7461,
+      "step": 1862
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926527516691413e-05,
+      "loss": 0.7663,
+      "step": 1863
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926385509552004e-05,
+      "loss": 0.7149,
+      "step": 1864
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926243365816806e-05,
+      "loss": 0.776,
+      "step": 1865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992610108548778e-05,
+      "loss": 0.6185,
+      "step": 1866
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992595866856688e-05,
+      "loss": 0.7142,
+      "step": 1867
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992581611505607e-05,
+      "loss": 0.603,
+      "step": 1868
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925673424957305e-05,
+      "loss": 0.76,
+      "step": 1869
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992553059827255e-05,
+      "loss": 0.5943,
+      "step": 1870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925387635003777e-05,
+      "loss": 0.8319,
+      "step": 1871
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925244535152947e-05,
+      "loss": 0.7002,
+      "step": 1872
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992510129872203e-05,
+      "loss": 0.6105,
+      "step": 1873
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924957925713003e-05,
+      "loss": 0.6016,
+      "step": 1874
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992481441612783e-05,
+      "loss": 0.7532,
+      "step": 1875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992467076996849e-05,
+      "loss": 0.775,
+      "step": 1876
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992452698723696e-05,
+      "loss": 0.5953,
+      "step": 1877
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992438306793522e-05,
+      "loss": 0.6419,
+      "step": 1878
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924239012065247e-05,
+      "loss": 0.7708,
+      "step": 1879
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924094819629024e-05,
+      "loss": 0.7711,
+      "step": 1880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992395049062854e-05,
+      "loss": 0.641,
+      "step": 1881
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923806025065775e-05,
+      "loss": 0.7361,
+      "step": 1882
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992366142294272e-05,
+      "loss": 0.7402,
+      "step": 1883
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992351668426136e-05,
+      "loss": 0.8033,
+      "step": 1884
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923371809023696e-05,
+      "loss": 0.7492,
+      "step": 1885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923226797231713e-05,
+      "loss": 0.6036,
+      "step": 1886
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923081648887415e-05,
+      "loss": 0.7565,
+      "step": 1887
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992293636399279e-05,
+      "loss": 0.7083,
+      "step": 1888
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922790942549842e-05,
+      "loss": 0.7133,
+      "step": 1889
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922645384560572e-05,
+      "loss": 0.635,
+      "step": 1890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922499690026983e-05,
+      "loss": 0.7677,
+      "step": 1891
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992235385895108e-05,
+      "loss": 0.6692,
+      "step": 1892
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922207891334867e-05,
+      "loss": 0.73,
+      "step": 1893
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992206178718036e-05,
+      "loss": 0.7269,
+      "step": 1894
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921915546489558e-05,
+      "loss": 0.7877,
+      "step": 1895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921769169264486e-05,
+      "loss": 0.7386,
+      "step": 1896
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992162265550715e-05,
+      "loss": 0.6545,
+      "step": 1897
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921476005219566e-05,
+      "loss": 0.6361,
+      "step": 1898
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921329218403757e-05,
+      "loss": 0.7184,
+      "step": 1899
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921182295061737e-05,
+      "loss": 0.8061,
+      "step": 1900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921035235195537e-05,
+      "loss": 0.7896,
+      "step": 1901
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920888038807172e-05,
+      "loss": 0.7286,
+      "step": 1902
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920740705898668e-05,
+      "loss": 0.6261,
+      "step": 1903
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920593236472057e-05,
+      "loss": 0.7065,
+      "step": 1904
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920445630529367e-05,
+      "loss": 0.8456,
+      "step": 1905
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920297888072623e-05,
+      "loss": 0.6253,
+      "step": 1906
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992015000910387e-05,
+      "loss": 0.7846,
+      "step": 1907
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920001993625133e-05,
+      "loss": 0.7609,
+      "step": 1908
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919853841638452e-05,
+      "loss": 0.6104,
+      "step": 1909
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919705553145866e-05,
+      "loss": 0.694,
+      "step": 1910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919557128149418e-05,
+      "loss": 0.7229,
+      "step": 1911
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919408566651143e-05,
+      "loss": 0.7832,
+      "step": 1912
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919259868653093e-05,
+      "loss": 0.7283,
+      "step": 1913
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919111034157312e-05,
+      "loss": 0.7688,
+      "step": 1914
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918962063165845e-05,
+      "loss": 0.6476,
+      "step": 1915
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918812955680748e-05,
+      "loss": 0.7781,
+      "step": 1916
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918663711704067e-05,
+      "loss": 0.8193,
+      "step": 1917
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991851433123786e-05,
+      "loss": 0.7486,
+      "step": 1918
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918364814284177e-05,
+      "loss": 0.7983,
+      "step": 1919
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991821516084508e-05,
+      "loss": 0.8162,
+      "step": 1920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918065370922632e-05,
+      "loss": 0.5847,
+      "step": 1921
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991791544451888e-05,
+      "loss": 0.6591,
+      "step": 1922
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917765381635907e-05,
+      "loss": 0.6388,
+      "step": 1923
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917615182275766e-05,
+      "loss": 0.7773,
+      "step": 1924
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991746484644052e-05,
+      "loss": 0.7929,
+      "step": 1925
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991731437413225e-05,
+      "loss": 0.6209,
+      "step": 1926
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917163765353016e-05,
+      "loss": 0.7703,
+      "step": 1927
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917013020104898e-05,
+      "loss": 0.7258,
+      "step": 1928
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916862138389962e-05,
+      "loss": 0.6445,
+      "step": 1929
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991671112021029e-05,
+      "loss": 0.8158,
+      "step": 1930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916559965567964e-05,
+      "loss": 0.6008,
+      "step": 1931
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916408674465058e-05,
+      "loss": 0.5781,
+      "step": 1932
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916257246903653e-05,
+      "loss": 0.811,
+      "step": 1933
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916105682885837e-05,
+      "loss": 0.7132,
+      "step": 1934
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915953982413693e-05,
+      "loss": 0.7347,
+      "step": 1935
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991580214548931e-05,
+      "loss": 0.698,
+      "step": 1936
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915650172114777e-05,
+      "loss": 0.7934,
+      "step": 1937
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915498062292182e-05,
+      "loss": 0.7807,
+      "step": 1938
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915345816023622e-05,
+      "loss": 0.6285,
+      "step": 1939
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915193433311196e-05,
+      "loss": 0.6504,
+      "step": 1940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915040914156992e-05,
+      "loss": 0.7386,
+      "step": 1941
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991488825856311e-05,
+      "loss": 0.7764,
+      "step": 1942
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991473546653166e-05,
+      "loss": 0.7349,
+      "step": 1943
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914582538064735e-05,
+      "loss": 0.6627,
+      "step": 1944
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914429473164442e-05,
+      "loss": 0.7535,
+      "step": 1945
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914276271832888e-05,
+      "loss": 0.7912,
+      "step": 1946
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914122934072182e-05,
+      "loss": 0.8204,
+      "step": 1947
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991396945988443e-05,
+      "loss": 0.775,
+      "step": 1948
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991381584927175e-05,
+      "loss": 0.801,
+      "step": 1949
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913662102236255e-05,
+      "loss": 0.652,
+      "step": 1950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913508218780055e-05,
+      "loss": 0.6204,
+      "step": 1951
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913354198905273e-05,
+      "loss": 0.7881,
+      "step": 1952
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913200042614025e-05,
+      "loss": 0.7536,
+      "step": 1953
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913045749908436e-05,
+      "loss": 0.782,
+      "step": 1954
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912891320790628e-05,
+      "loss": 0.6884,
+      "step": 1955
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991273675526272e-05,
+      "loss": 0.701,
+      "step": 1956
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912582053326847e-05,
+      "loss": 0.7822,
+      "step": 1957
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912427214985137e-05,
+      "loss": 0.7191,
+      "step": 1958
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912272240239715e-05,
+      "loss": 0.7644,
+      "step": 1959
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912117129092722e-05,
+      "loss": 0.7181,
+      "step": 1960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911961881546286e-05,
+      "loss": 0.6023,
+      "step": 1961
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991180649760254e-05,
+      "loss": 0.7715,
+      "step": 1962
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911650977263634e-05,
+      "loss": 0.7842,
+      "step": 1963
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911495320531698e-05,
+      "loss": 0.7338,
+      "step": 1964
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911339527408877e-05,
+      "loss": 0.7481,
+      "step": 1965
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911183597897315e-05,
+      "loss": 0.7522,
+      "step": 1966
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991102753199916e-05,
+      "loss": 0.7598,
+      "step": 1967
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991087132971656e-05,
+      "loss": 0.741,
+      "step": 1968
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991071499105166e-05,
+      "loss": 0.78,
+      "step": 1969
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991055851600661e-05,
+      "loss": 0.7264,
+      "step": 1970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991040190458357e-05,
+      "loss": 0.6124,
+      "step": 1971
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910245156784688e-05,
+      "loss": 0.7391,
+      "step": 1972
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910088272612128e-05,
+      "loss": 0.7412,
+      "step": 1973
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909931252068046e-05,
+      "loss": 0.7709,
+      "step": 1974
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909774095154603e-05,
+      "loss": 0.6091,
+      "step": 1975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990961680187396e-05,
+      "loss": 0.7356,
+      "step": 1976
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990945937222828e-05,
+      "loss": 0.7631,
+      "step": 1977
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909301806219737e-05,
+      "loss": 0.754,
+      "step": 1978
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909144103850487e-05,
+      "loss": 0.8175,
+      "step": 1979
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908986265122712e-05,
+      "loss": 0.6447,
+      "step": 1980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908828290038577e-05,
+      "loss": 0.7652,
+      "step": 1981
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908670178600264e-05,
+      "loss": 0.7203,
+      "step": 1982
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908511930809935e-05,
+      "loss": 0.7135,
+      "step": 1983
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908353546669777e-05,
+      "loss": 0.7709,
+      "step": 1984
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990819502618197e-05,
+      "loss": 0.7139,
+      "step": 1985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908036369348695e-05,
+      "loss": 0.7249,
+      "step": 1986
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907877576172133e-05,
+      "loss": 0.7783,
+      "step": 1987
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907718646654464e-05,
+      "loss": 0.7418,
+      "step": 1988
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907559580797885e-05,
+      "loss": 0.6582,
+      "step": 1989
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907400378604583e-05,
+      "loss": 0.6237,
+      "step": 1990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907241040076744e-05,
+      "loss": 0.7183,
+      "step": 1991
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907081565216564e-05,
+      "loss": 0.7843,
+      "step": 1992
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906921954026234e-05,
+      "loss": 0.7073,
+      "step": 1993
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906762206507954e-05,
+      "loss": 0.7746,
+      "step": 1994
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906602322663922e-05,
+      "loss": 0.6317,
+      "step": 1995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906442302496338e-05,
+      "loss": 0.7762,
+      "step": 1996
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99062821460074e-05,
+      "loss": 0.7485,
+      "step": 1997
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990612185319932e-05,
+      "loss": 0.7249,
+      "step": 1998
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905961424074294e-05,
+      "loss": 0.7469,
+      "step": 1999
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990580085863454e-05,
+      "loss": 0.7471,
+      "step": 2000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990564015688226e-05,
+      "loss": 0.6093,
+      "step": 2001
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905479318819664e-05,
+      "loss": 0.7881,
+      "step": 2002
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990531834444897e-05,
+      "loss": 0.7351,
+      "step": 2003
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905157233772396e-05,
+      "loss": 0.7409,
+      "step": 2004
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904995986792153e-05,
+      "loss": 0.6551,
+      "step": 2005
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904834603510463e-05,
+      "loss": 0.7813,
+      "step": 2006
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904673083929546e-05,
+      "loss": 0.6209,
+      "step": 2007
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990451142805162e-05,
+      "loss": 0.6562,
+      "step": 2008
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990434963587892e-05,
+      "loss": 0.7184,
+      "step": 2009
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990418770741366e-05,
+      "loss": 0.5629,
+      "step": 2010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990402564265808e-05,
+      "loss": 0.737,
+      "step": 2011
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99038634416144e-05,
+      "loss": 0.7327,
+      "step": 2012
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990370110428486e-05,
+      "loss": 0.6302,
+      "step": 2013
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903538630671687e-05,
+      "loss": 0.7718,
+      "step": 2014
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903376020777126e-05,
+      "loss": 0.7025,
+      "step": 2015
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903213274603404e-05,
+      "loss": 0.643,
+      "step": 2016
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990305039215277e-05,
+      "loss": 0.7274,
+      "step": 2017
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902887373427458e-05,
+      "loss": 0.7276,
+      "step": 2018
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902724218429716e-05,
+      "loss": 0.7315,
+      "step": 2019
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990256092716179e-05,
+      "loss": 0.7403,
+      "step": 2020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902397499625917e-05,
+      "loss": 0.7878,
+      "step": 2021
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990223393582436e-05,
+      "loss": 0.6123,
+      "step": 2022
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902070235759358e-05,
+      "loss": 0.7573,
+      "step": 2023
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901906399433172e-05,
+      "loss": 0.7107,
+      "step": 2024
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901742426848055e-05,
+      "loss": 0.7974,
+      "step": 2025
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901578318006258e-05,
+      "loss": 0.7273,
+      "step": 2026
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901414072910045e-05,
+      "loss": 0.7502,
+      "step": 2027
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901249691561672e-05,
+      "loss": 0.8131,
+      "step": 2028
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901085173963407e-05,
+      "loss": 0.6451,
+      "step": 2029
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900920520117507e-05,
+      "loss": 0.7713,
+      "step": 2030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900755730026243e-05,
+      "loss": 0.6455,
+      "step": 2031
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990059080369188e-05,
+      "loss": 0.7219,
+      "step": 2032
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990042574111669e-05,
+      "loss": 0.725,
+      "step": 2033
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990026054230294e-05,
+      "loss": 0.7572,
+      "step": 2034
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900095207252905e-05,
+      "loss": 0.5957,
+      "step": 2035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899929735968864e-05,
+      "loss": 0.7355,
+      "step": 2036
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989976412845309e-05,
+      "loss": 0.77,
+      "step": 2037
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989959838470786e-05,
+      "loss": 0.7894,
+      "step": 2038
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989943250473546e-05,
+      "loss": 0.7399,
+      "step": 2039
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989926648853817e-05,
+      "loss": 0.6893,
+      "step": 2040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899100336118276e-05,
+      "loss": 0.7476,
+      "step": 2041
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898934047478063e-05,
+      "loss": 0.5945,
+      "step": 2042
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898767622619818e-05,
+      "loss": 0.6177,
+      "step": 2043
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898601061545833e-05,
+      "loss": 0.7369,
+      "step": 2044
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98984343642584e-05,
+      "loss": 0.5629,
+      "step": 2045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898267530759815e-05,
+      "loss": 0.7496,
+      "step": 2046
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989810056105237e-05,
+      "loss": 0.6954,
+      "step": 2047
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989793345513836e-05,
+      "loss": 0.7803,
+      "step": 2048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897766213020094e-05,
+      "loss": 0.5926,
+      "step": 2049
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897598834699865e-05,
+      "loss": 0.6266,
+      "step": 2050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989743132017998e-05,
+      "loss": 0.589,
+      "step": 2051
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897263669462745e-05,
+      "loss": 0.8119,
+      "step": 2052
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989709588255046e-05,
+      "loss": 0.7331,
+      "step": 2053
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896927959445444e-05,
+      "loss": 0.7692,
+      "step": 2054
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896759900150004e-05,
+      "loss": 0.7307,
+      "step": 2055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989659170466645e-05,
+      "loss": 0.8123,
+      "step": 2056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98964233729971e-05,
+      "loss": 0.7483,
+      "step": 2057
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896254905144265e-05,
+      "loss": 0.7873,
+      "step": 2058
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896086301110268e-05,
+      "loss": 0.6793,
+      "step": 2059
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989591756089743e-05,
+      "loss": 0.6666,
+      "step": 2060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895748684508068e-05,
+      "loss": 0.6565,
+      "step": 2061
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895579671944514e-05,
+      "loss": 0.7575,
+      "step": 2062
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895410523209084e-05,
+      "loss": 0.6277,
+      "step": 2063
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989524123830411e-05,
+      "loss": 0.6767,
+      "step": 2064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895071817231925e-05,
+      "loss": 0.7611,
+      "step": 2065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894902259994856e-05,
+      "loss": 0.8244,
+      "step": 2066
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894732566595238e-05,
+      "loss": 0.5985,
+      "step": 2067
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894562737035406e-05,
+      "loss": 0.6417,
+      "step": 2068
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894392771317694e-05,
+      "loss": 0.7592,
+      "step": 2069
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894222669444447e-05,
+      "loss": 0.7579,
+      "step": 2070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9894052431418e-05,
+      "loss": 0.7055,
+      "step": 2071
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893882057240698e-05,
+      "loss": 0.6993,
+      "step": 2072
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893711546914885e-05,
+      "loss": 0.7655,
+      "step": 2073
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989354090044291e-05,
+      "loss": 0.5796,
+      "step": 2074
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893370117827114e-05,
+      "loss": 0.5793,
+      "step": 2075
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893199199069858e-05,
+      "loss": 0.754,
+      "step": 2076
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893028144173482e-05,
+      "loss": 0.7345,
+      "step": 2077
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989285695314035e-05,
+      "loss": 0.6227,
+      "step": 2078
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892685625972814e-05,
+      "loss": 0.7633,
+      "step": 2079
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989251416267323e-05,
+      "loss": 0.6473,
+      "step": 2080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989234256324396e-05,
+      "loss": 0.7099,
+      "step": 2081
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892170827687366e-05,
+      "loss": 0.5515,
+      "step": 2082
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891998956005806e-05,
+      "loss": 0.6627,
+      "step": 2083
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891826948201647e-05,
+      "loss": 0.8368,
+      "step": 2084
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989165480427726e-05,
+      "loss": 0.7393,
+      "step": 2085
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989148252423501e-05,
+      "loss": 0.7159,
+      "step": 2086
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891310108077272e-05,
+      "loss": 0.7385,
+      "step": 2087
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891137555806416e-05,
+      "loss": 0.665,
+      "step": 2088
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890964867424813e-05,
+      "loss": 0.5709,
+      "step": 2089
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890792042934844e-05,
+      "loss": 0.7148,
+      "step": 2090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989061908233888e-05,
+      "loss": 0.7229,
+      "step": 2091
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890445985639316e-05,
+      "loss": 0.7715,
+      "step": 2092
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989027275283852e-05,
+      "loss": 0.7378,
+      "step": 2093
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989009938393888e-05,
+      "loss": 0.7854,
+      "step": 2094
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889925878942785e-05,
+      "loss": 0.7388,
+      "step": 2095
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889752237852617e-05,
+      "loss": 0.7727,
+      "step": 2096
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988957846067077e-05,
+      "loss": 0.7353,
+      "step": 2097
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889404547399634e-05,
+      "loss": 0.6414,
+      "step": 2098
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889230498041598e-05,
+      "loss": 0.8103,
+      "step": 2099
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889056312599066e-05,
+      "loss": 0.7143,
+      "step": 2100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988888199107443e-05,
+      "loss": 0.6358,
+      "step": 2101
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888707533470085e-05,
+      "loss": 0.7654,
+      "step": 2102
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888532939788437e-05,
+      "loss": 0.7504,
+      "step": 2103
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888358210031885e-05,
+      "loss": 0.7768,
+      "step": 2104
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888183344202837e-05,
+      "loss": 0.6469,
+      "step": 2105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98880083423037e-05,
+      "loss": 0.712,
+      "step": 2106
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887833204336876e-05,
+      "loss": 0.7299,
+      "step": 2107
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887657930304782e-05,
+      "loss": 0.7451,
+      "step": 2108
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887482520209824e-05,
+      "loss": 0.7591,
+      "step": 2109
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988730697405442e-05,
+      "loss": 0.7346,
+      "step": 2110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887131291840983e-05,
+      "loss": 0.7512,
+      "step": 2111
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886955473571935e-05,
+      "loss": 0.777,
+      "step": 2112
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988677951924969e-05,
+      "loss": 0.7546,
+      "step": 2113
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886603428876673e-05,
+      "loss": 0.7736,
+      "step": 2114
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98864272024553e-05,
+      "loss": 0.7547,
+      "step": 2115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886250839988008e-05,
+      "loss": 0.6865,
+      "step": 2116
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886074341477218e-05,
+      "loss": 0.6103,
+      "step": 2117
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885897706925353e-05,
+      "loss": 0.7886,
+      "step": 2118
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885720936334855e-05,
+      "loss": 0.7455,
+      "step": 2119
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885544029708147e-05,
+      "loss": 0.7533,
+      "step": 2120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988536698704767e-05,
+      "loss": 0.7414,
+      "step": 2121
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885189808355855e-05,
+      "loss": 0.7689,
+      "step": 2122
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988501249363514e-05,
+      "loss": 0.7427,
+      "step": 2123
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988483504288797e-05,
+      "loss": 0.6903,
+      "step": 2124
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884657456116785e-05,
+      "loss": 0.7811,
+      "step": 2125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884479733324027e-05,
+      "loss": 0.7151,
+      "step": 2126
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988430187451214e-05,
+      "loss": 0.688,
+      "step": 2127
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884123879683578e-05,
+      "loss": 0.6015,
+      "step": 2128
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883945748840786e-05,
+      "loss": 0.6893,
+      "step": 2129
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883767481986217e-05,
+      "loss": 0.7814,
+      "step": 2130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883589079122317e-05,
+      "loss": 0.7145,
+      "step": 2131
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883410540251553e-05,
+      "loss": 0.7225,
+      "step": 2132
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883231865376372e-05,
+      "loss": 0.776,
+      "step": 2133
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883053054499236e-05,
+      "loss": 0.637,
+      "step": 2134
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882874107622608e-05,
+      "loss": 0.7838,
+      "step": 2135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988269502474895e-05,
+      "loss": 0.7588,
+      "step": 2136
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988251580588072e-05,
+      "loss": 0.6925,
+      "step": 2137
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882336451020392e-05,
+      "loss": 0.6777,
+      "step": 2138
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988215696017043e-05,
+      "loss": 0.7524,
+      "step": 2139
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881977333333307e-05,
+      "loss": 0.8115,
+      "step": 2140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988179757051149e-05,
+      "loss": 0.7107,
+      "step": 2141
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881617671707457e-05,
+      "loss": 0.6345,
+      "step": 2142
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988143763692368e-05,
+      "loss": 0.7561,
+      "step": 2143
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881257466162642e-05,
+      "loss": 0.6745,
+      "step": 2144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988107715942682e-05,
+      "loss": 0.6354,
+      "step": 2145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988089671671869e-05,
+      "loss": 0.7293,
+      "step": 2146
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988071613804074e-05,
+      "loss": 0.7614,
+      "step": 2147
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880535423395456e-05,
+      "loss": 0.7116,
+      "step": 2148
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988035457278532e-05,
+      "loss": 0.7316,
+      "step": 2149
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880173586212827e-05,
+      "loss": 0.6563,
+      "step": 2150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987999246368046e-05,
+      "loss": 0.7394,
+      "step": 2151
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987981120519072e-05,
+      "loss": 0.7292,
+      "step": 2152
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987962981074609e-05,
+      "loss": 0.6789,
+      "step": 2153
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879448280349077e-05,
+      "loss": 0.7155,
+      "step": 2154
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879266614002174e-05,
+      "loss": 0.6253,
+      "step": 2155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879084811707882e-05,
+      "loss": 0.6269,
+      "step": 2156
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878902873468702e-05,
+      "loss": 0.7716,
+      "step": 2157
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878720799287138e-05,
+      "loss": 0.7889,
+      "step": 2158
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878538589165696e-05,
+      "loss": 0.7073,
+      "step": 2159
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878356243106884e-05,
+      "loss": 0.7013,
+      "step": 2160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987817376111321e-05,
+      "loss": 0.7773,
+      "step": 2161
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877991143187186e-05,
+      "loss": 0.7679,
+      "step": 2162
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987780838933132e-05,
+      "loss": 0.5933,
+      "step": 2163
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877625499548136e-05,
+      "loss": 0.7611,
+      "step": 2164
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877442473840143e-05,
+      "loss": 0.7025,
+      "step": 2165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877259312209864e-05,
+      "loss": 0.7289,
+      "step": 2166
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877076014659817e-05,
+      "loss": 0.7774,
+      "step": 2167
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876892581192525e-05,
+      "loss": 0.7574,
+      "step": 2168
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876709011810513e-05,
+      "loss": 0.7164,
+      "step": 2169
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876525306516308e-05,
+      "loss": 0.5708,
+      "step": 2170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876341465312437e-05,
+      "loss": 0.6253,
+      "step": 2171
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876157488201426e-05,
+      "loss": 0.7318,
+      "step": 2172
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987597337518581e-05,
+      "loss": 0.7077,
+      "step": 2173
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875789126268127e-05,
+      "loss": 0.7377,
+      "step": 2174
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98756047414509e-05,
+      "loss": 0.6216,
+      "step": 2175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875420220736682e-05,
+      "loss": 0.7626,
+      "step": 2176
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875235564128003e-05,
+      "loss": 0.5922,
+      "step": 2177
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875050771627403e-05,
+      "loss": 0.6707,
+      "step": 2178
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987486584323743e-05,
+      "loss": 0.8209,
+      "step": 2179
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874680778960626e-05,
+      "loss": 0.5956,
+      "step": 2180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874495578799538e-05,
+      "loss": 0.6068,
+      "step": 2181
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987431024275671e-05,
+      "loss": 0.6889,
+      "step": 2182
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874124770834703e-05,
+      "loss": 0.7326,
+      "step": 2183
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873939163036058e-05,
+      "loss": 0.651,
+      "step": 2184
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873753419363336e-05,
+      "loss": 0.6195,
+      "step": 2185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873567539819093e-05,
+      "loss": 0.6328,
+      "step": 2186
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987338152440588e-05,
+      "loss": 0.6475,
+      "step": 2187
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873195373126264e-05,
+      "loss": 0.5888,
+      "step": 2188
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873009085982805e-05,
+      "loss": 0.608,
+      "step": 2189
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872822662978066e-05,
+      "loss": 0.7367,
+      "step": 2190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872636104114612e-05,
+      "loss": 0.7544,
+      "step": 2191
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872449409395012e-05,
+      "loss": 0.6314,
+      "step": 2192
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987226257882183e-05,
+      "loss": 0.6525,
+      "step": 2193
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872075612397642e-05,
+      "loss": 0.7475,
+      "step": 2194
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871888510125022e-05,
+      "loss": 0.673,
+      "step": 2195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987170127200654e-05,
+      "loss": 0.7932,
+      "step": 2196
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987151389804477e-05,
+      "loss": 0.7592,
+      "step": 2197
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98713263882423e-05,
+      "loss": 0.5763,
+      "step": 2198
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871138742601707e-05,
+      "loss": 0.7147,
+      "step": 2199
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870950961125564e-05,
+      "loss": 0.7295,
+      "step": 2200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870763043816467e-05,
+      "loss": 0.7619,
+      "step": 2201
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870574990676997e-05,
+      "loss": 0.7485,
+      "step": 2202
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870386801709746e-05,
+      "loss": 0.8059,
+      "step": 2203
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870198476917296e-05,
+      "loss": 0.7663,
+      "step": 2204
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987001001630224e-05,
+      "loss": 0.638,
+      "step": 2205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869821419867176e-05,
+      "loss": 0.7505,
+      "step": 2206
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869632687614693e-05,
+      "loss": 0.7821,
+      "step": 2207
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98694438195474e-05,
+      "loss": 0.683,
+      "step": 2208
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869254815667884e-05,
+      "loss": 0.78,
+      "step": 2209
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869065675978747e-05,
+      "loss": 0.65,
+      "step": 2210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868876400482598e-05,
+      "loss": 0.7467,
+      "step": 2211
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868686989182037e-05,
+      "loss": 0.6718,
+      "step": 2212
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986849744207967e-05,
+      "loss": 0.6921,
+      "step": 2213
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986830775917811e-05,
+      "loss": 0.79,
+      "step": 2214
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868117940479963e-05,
+      "loss": 0.757,
+      "step": 2215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986792798598784e-05,
+      "loss": 0.717,
+      "step": 2216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986773789570436e-05,
+      "loss": 0.7086,
+      "step": 2217
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867547669632133e-05,
+      "loss": 0.6278,
+      "step": 2218
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867357307773783e-05,
+      "loss": 0.7618,
+      "step": 2219
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867166810131925e-05,
+      "loss": 0.7854,
+      "step": 2220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866976176709185e-05,
+      "loss": 0.7861,
+      "step": 2221
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866785407508178e-05,
+      "loss": 0.747,
+      "step": 2222
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866594502531536e-05,
+      "loss": 0.7477,
+      "step": 2223
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866403461781883e-05,
+      "loss": 0.7774,
+      "step": 2224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986621228526185e-05,
+      "loss": 0.7308,
+      "step": 2225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986602097297407e-05,
+      "loss": 0.6415,
+      "step": 2226
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986582952492117e-05,
+      "loss": 0.6404,
+      "step": 2227
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986563794110578e-05,
+      "loss": 0.7384,
+      "step": 2228
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865446221530555e-05,
+      "loss": 0.7978,
+      "step": 2229
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865254366198112e-05,
+      "loss": 0.7278,
+      "step": 2230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865062375111105e-05,
+      "loss": 0.6247,
+      "step": 2231
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986487024827217e-05,
+      "loss": 0.72,
+      "step": 2232
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864677985683954e-05,
+      "loss": 0.691,
+      "step": 2233
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864485587349097e-05,
+      "loss": 0.6971,
+      "step": 2234
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864293053270258e-05,
+      "loss": 0.656,
+      "step": 2235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864100383450072e-05,
+      "loss": 0.6983,
+      "step": 2236
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98639075778912e-05,
+      "loss": 0.7757,
+      "step": 2237
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986371463659629e-05,
+      "loss": 0.6073,
+      "step": 2238
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863521559568008e-05,
+      "loss": 0.6975,
+      "step": 2239
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863328346808994e-05,
+      "loss": 0.7448,
+      "step": 2240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863134998321918e-05,
+      "loss": 0.7036,
+      "step": 2241
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862941514109437e-05,
+      "loss": 0.5992,
+      "step": 2242
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862747894174218e-05,
+      "loss": 0.7248,
+      "step": 2243
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862554138518916e-05,
+      "loss": 0.7577,
+      "step": 2244
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862360247146205e-05,
+      "loss": 0.7739,
+      "step": 2245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862166220058754e-05,
+      "loss": 0.7189,
+      "step": 2246
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861972057259228e-05,
+      "loss": 0.7034,
+      "step": 2247
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861777758750302e-05,
+      "loss": 0.7613,
+      "step": 2248
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986158332453465e-05,
+      "loss": 0.726,
+      "step": 2249
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861388754614944e-05,
+      "loss": 0.7755,
+      "step": 2250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861194048993865e-05,
+      "loss": 0.8004,
+      "step": 2251
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986099920767409e-05,
+      "loss": 0.7015,
+      "step": 2252
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98608042306583e-05,
+      "loss": 0.6458,
+      "step": 2253
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986060911794918e-05,
+      "loss": 0.6164,
+      "step": 2254
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860413869549417e-05,
+      "loss": 0.6199,
+      "step": 2255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860218485461693e-05,
+      "loss": 0.6276,
+      "step": 2256
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98600229656887e-05,
+      "loss": 0.7497,
+      "step": 2257
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859827310233125e-05,
+      "loss": 0.7309,
+      "step": 2258
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859631519097664e-05,
+      "loss": 0.7415,
+      "step": 2259
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859435592285007e-05,
+      "loss": 0.7493,
+      "step": 2260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985923952979786e-05,
+      "loss": 0.78,
+      "step": 2261
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859043331638906e-05,
+      "loss": 0.7858,
+      "step": 2262
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858846997810856e-05,
+      "loss": 0.6184,
+      "step": 2263
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985865052831641e-05,
+      "loss": 0.8011,
+      "step": 2264
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985845392315827e-05,
+      "loss": 0.7488,
+      "step": 2265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985825718233914e-05,
+      "loss": 0.7374,
+      "step": 2266
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985806030586173e-05,
+      "loss": 0.6981,
+      "step": 2267
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985786329372875e-05,
+      "loss": 0.7159,
+      "step": 2268
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857666145942908e-05,
+      "loss": 0.6854,
+      "step": 2269
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857468862506914e-05,
+      "loss": 0.6928,
+      "step": 2270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985727144342349e-05,
+      "loss": 0.7745,
+      "step": 2271
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985707388869535e-05,
+      "loss": 0.7124,
+      "step": 2272
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856876198325213e-05,
+      "loss": 0.7132,
+      "step": 2273
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856678372315797e-05,
+      "loss": 0.7107,
+      "step": 2274
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856480410669827e-05,
+      "loss": 0.7278,
+      "step": 2275
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856282313390026e-05,
+      "loss": 0.7138,
+      "step": 2276
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856084080479115e-05,
+      "loss": 0.7641,
+      "step": 2277
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985588571193983e-05,
+      "loss": 0.851,
+      "step": 2278
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98556872077749e-05,
+      "loss": 0.788,
+      "step": 2279
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985548856798705e-05,
+      "loss": 0.6978,
+      "step": 2280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985528979257902e-05,
+      "loss": 0.6812,
+      "step": 2281
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855090881553546e-05,
+      "loss": 0.7863,
+      "step": 2282
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854891834913357e-05,
+      "loss": 0.7037,
+      "step": 2283
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98546926526612e-05,
+      "loss": 0.7858,
+      "step": 2284
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985449333479981e-05,
+      "loss": 0.7976,
+      "step": 2285
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854293881331935e-05,
+      "loss": 0.7058,
+      "step": 2286
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854094292260316e-05,
+      "loss": 0.6127,
+      "step": 2287
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98538945675877e-05,
+      "loss": 0.6452,
+      "step": 2288
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985369470731684e-05,
+      "loss": 0.5881,
+      "step": 2289
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985349471145048e-05,
+      "loss": 0.7522,
+      "step": 2290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853294579991376e-05,
+      "loss": 0.706,
+      "step": 2291
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985309431294228e-05,
+      "loss": 0.7645,
+      "step": 2292
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985289391030595e-05,
+      "loss": 0.6109,
+      "step": 2293
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985269337208514e-05,
+      "loss": 0.7754,
+      "step": 2294
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852492698282613e-05,
+      "loss": 0.746,
+      "step": 2295
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985229188890113e-05,
+      "loss": 0.7509,
+      "step": 2296
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852090943943452e-05,
+      "loss": 0.678,
+      "step": 2297
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851889863412347e-05,
+      "loss": 0.7497,
+      "step": 2298
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985168864731058e-05,
+      "loss": 0.7575,
+      "step": 2299
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985148729564092e-05,
+      "loss": 0.758,
+      "step": 2300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985128580840614e-05,
+      "loss": 0.7767,
+      "step": 2301
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851084185609012e-05,
+      "loss": 0.7586,
+      "step": 2302
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985088242725231e-05,
+      "loss": 0.7192,
+      "step": 2303
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985068053333881e-05,
+      "loss": 0.774,
+      "step": 2304
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850478503871286e-05,
+      "loss": 0.7071,
+      "step": 2305
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850276338852525e-05,
+      "loss": 0.7019,
+      "step": 2306
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850074038285307e-05,
+      "loss": 0.6944,
+      "step": 2307
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849871602172417e-05,
+      "loss": 0.5848,
+      "step": 2308
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849669030516635e-05,
+      "loss": 0.7942,
+      "step": 2309
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849466323320757e-05,
+      "loss": 0.7078,
+      "step": 2310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849263480587563e-05,
+      "loss": 0.7371,
+      "step": 2311
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849060502319852e-05,
+      "loss": 0.7067,
+      "step": 2312
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848857388520414e-05,
+      "loss": 0.7177,
+      "step": 2313
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848654139192046e-05,
+      "loss": 0.6018,
+      "step": 2314
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984845075433754e-05,
+      "loss": 0.7293,
+      "step": 2315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98482472339597e-05,
+      "loss": 0.5477,
+      "step": 2316
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848043578061326e-05,
+      "loss": 0.7651,
+      "step": 2317
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847839786645214e-05,
+      "loss": 0.6063,
+      "step": 2318
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847635859714178e-05,
+      "loss": 0.8037,
+      "step": 2319
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847431797271016e-05,
+      "loss": 0.6977,
+      "step": 2320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847227599318544e-05,
+      "loss": 0.7246,
+      "step": 2321
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847023265859567e-05,
+      "loss": 0.7318,
+      "step": 2322
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846818796896895e-05,
+      "loss": 0.7142,
+      "step": 2323
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846614192433345e-05,
+      "loss": 0.788,
+      "step": 2324
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846409452471734e-05,
+      "loss": 0.703,
+      "step": 2325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846204577014874e-05,
+      "loss": 0.7211,
+      "step": 2326
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845999566065586e-05,
+      "loss": 0.7589,
+      "step": 2327
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845794419626698e-05,
+      "loss": 0.7111,
+      "step": 2328
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845589137701023e-05,
+      "loss": 0.7074,
+      "step": 2329
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845383720291392e-05,
+      "loss": 0.7568,
+      "step": 2330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845178167400633e-05,
+      "loss": 0.7656,
+      "step": 2331
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984497247903157e-05,
+      "loss": 0.8103,
+      "step": 2332
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844766655187032e-05,
+      "loss": 0.7148,
+      "step": 2333
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844560695869855e-05,
+      "loss": 0.793,
+      "step": 2334
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844354601082877e-05,
+      "loss": 0.7677,
+      "step": 2335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844148370828927e-05,
+      "loss": 0.7886,
+      "step": 2336
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843942005110844e-05,
+      "loss": 0.805,
+      "step": 2337
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843735503931468e-05,
+      "loss": 0.7726,
+      "step": 2338
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843528867293647e-05,
+      "loss": 0.7182,
+      "step": 2339
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984332209520022e-05,
+      "loss": 0.7427,
+      "step": 2340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843115187654028e-05,
+      "loss": 0.6746,
+      "step": 2341
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984290814465792e-05,
+      "loss": 0.7484,
+      "step": 2342
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984270096621475e-05,
+      "loss": 0.7014,
+      "step": 2343
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842493652327367e-05,
+      "loss": 0.7301,
+      "step": 2344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842286202998625e-05,
+      "loss": 0.7622,
+      "step": 2345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842078618231375e-05,
+      "loss": 0.6494,
+      "step": 2346
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841870898028476e-05,
+      "loss": 0.7653,
+      "step": 2347
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984166304239278e-05,
+      "loss": 0.6443,
+      "step": 2348
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841455051327157e-05,
+      "loss": 0.6694,
+      "step": 2349
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841246924834467e-05,
+      "loss": 0.6118,
+      "step": 2350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984103866291757e-05,
+      "loss": 0.6164,
+      "step": 2351
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840830265579334e-05,
+      "loss": 0.6971,
+      "step": 2352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840621732822628e-05,
+      "loss": 0.6923,
+      "step": 2353
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840413064650318e-05,
+      "loss": 0.769,
+      "step": 2354
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840204261065276e-05,
+      "loss": 0.6989,
+      "step": 2355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983999532207038e-05,
+      "loss": 0.7575,
+      "step": 2356
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98397862476685e-05,
+      "loss": 0.7264,
+      "step": 2357
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839577037862518e-05,
+      "loss": 0.5821,
+      "step": 2358
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983936769265531e-05,
+      "loss": 0.615,
+      "step": 2359
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839158212049755e-05,
+      "loss": 0.7617,
+      "step": 2360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983894859604874e-05,
+      "loss": 0.7761,
+      "step": 2361
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838738844655144e-05,
+      "loss": 0.6898,
+      "step": 2362
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983852895787186e-05,
+      "loss": 0.7832,
+      "step": 2363
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983831893570177e-05,
+      "loss": 0.7032,
+      "step": 2364
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838108778147767e-05,
+      "loss": 0.6151,
+      "step": 2365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837898485212744e-05,
+      "loss": 0.636,
+      "step": 2366
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837688056899593e-05,
+      "loss": 0.7241,
+      "step": 2367
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837477493211207e-05,
+      "loss": 0.739,
+      "step": 2368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837266794150493e-05,
+      "loss": 0.7175,
+      "step": 2369
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837055959720337e-05,
+      "loss": 0.7396,
+      "step": 2370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983684498992365e-05,
+      "loss": 0.7794,
+      "step": 2371
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836633884763338e-05,
+      "loss": 0.7744,
+      "step": 2372
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836422644242295e-05,
+      "loss": 0.7179,
+      "step": 2373
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983621126836343e-05,
+      "loss": 0.622,
+      "step": 2374
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835999757129665e-05,
+      "loss": 0.7625,
+      "step": 2375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835788110543893e-05,
+      "loss": 0.6365,
+      "step": 2376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835576328609035e-05,
+      "loss": 0.6287,
+      "step": 2377
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835364411328006e-05,
+      "loss": 0.7274,
+      "step": 2378
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835152358703723e-05,
+      "loss": 0.7045,
+      "step": 2379
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98349401707391e-05,
+      "loss": 0.7332,
+      "step": 2380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834727847437055e-05,
+      "loss": 0.7562,
+      "step": 2381
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834515388800516e-05,
+      "loss": 0.7486,
+      "step": 2382
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834302794832404e-05,
+      "loss": 0.7576,
+      "step": 2383
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834090065535644e-05,
+      "loss": 0.6695,
+      "step": 2384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833877200913166e-05,
+      "loss": 0.7623,
+      "step": 2385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833664200967892e-05,
+      "loss": 0.7146,
+      "step": 2386
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833451065702762e-05,
+      "loss": 0.8379,
+      "step": 2387
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833237795120703e-05,
+      "loss": 0.7154,
+      "step": 2388
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833024389224653e-05,
+      "loss": 0.748,
+      "step": 2389
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832810848017547e-05,
+      "loss": 0.7583,
+      "step": 2390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983259717150232e-05,
+      "loss": 0.7436,
+      "step": 2391
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832383359681923e-05,
+      "loss": 0.7349,
+      "step": 2392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832169412559284e-05,
+      "loss": 0.6658,
+      "step": 2393
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831955330137362e-05,
+      "loss": 0.6705,
+      "step": 2394
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983174111241909e-05,
+      "loss": 0.7585,
+      "step": 2395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831526759407425e-05,
+      "loss": 0.6733,
+      "step": 2396
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831312271105313e-05,
+      "loss": 0.7256,
+      "step": 2397
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831097647515704e-05,
+      "loss": 0.7347,
+      "step": 2398
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830882888641555e-05,
+      "loss": 0.6238,
+      "step": 2399
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830667994485818e-05,
+      "loss": 0.7885,
+      "step": 2400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983045296505145e-05,
+      "loss": 0.7647,
+      "step": 2401
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983023780034141e-05,
+      "loss": 0.642,
+      "step": 2402
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830022500358664e-05,
+      "loss": 0.6279,
+      "step": 2403
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982980706510617e-05,
+      "loss": 0.7109,
+      "step": 2404
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829591494586894e-05,
+      "loss": 0.6236,
+      "step": 2405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98293757888038e-05,
+      "loss": 0.6446,
+      "step": 2406
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982915994775986e-05,
+      "loss": 0.6166,
+      "step": 2407
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828943971458042e-05,
+      "loss": 0.7124,
+      "step": 2408
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982872785990132e-05,
+      "loss": 0.6735,
+      "step": 2409
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982851161309266e-05,
+      "loss": 0.8021,
+      "step": 2410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828295231035054e-05,
+      "loss": 0.7177,
+      "step": 2411
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828078713731463e-05,
+      "loss": 0.7327,
+      "step": 2412
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827862061184876e-05,
+      "loss": 0.7246,
+      "step": 2413
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982764527339827e-05,
+      "loss": 0.6402,
+      "step": 2414
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982742835037463e-05,
+      "loss": 0.614,
+      "step": 2415
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827211292116943e-05,
+      "loss": 0.7176,
+      "step": 2416
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982699409862819e-05,
+      "loss": 0.6238,
+      "step": 2417
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826776769911365e-05,
+      "loss": 0.801,
+      "step": 2418
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826559305969458e-05,
+      "loss": 0.7172,
+      "step": 2419
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982634170680546e-05,
+      "loss": 0.6467,
+      "step": 2420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826123972422364e-05,
+      "loss": 0.702,
+      "step": 2421
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982590610282317e-05,
+      "loss": 0.5782,
+      "step": 2422
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825688098010873e-05,
+      "loss": 0.7088,
+      "step": 2423
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825469957988477e-05,
+      "loss": 0.6798,
+      "step": 2424
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982525168275898e-05,
+      "loss": 0.6628,
+      "step": 2425
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825033272325386e-05,
+      "loss": 0.6508,
+      "step": 2426
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98248147266907e-05,
+      "loss": 0.6296,
+      "step": 2427
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824596045857932e-05,
+      "loss": 0.7131,
+      "step": 2428
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824377229830088e-05,
+      "loss": 0.7567,
+      "step": 2429
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982415827861018e-05,
+      "loss": 0.612,
+      "step": 2430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823939192201224e-05,
+      "loss": 0.7188,
+      "step": 2431
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823719970606235e-05,
+      "loss": 0.6989,
+      "step": 2432
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823500613828223e-05,
+      "loss": 0.6648,
+      "step": 2433
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823281121870215e-05,
+      "loss": 0.7174,
+      "step": 2434
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823061494735224e-05,
+      "loss": 0.7508,
+      "step": 2435
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822841732426277e-05,
+      "loss": 0.7549,
+      "step": 2436
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822621834946397e-05,
+      "loss": 0.8037,
+      "step": 2437
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822401802298608e-05,
+      "loss": 0.755,
+      "step": 2438
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982218163448594e-05,
+      "loss": 0.7062,
+      "step": 2439
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821961331511424e-05,
+      "loss": 0.7806,
+      "step": 2440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821740893378084e-05,
+      "loss": 0.746,
+      "step": 2441
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821520320088968e-05,
+      "loss": 0.7667,
+      "step": 2442
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821299611647096e-05,
+      "loss": 0.7398,
+      "step": 2443
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821078768055516e-05,
+      "loss": 0.6539,
+      "step": 2444
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820857789317255e-05,
+      "loss": 0.6263,
+      "step": 2445
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820636675435368e-05,
+      "loss": 0.7182,
+      "step": 2446
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982041542641289e-05,
+      "loss": 0.6186,
+      "step": 2447
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820194042252867e-05,
+      "loss": 0.6497,
+      "step": 2448
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819972522958345e-05,
+      "loss": 0.7368,
+      "step": 2449
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981975086853237e-05,
+      "loss": 0.6847,
+      "step": 2450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819529078978e-05,
+      "loss": 0.7381,
+      "step": 2451
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819307154298277e-05,
+      "loss": 0.778,
+      "step": 2452
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981908509449626e-05,
+      "loss": 0.7577,
+      "step": 2453
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818862899575008e-05,
+      "loss": 0.7322,
+      "step": 2454
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981864056953757e-05,
+      "loss": 0.7083,
+      "step": 2455
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818418104387015e-05,
+      "loss": 0.7813,
+      "step": 2456
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818195504126396e-05,
+      "loss": 0.6562,
+      "step": 2457
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981797276875878e-05,
+      "loss": 0.6918,
+      "step": 2458
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817749898287232e-05,
+      "loss": 0.7767,
+      "step": 2459
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817526892714822e-05,
+      "loss": 0.6954,
+      "step": 2460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981730375204461e-05,
+      "loss": 0.701,
+      "step": 2461
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817080476279676e-05,
+      "loss": 0.7209,
+      "step": 2462
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981685706542309e-05,
+      "loss": 0.7171,
+      "step": 2463
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816633519477923e-05,
+      "loss": 0.5875,
+      "step": 2464
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816409838447255e-05,
+      "loss": 0.6163,
+      "step": 2465
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981618602233416e-05,
+      "loss": 0.6982,
+      "step": 2466
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981596207114172e-05,
+      "loss": 0.5867,
+      "step": 2467
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815737984873016e-05,
+      "loss": 0.6319,
+      "step": 2468
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815513763531134e-05,
+      "loss": 0.7083,
+      "step": 2469
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815289407119156e-05,
+      "loss": 0.7962,
+      "step": 2470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815064915640174e-05,
+      "loss": 0.7714,
+      "step": 2471
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814840289097273e-05,
+      "loss": 0.5989,
+      "step": 2472
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981461552749355e-05,
+      "loss": 0.7133,
+      "step": 2473
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814390630832084e-05,
+      "loss": 0.6075,
+      "step": 2474
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814165599115987e-05,
+      "loss": 0.7125,
+      "step": 2475
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813940432348343e-05,
+      "loss": 0.7334,
+      "step": 2476
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981371513053226e-05,
+      "loss": 0.6435,
+      "step": 2477
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981348969367083e-05,
+      "loss": 0.7165,
+      "step": 2478
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813264121767163e-05,
+      "loss": 0.5943,
+      "step": 2479
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813038414824356e-05,
+      "loss": 0.7191,
+      "step": 2480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812812572845516e-05,
+      "loss": 0.7548,
+      "step": 2481
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812586595833755e-05,
+      "loss": 0.6259,
+      "step": 2482
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812360483792183e-05,
+      "loss": 0.6409,
+      "step": 2483
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812134236723905e-05,
+      "loss": 0.7156,
+      "step": 2484
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981190785463204e-05,
+      "loss": 0.7651,
+      "step": 2485
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811681337519702e-05,
+      "loss": 0.5856,
+      "step": 2486
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811454685390007e-05,
+      "loss": 0.7409,
+      "step": 2487
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811227898246072e-05,
+      "loss": 0.8142,
+      "step": 2488
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811000976091025e-05,
+      "loss": 0.792,
+      "step": 2489
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981077391892798e-05,
+      "loss": 0.6811,
+      "step": 2490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810546726760064e-05,
+      "loss": 0.7215,
+      "step": 2491
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810319399590406e-05,
+      "loss": 0.5953,
+      "step": 2492
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810091937422134e-05,
+      "loss": 0.6125,
+      "step": 2493
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980986434025838e-05,
+      "loss": 0.7637,
+      "step": 2494
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980963660810227e-05,
+      "loss": 0.6968,
+      "step": 2495
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980940874095694e-05,
+      "loss": 0.7522,
+      "step": 2496
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809180738825527e-05,
+      "loss": 0.7319,
+      "step": 2497
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980895260171117e-05,
+      "loss": 0.7834,
+      "step": 2498
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808724329617e-05,
+      "loss": 0.5749,
+      "step": 2499
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980849592254617e-05,
+      "loss": 0.6427,
+      "step": 2500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980826738050182e-05,
+      "loss": 0.6143,
+      "step": 2501
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808038703487088e-05,
+      "loss": 0.7473,
+      "step": 2502
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807809891505126e-05,
+      "loss": 0.6592,
+      "step": 2503
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807580944559083e-05,
+      "loss": 0.6767,
+      "step": 2504
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807351862652108e-05,
+      "loss": 0.617,
+      "step": 2505
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807122645787354e-05,
+      "loss": 0.7728,
+      "step": 2506
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806893293967974e-05,
+      "loss": 0.581,
+      "step": 2507
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806663807197127e-05,
+      "loss": 0.7475,
+      "step": 2508
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806434185477966e-05,
+      "loss": 0.7415,
+      "step": 2509
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806204428813656e-05,
+      "loss": 0.6251,
+      "step": 2510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980597453720736e-05,
+      "loss": 0.7381,
+      "step": 2511
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805744510662235e-05,
+      "loss": 0.7324,
+      "step": 2512
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980551434918145e-05,
+      "loss": 0.6167,
+      "step": 2513
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980528405276817e-05,
+      "loss": 0.7687,
+      "step": 2514
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805053621425565e-05,
+      "loss": 0.7785,
+      "step": 2515
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804823055156807e-05,
+      "loss": 0.6155,
+      "step": 2516
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804592353965072e-05,
+      "loss": 0.6764,
+      "step": 2517
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980436151785353e-05,
+      "loss": 0.7167,
+      "step": 2518
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804130546825356e-05,
+      "loss": 0.6578,
+      "step": 2519
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803899440883735e-05,
+      "loss": 0.7423,
+      "step": 2520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980366820003184e-05,
+      "loss": 0.7848,
+      "step": 2521
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803436824272856e-05,
+      "loss": 0.7471,
+      "step": 2522
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980320531360997e-05,
+      "loss": 0.7565,
+      "step": 2523
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802973668046364e-05,
+      "loss": 0.7705,
+      "step": 2524
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802741887585224e-05,
+      "loss": 0.7587,
+      "step": 2525
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802509972229743e-05,
+      "loss": 0.7589,
+      "step": 2526
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802277921983114e-05,
+      "loss": 0.725,
+      "step": 2527
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802045736848527e-05,
+      "loss": 0.686,
+      "step": 2528
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980181341682918e-05,
+      "loss": 0.7685,
+      "step": 2529
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801580961928263e-05,
+      "loss": 0.599,
+      "step": 2530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980134837214898e-05,
+      "loss": 0.7285,
+      "step": 2531
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801115647494537e-05,
+      "loss": 0.7434,
+      "step": 2532
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800882787968126e-05,
+      "loss": 0.7257,
+      "step": 2533
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800649793572956e-05,
+      "loss": 0.7008,
+      "step": 2534
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800416664312235e-05,
+      "loss": 0.699,
+      "step": 2535
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800183400189167e-05,
+      "loss": 0.776,
+      "step": 2536
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799950001206968e-05,
+      "loss": 0.736,
+      "step": 2537
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979971646736884e-05,
+      "loss": 0.7234,
+      "step": 2538
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799482798678008e-05,
+      "loss": 0.5982,
+      "step": 2539
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979924899513768e-05,
+      "loss": 0.7627,
+      "step": 2540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799015056751078e-05,
+      "loss": 0.7037,
+      "step": 2541
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798780983521417e-05,
+      "loss": 0.7163,
+      "step": 2542
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798546775451917e-05,
+      "loss": 0.7244,
+      "step": 2543
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798312432545807e-05,
+      "loss": 0.7324,
+      "step": 2544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798077954806306e-05,
+      "loss": 0.6992,
+      "step": 2545
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797843342236642e-05,
+      "loss": 0.7344,
+      "step": 2546
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797608594840048e-05,
+      "loss": 0.7804,
+      "step": 2547
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797373712619746e-05,
+      "loss": 0.7668,
+      "step": 2548
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979713869557898e-05,
+      "loss": 0.6305,
+      "step": 2549
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796903543720974e-05,
+      "loss": 0.7394,
+      "step": 2550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796668257048965e-05,
+      "loss": 0.7787,
+      "step": 2551
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979643283556619e-05,
+      "loss": 0.6192,
+      "step": 2552
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97961972792759e-05,
+      "loss": 0.6471,
+      "step": 2553
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979596158818132e-05,
+      "loss": 0.8008,
+      "step": 2554
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795725762285704e-05,
+      "loss": 0.5728,
+      "step": 2555
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795489801592293e-05,
+      "loss": 0.6971,
+      "step": 2556
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979525370610434e-05,
+      "loss": 0.7868,
+      "step": 2557
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795017475825084e-05,
+      "loss": 0.7439,
+      "step": 2558
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794781110757785e-05,
+      "loss": 0.7897,
+      "step": 2559
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794544610905686e-05,
+      "loss": 0.6968,
+      "step": 2560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794307976272052e-05,
+      "loss": 0.7632,
+      "step": 2561
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979407120686013e-05,
+      "loss": 0.7248,
+      "step": 2562
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793834302673186e-05,
+      "loss": 0.6942,
+      "step": 2563
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793597263714474e-05,
+      "loss": 0.7137,
+      "step": 2564
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979336008998726e-05,
+      "loss": 0.6225,
+      "step": 2565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793122781494802e-05,
+      "loss": 0.73,
+      "step": 2566
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792885338240375e-05,
+      "loss": 0.7424,
+      "step": 2567
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792647760227238e-05,
+      "loss": 0.7154,
+      "step": 2568
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792410047458664e-05,
+      "loss": 0.6286,
+      "step": 2569
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792172199937926e-05,
+      "loss": 0.6623,
+      "step": 2570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791934217668286e-05,
+      "loss": 0.7893,
+      "step": 2571
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791696100653035e-05,
+      "loss": 0.6532,
+      "step": 2572
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979145784889544e-05,
+      "loss": 0.7051,
+      "step": 2573
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791219462398782e-05,
+      "loss": 0.7371,
+      "step": 2574
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790980941166337e-05,
+      "loss": 0.8121,
+      "step": 2575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790742285201396e-05,
+      "loss": 0.729,
+      "step": 2576
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790503494507235e-05,
+      "loss": 0.7575,
+      "step": 2577
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790264569087147e-05,
+      "loss": 0.7387,
+      "step": 2578
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790025508944412e-05,
+      "loss": 0.749,
+      "step": 2579
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789786314082327e-05,
+      "loss": 0.7637,
+      "step": 2580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978954698450418e-05,
+      "loss": 0.6836,
+      "step": 2581
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978930752021326e-05,
+      "loss": 0.7163,
+      "step": 2582
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789067921212874e-05,
+      "loss": 0.6042,
+      "step": 2583
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788828187506307e-05,
+      "loss": 0.7595,
+      "step": 2584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788588319096864e-05,
+      "loss": 0.7115,
+      "step": 2585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788348315987843e-05,
+      "loss": 0.7468,
+      "step": 2586
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788108178182552e-05,
+      "loss": 0.5914,
+      "step": 2587
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978786790568429e-05,
+      "loss": 0.6305,
+      "step": 2588
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787627498496366e-05,
+      "loss": 0.5983,
+      "step": 2589
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787386956622085e-05,
+      "loss": 0.7441,
+      "step": 2590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978714628006476e-05,
+      "loss": 0.6248,
+      "step": 2591
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786905468827707e-05,
+      "loss": 0.8002,
+      "step": 2592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978666452291423e-05,
+      "loss": 0.6777,
+      "step": 2593
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786423442327653e-05,
+      "loss": 0.6892,
+      "step": 2594
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978618222707129e-05,
+      "loss": 0.7853,
+      "step": 2595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785940877148457e-05,
+      "loss": 0.7495,
+      "step": 2596
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785699392562485e-05,
+      "loss": 0.6986,
+      "step": 2597
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785457773316684e-05,
+      "loss": 0.7939,
+      "step": 2598
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978521601941439e-05,
+      "loss": 0.6202,
+      "step": 2599
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784974130858926e-05,
+      "loss": 0.6841,
+      "step": 2600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978473210765362e-05,
+      "loss": 0.7131,
+      "step": 2601
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784489949801798e-05,
+      "loss": 0.7967,
+      "step": 2602
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784247657306804e-05,
+      "loss": 0.7822,
+      "step": 2603
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978400523017196e-05,
+      "loss": 0.7193,
+      "step": 2604
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978376266840061e-05,
+      "loss": 0.6609,
+      "step": 2605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783519971996087e-05,
+      "loss": 0.6217,
+      "step": 2606
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783277140961734e-05,
+      "loss": 0.6061,
+      "step": 2607
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783034175300888e-05,
+      "loss": 0.786,
+      "step": 2608
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97827910750169e-05,
+      "loss": 0.7242,
+      "step": 2609
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782547840113106e-05,
+      "loss": 0.7178,
+      "step": 2610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978230447059286e-05,
+      "loss": 0.7901,
+      "step": 2611
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978206096645951e-05,
+      "loss": 0.66,
+      "step": 2612
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781817327716408e-05,
+      "loss": 0.7411,
+      "step": 2613
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781573554366903e-05,
+      "loss": 0.7786,
+      "step": 2614
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781329646414348e-05,
+      "loss": 0.8118,
+      "step": 2615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781085603862106e-05,
+      "loss": 0.6937,
+      "step": 2616
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978084142671353e-05,
+      "loss": 0.7422,
+      "step": 2617
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780597114971983e-05,
+      "loss": 0.602,
+      "step": 2618
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780352668640822e-05,
+      "loss": 0.6537,
+      "step": 2619
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978010808772342e-05,
+      "loss": 0.7131,
+      "step": 2620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779863372223132e-05,
+      "loss": 0.6859,
+      "step": 2621
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779618522143335e-05,
+      "loss": 0.7322,
+      "step": 2622
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977937353748739e-05,
+      "loss": 0.5546,
+      "step": 2623
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9779128418258678e-05,
+      "loss": 0.689,
+      "step": 2624
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977888316446056e-05,
+      "loss": 0.7268,
+      "step": 2625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977863777609642e-05,
+      "loss": 0.741,
+      "step": 2626
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778392253169635e-05,
+      "loss": 0.7349,
+      "step": 2627
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778146595683573e-05,
+      "loss": 0.7025,
+      "step": 2628
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777900803641627e-05,
+      "loss": 0.7497,
+      "step": 2629
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777654877047173e-05,
+      "loss": 0.6711,
+      "step": 2630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97774088159036e-05,
+      "loss": 0.6948,
+      "step": 2631
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977716262021429e-05,
+      "loss": 0.7145,
+      "step": 2632
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776916289982625e-05,
+      "loss": 0.7302,
+      "step": 2633
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776669825212005e-05,
+      "loss": 0.7858,
+      "step": 2634
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776423225905817e-05,
+      "loss": 0.8005,
+      "step": 2635
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977617649206746e-05,
+      "loss": 0.742,
+      "step": 2636
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775929623700318e-05,
+      "loss": 0.7121,
+      "step": 2637
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775682620807796e-05,
+      "loss": 0.6948,
+      "step": 2638
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775435483393293e-05,
+      "loss": 0.7684,
+      "step": 2639
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775188211460207e-05,
+      "loss": 0.7518,
+      "step": 2640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774940805011943e-05,
+      "loss": 0.7429,
+      "step": 2641
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774693264051903e-05,
+      "loss": 0.7426,
+      "step": 2642
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774445588583498e-05,
+      "loss": 0.6007,
+      "step": 2643
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977419777861013e-05,
+      "loss": 0.7174,
+      "step": 2644
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773949834135212e-05,
+      "loss": 0.6123,
+      "step": 2645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773701755162157e-05,
+      "loss": 0.7854,
+      "step": 2646
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977345354169438e-05,
+      "loss": 0.663,
+      "step": 2647
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773205193735293e-05,
+      "loss": 0.7864,
+      "step": 2648
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772956711288313e-05,
+      "loss": 0.691,
+      "step": 2649
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772708094356865e-05,
+      "loss": 0.7647,
+      "step": 2650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772459342944364e-05,
+      "loss": 0.7126,
+      "step": 2651
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772210457054236e-05,
+      "loss": 0.7576,
+      "step": 2652
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771961436689906e-05,
+      "loss": 0.6531,
+      "step": 2653
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97717122818548e-05,
+      "loss": 0.7231,
+      "step": 2654
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977146299255235e-05,
+      "loss": 0.596,
+      "step": 2655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977121356878598e-05,
+      "loss": 0.6302,
+      "step": 2656
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770964010559122e-05,
+      "loss": 0.7679,
+      "step": 2657
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770714317875218e-05,
+      "loss": 0.6355,
+      "step": 2658
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97704644907377e-05,
+      "loss": 0.7248,
+      "step": 2659
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770214529150004e-05,
+      "loss": 0.7207,
+      "step": 2660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769964433115573e-05,
+      "loss": 0.6425,
+      "step": 2661
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769714202637847e-05,
+      "loss": 0.6042,
+      "step": 2662
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769463837720267e-05,
+      "loss": 0.7134,
+      "step": 2663
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976921333836628e-05,
+      "loss": 0.7426,
+      "step": 2664
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768962704579335e-05,
+      "loss": 0.6915,
+      "step": 2665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768711936362878e-05,
+      "loss": 0.7673,
+      "step": 2666
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768461033720364e-05,
+      "loss": 0.703,
+      "step": 2667
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768209996655242e-05,
+      "loss": 0.6993,
+      "step": 2668
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767958825170965e-05,
+      "loss": 0.7469,
+      "step": 2669
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767707519270993e-05,
+      "loss": 0.5774,
+      "step": 2670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767456078958785e-05,
+      "loss": 0.657,
+      "step": 2671
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767204504237796e-05,
+      "loss": 0.73,
+      "step": 2672
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766952795111492e-05,
+      "loss": 0.7122,
+      "step": 2673
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766700951583334e-05,
+      "loss": 0.6439,
+      "step": 2674
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766448973656793e-05,
+      "loss": 0.7499,
+      "step": 2675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976619686133533e-05,
+      "loss": 0.5834,
+      "step": 2676
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765944614622414e-05,
+      "loss": 0.5945,
+      "step": 2677
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765692233521523e-05,
+      "loss": 0.588,
+      "step": 2678
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765439718036123e-05,
+      "loss": 0.7088,
+      "step": 2679
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765187068169696e-05,
+      "loss": 0.5899,
+      "step": 2680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764934283925712e-05,
+      "loss": 0.7719,
+      "step": 2681
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764681365307653e-05,
+      "loss": 0.7028,
+      "step": 2682
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764428312318995e-05,
+      "loss": 0.7486,
+      "step": 2683
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764175124963226e-05,
+      "loss": 0.7089,
+      "step": 2684
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763921803243827e-05,
+      "loss": 0.767,
+      "step": 2685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763668347164288e-05,
+      "loss": 0.7461,
+      "step": 2686
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976341475672809e-05,
+      "loss": 0.7217,
+      "step": 2687
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976316103193873e-05,
+      "loss": 0.6869,
+      "step": 2688
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976290717279969e-05,
+      "loss": 0.5798,
+      "step": 2689
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762653179314474e-05,
+      "loss": 0.7171,
+      "step": 2690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976239905148657e-05,
+      "loss": 0.7186,
+      "step": 2691
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762144789319475e-05,
+      "loss": 0.7266,
+      "step": 2692
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976189039281669e-05,
+      "loss": 0.61,
+      "step": 2693
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976163586198172e-05,
+      "loss": 0.5592,
+      "step": 2694
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761381196818058e-05,
+      "loss": 0.8084,
+      "step": 2695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761126397329217e-05,
+      "loss": 0.7834,
+      "step": 2696
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760871463518702e-05,
+      "loss": 0.5702,
+      "step": 2697
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976061639539001e-05,
+      "loss": 0.7758,
+      "step": 2698
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976036119294667e-05,
+      "loss": 0.7008,
+      "step": 2699
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760105856192177e-05,
+      "loss": 0.6436,
+      "step": 2700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759850385130055e-05,
+      "loss": 0.6973,
+      "step": 2701
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759594779763817e-05,
+      "loss": 0.6168,
+      "step": 2702
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759339040096977e-05,
+      "loss": 0.593,
+      "step": 2703
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759083166133057e-05,
+      "loss": 0.7233,
+      "step": 2704
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758827157875576e-05,
+      "loss": 0.6572,
+      "step": 2705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758571015328062e-05,
+      "loss": 0.6931,
+      "step": 2706
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758314738494032e-05,
+      "loss": 0.6895,
+      "step": 2707
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975805832737702e-05,
+      "loss": 0.6289,
+      "step": 2708
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975780178198055e-05,
+      "loss": 0.8042,
+      "step": 2709
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757545102308154e-05,
+      "loss": 0.6823,
+      "step": 2710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757288288363363e-05,
+      "loss": 0.7157,
+      "step": 2711
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975703134014971e-05,
+      "loss": 0.6951,
+      "step": 2712
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756774257670735e-05,
+      "loss": 0.693,
+      "step": 2713
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756517040929973e-05,
+      "loss": 0.7317,
+      "step": 2714
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756259689930965e-05,
+      "loss": 0.753,
+      "step": 2715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756002204677245e-05,
+      "loss": 0.7007,
+      "step": 2716
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975574458517237e-05,
+      "loss": 0.7259,
+      "step": 2717
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755486831419876e-05,
+      "loss": 0.7084,
+      "step": 2718
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755228943423308e-05,
+      "loss": 0.7118,
+      "step": 2719
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754970921186217e-05,
+      "loss": 0.7257,
+      "step": 2720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754712764712158e-05,
+      "loss": 0.7963,
+      "step": 2721
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754454474004682e-05,
+      "loss": 0.6156,
+      "step": 2722
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975419604906734e-05,
+      "loss": 0.6946,
+      "step": 2723
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753937489903686e-05,
+      "loss": 0.7681,
+      "step": 2724
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975367879651728e-05,
+      "loss": 0.6626,
+      "step": 2725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975341996891169e-05,
+      "loss": 0.6397,
+      "step": 2726
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753161007090468e-05,
+      "loss": 0.6154,
+      "step": 2727
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975290191105718e-05,
+      "loss": 0.6663,
+      "step": 2728
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752642680815396e-05,
+      "loss": 0.7647,
+      "step": 2729
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752383316368674e-05,
+      "loss": 0.7446,
+      "step": 2730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975212381772059e-05,
+      "loss": 0.6324,
+      "step": 2731
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751864184874713e-05,
+      "loss": 0.6762,
+      "step": 2732
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751604417834617e-05,
+      "loss": 0.7416,
+      "step": 2733
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751344516603875e-05,
+      "loss": 0.7009,
+      "step": 2734
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751084481186066e-05,
+      "loss": 0.7062,
+      "step": 2735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750824311584766e-05,
+      "loss": 0.6998,
+      "step": 2736
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750564007803554e-05,
+      "loss": 0.6456,
+      "step": 2737
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750303569846014e-05,
+      "loss": 0.7728,
+      "step": 2738
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975004299771573e-05,
+      "loss": 0.8173,
+      "step": 2739
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749782291416287e-05,
+      "loss": 0.7742,
+      "step": 2740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749521450951277e-05,
+      "loss": 0.7682,
+      "step": 2741
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974926047632428e-05,
+      "loss": 0.7175,
+      "step": 2742
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748999367538896e-05,
+      "loss": 0.7267,
+      "step": 2743
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974873812459871e-05,
+      "loss": 0.739,
+      "step": 2744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748476747507325e-05,
+      "loss": 0.7367,
+      "step": 2745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974821523626833e-05,
+      "loss": 0.7425,
+      "step": 2746
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974795359088533e-05,
+      "loss": 0.7122,
+      "step": 2747
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747691811361926e-05,
+      "loss": 0.7735,
+      "step": 2748
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747429897701712e-05,
+      "loss": 0.6914,
+      "step": 2749
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747167849908305e-05,
+      "loss": 0.7188,
+      "step": 2750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97469056679853e-05,
+      "loss": 0.7048,
+      "step": 2751
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746643351936307e-05,
+      "loss": 0.7251,
+      "step": 2752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974638090176494e-05,
+      "loss": 0.8244,
+      "step": 2753
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746118317474806e-05,
+      "loss": 0.7422,
+      "step": 2754
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974585559906952e-05,
+      "loss": 0.805,
+      "step": 2755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97455927465527e-05,
+      "loss": 0.5757,
+      "step": 2756
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974532975992796e-05,
+      "loss": 0.7345,
+      "step": 2757
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745066639198918e-05,
+      "loss": 0.7235,
+      "step": 2758
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97448033843692e-05,
+      "loss": 0.6872,
+      "step": 2759
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744539995442417e-05,
+      "loss": 0.7496,
+      "step": 2760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974427647242221e-05,
+      "loss": 0.7505,
+      "step": 2761
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744012815312192e-05,
+      "loss": 0.6158,
+      "step": 2762
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743749024115993e-05,
+      "loss": 0.7568,
+      "step": 2763
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743485098837253e-05,
+      "loss": 0.5775,
+      "step": 2764
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974322103947959e-05,
+      "loss": 0.6566,
+      "step": 2765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974295684604665e-05,
+      "loss": 0.6635,
+      "step": 2766
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742692518542062e-05,
+      "loss": 0.7009,
+      "step": 2767
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974242805696946e-05,
+      "loss": 0.7131,
+      "step": 2768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974216346133249e-05,
+      "loss": 0.7025,
+      "step": 2769
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974189873163479e-05,
+      "loss": 0.583,
+      "step": 2770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741633867880005e-05,
+      "loss": 0.8015,
+      "step": 2771
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974136887007178e-05,
+      "loss": 0.7176,
+      "step": 2772
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741103738213755e-05,
+      "loss": 0.6272,
+      "step": 2773
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740838472309586e-05,
+      "loss": 0.6869,
+      "step": 2774
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974057307236292e-05,
+      "loss": 0.7222,
+      "step": 2775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974030753837741e-05,
+      "loss": 0.7348,
+      "step": 2776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974004187035671e-05,
+      "loss": 0.773,
+      "step": 2777
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739776068304474e-05,
+      "loss": 0.7162,
+      "step": 2778
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973951013222436e-05,
+      "loss": 0.7147,
+      "step": 2779
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739244062120037e-05,
+      "loss": 0.741,
+      "step": 2780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973897785799515e-05,
+      "loss": 0.644,
+      "step": 2781
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738711519853374e-05,
+      "loss": 0.7353,
+      "step": 2782
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738445047698364e-05,
+      "loss": 0.7309,
+      "step": 2783
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97381784415338e-05,
+      "loss": 0.7344,
+      "step": 2784
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973791170136334e-05,
+      "loss": 0.6253,
+      "step": 2785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737644827190664e-05,
+      "loss": 0.7887,
+      "step": 2786
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737377819019432e-05,
+      "loss": 0.7427,
+      "step": 2787
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737110676853327e-05,
+      "loss": 0.7775,
+      "step": 2788
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736843400696025e-05,
+      "loss": 0.6208,
+      "step": 2789
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736575990551203e-05,
+      "loss": 0.6917,
+      "step": 2790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973630844642254e-05,
+      "loss": 0.6998,
+      "step": 2791
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736040768313715e-05,
+      "loss": 0.752,
+      "step": 2792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735772956228415e-05,
+      "loss": 0.7097,
+      "step": 2793
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973550501017033e-05,
+      "loss": 0.7297,
+      "step": 2794
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735236930143134e-05,
+      "loss": 0.7403,
+      "step": 2795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734968716150527e-05,
+      "loss": 0.7556,
+      "step": 2796
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97347003681962e-05,
+      "loss": 0.7534,
+      "step": 2797
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734431886283834e-05,
+      "loss": 0.7436,
+      "step": 2798
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734163270417136e-05,
+      "loss": 0.7519,
+      "step": 2799
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97338945205998e-05,
+      "loss": 0.7021,
+      "step": 2800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733625636835523e-05,
+      "loss": 0.5679,
+      "step": 2801
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733356619128002e-05,
+      "loss": 0.7893,
+      "step": 2802
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973308746748094e-05,
+      "loss": 0.7137,
+      "step": 2803
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732818181898046e-05,
+      "loss": 0.7099,
+      "step": 2804
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732548762383018e-05,
+      "loss": 0.678,
+      "step": 2805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732279208939572e-05,
+      "loss": 0.718,
+      "step": 2806
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732009521571408e-05,
+      "loss": 0.7096,
+      "step": 2807
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731739700282246e-05,
+      "loss": 0.823,
+      "step": 2808
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731469745075792e-05,
+      "loss": 0.7154,
+      "step": 2809
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973119965595576e-05,
+      "loss": 0.6849,
+      "step": 2810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730929432925875e-05,
+      "loss": 0.7172,
+      "step": 2811
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730659075989852e-05,
+      "loss": 0.5919,
+      "step": 2812
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730388585151408e-05,
+      "loss": 0.732,
+      "step": 2813
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730117960414266e-05,
+      "loss": 0.6923,
+      "step": 2814
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729847201782154e-05,
+      "loss": 0.7531,
+      "step": 2815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729576309258793e-05,
+      "loss": 0.708,
+      "step": 2816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972930528284791e-05,
+      "loss": 0.7597,
+      "step": 2817
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972903412255324e-05,
+      "loss": 0.7262,
+      "step": 2818
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728762828378513e-05,
+      "loss": 0.7062,
+      "step": 2819
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972849140032746e-05,
+      "loss": 0.7104,
+      "step": 2820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972821983840382e-05,
+      "loss": 0.6945,
+      "step": 2821
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972794814261132e-05,
+      "loss": 0.596,
+      "step": 2822
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727676312953716e-05,
+      "loss": 0.7094,
+      "step": 2823
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972740434943473e-05,
+      "loss": 0.7131,
+      "step": 2824
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727132252058113e-05,
+      "loss": 0.732,
+      "step": 2825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726860020827615e-05,
+      "loss": 0.7802,
+      "step": 2826
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972658765574697e-05,
+      "loss": 0.6264,
+      "step": 2827
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726315156819936e-05,
+      "loss": 0.799,
+      "step": 2828
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972604252405026e-05,
+      "loss": 0.8007,
+      "step": 2829
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972576975744169e-05,
+      "loss": 0.6252,
+      "step": 2830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725496856997987e-05,
+      "loss": 0.6961,
+      "step": 2831
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725223822722897e-05,
+      "loss": 0.6771,
+      "step": 2832
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724950654620182e-05,
+      "loss": 0.7076,
+      "step": 2833
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97246773526936e-05,
+      "loss": 0.7642,
+      "step": 2834
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724403916946917e-05,
+      "loss": 0.7119,
+      "step": 2835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972413034738389e-05,
+      "loss": 0.7418,
+      "step": 2836
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723856644008283e-05,
+      "loss": 0.7875,
+      "step": 2837
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723582806823866e-05,
+      "loss": 0.7539,
+      "step": 2838
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723308835834406e-05,
+      "loss": 0.6424,
+      "step": 2839
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723034731043674e-05,
+      "loss": 0.7179,
+      "step": 2840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972276049245544e-05,
+      "loss": 0.7422,
+      "step": 2841
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722486120073478e-05,
+      "loss": 0.7564,
+      "step": 2842
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722211613901564e-05,
+      "loss": 0.7122,
+      "step": 2843
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721936973943474e-05,
+      "loss": 0.632,
+      "step": 2844
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972166220020299e-05,
+      "loss": 0.7033,
+      "step": 2845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972138729268389e-05,
+      "loss": 0.7001,
+      "step": 2846
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721112251389963e-05,
+      "loss": 0.6271,
+      "step": 2847
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720837076324984e-05,
+      "loss": 0.7765,
+      "step": 2848
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972056176749275e-05,
+      "loss": 0.6969,
+      "step": 2849
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720286324897048e-05,
+      "loss": 0.5719,
+      "step": 2850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972001074854166e-05,
+      "loss": 0.7007,
+      "step": 2851
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719735038430382e-05,
+      "loss": 0.787,
+      "step": 2852
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719459194567013e-05,
+      "loss": 0.7367,
+      "step": 2853
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719183216955346e-05,
+      "loss": 0.7303,
+      "step": 2854
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971890710559917e-05,
+      "loss": 0.7159,
+      "step": 2855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718630860502303e-05,
+      "loss": 0.706,
+      "step": 2856
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971835448166853e-05,
+      "loss": 0.6971,
+      "step": 2857
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971807796910166e-05,
+      "loss": 0.5924,
+      "step": 2858
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97178013228055e-05,
+      "loss": 0.7256,
+      "step": 2859
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717524542783854e-05,
+      "loss": 0.8148,
+      "step": 2860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971724762904053e-05,
+      "loss": 0.6632,
+      "step": 2861
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716970581579343e-05,
+      "loss": 0.6675,
+      "step": 2862
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97166934004041e-05,
+      "loss": 0.6556,
+      "step": 2863
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716416085518622e-05,
+      "loss": 0.7016,
+      "step": 2864
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971613863692672e-05,
+      "loss": 0.7524,
+      "step": 2865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971586105463221e-05,
+      "loss": 0.7768,
+      "step": 2866
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971558333863892e-05,
+      "loss": 0.6166,
+      "step": 2867
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971530548895066e-05,
+      "loss": 0.7054,
+      "step": 2868
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715027505571266e-05,
+      "loss": 0.6287,
+      "step": 2869
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971474938850455e-05,
+      "loss": 0.6887,
+      "step": 2870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971447113775435e-05,
+      "loss": 0.7181,
+      "step": 2871
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714192753324496e-05,
+      "loss": 0.7071,
+      "step": 2872
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713914235218806e-05,
+      "loss": 0.613,
+      "step": 2873
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713635583441125e-05,
+      "loss": 0.7319,
+      "step": 2874
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713356797995282e-05,
+      "loss": 0.78,
+      "step": 2875
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713077878885116e-05,
+      "loss": 0.6663,
+      "step": 2876
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712798826114462e-05,
+      "loss": 0.8008,
+      "step": 2877
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712519639687166e-05,
+      "loss": 0.7652,
+      "step": 2878
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712240319607063e-05,
+      "loss": 0.7722,
+      "step": 2879
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711960865877998e-05,
+      "loss": 0.6541,
+      "step": 2880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711681278503817e-05,
+      "loss": 0.6431,
+      "step": 2881
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711401557488366e-05,
+      "loss": 0.6417,
+      "step": 2882
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711121702835504e-05,
+      "loss": 0.7684,
+      "step": 2883
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971084171454907e-05,
+      "loss": 0.7323,
+      "step": 2884
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971056159263292e-05,
+      "loss": 0.6934,
+      "step": 2885
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971028133709091e-05,
+      "loss": 0.755,
+      "step": 2886
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97100009479269e-05,
+      "loss": 0.715,
+      "step": 2887
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709720425144745e-05,
+      "loss": 0.6848,
+      "step": 2888
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97094397687483e-05,
+      "loss": 0.7699,
+      "step": 2889
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709158978741437e-05,
+      "loss": 0.7287,
+      "step": 2890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708878055128015e-05,
+      "loss": 0.7223,
+      "step": 2891
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97085969979119e-05,
+      "loss": 0.6646,
+      "step": 2892
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970831580709696e-05,
+      "loss": 0.7671,
+      "step": 2893
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708034482687063e-05,
+      "loss": 0.7335,
+      "step": 2894
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707753024686077e-05,
+      "loss": 0.6052,
+      "step": 2895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707471433097888e-05,
+      "loss": 0.6837,
+      "step": 2896
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707189707926356e-05,
+      "loss": 0.7678,
+      "step": 2897
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9706907849175366e-05,
+      "loss": 0.6678,
+      "step": 2898
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9706625856848792e-05,
+      "loss": 0.7016,
+      "step": 2899
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9706343730950523e-05,
+      "loss": 0.7625,
+      "step": 2900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970606147148443e-05,
+      "loss": 0.6264,
+      "step": 2901
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705779078454407e-05,
+      "loss": 0.5981,
+      "step": 2902
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705496551864332e-05,
+      "loss": 0.7167,
+      "step": 2903
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.97052138917181e-05,
+      "loss": 0.6993,
+      "step": 2904
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704931098019594e-05,
+      "loss": 0.5813,
+      "step": 2905
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970464817077271e-05,
+      "loss": 0.7516,
+      "step": 2906
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970436510998134e-05,
+      "loss": 0.7399,
+      "step": 2907
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970408191564938e-05,
+      "loss": 0.7028,
+      "step": 2908
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703798587780723e-05,
+      "loss": 0.6258,
+      "step": 2909
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703515126379274e-05,
+      "loss": 0.784,
+      "step": 2910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703231531448928e-05,
+      "loss": 0.6087,
+      "step": 2911
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702947802993596e-05,
+      "loss": 0.7523,
+      "step": 2912
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702663941017167e-05,
+      "loss": 0.6932,
+      "step": 2913
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702379945523565e-05,
+      "loss": 0.7609,
+      "step": 2914
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702095816516687e-05,
+      "loss": 0.7044,
+      "step": 2915
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701811554000444e-05,
+      "loss": 0.7493,
+      "step": 2916
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970152715797875e-05,
+      "loss": 0.6165,
+      "step": 2917
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701242628455518e-05,
+      "loss": 0.6574,
+      "step": 2918
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700957965434664e-05,
+      "loss": 0.8186,
+      "step": 2919
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700673168920103e-05,
+      "loss": 0.6118,
+      "step": 2920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970038823891576e-05,
+      "loss": 0.7509,
+      "step": 2921
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700103175425544e-05,
+      "loss": 0.7553,
+      "step": 2922
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969981797845339e-05,
+      "loss": 0.6091,
+      "step": 2923
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699532648003214e-05,
+      "loss": 0.7898,
+      "step": 2924
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969924718407895e-05,
+      "loss": 0.739,
+      "step": 2925
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969896158668452e-05,
+      "loss": 0.7085,
+      "step": 2926
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698675855823858e-05,
+      "loss": 0.6841,
+      "step": 2927
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969838999150089e-05,
+      "loss": 0.7552,
+      "step": 2928
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698103993719554e-05,
+      "loss": 0.739,
+      "step": 2929
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697817862483793e-05,
+      "loss": 0.7067,
+      "step": 2930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969753159779753e-05,
+      "loss": 0.6067,
+      "step": 2931
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697245199664715e-05,
+      "loss": 0.714,
+      "step": 2932
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969695866808928e-05,
+      "loss": 0.7083,
+      "step": 2933
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969667200307518e-05,
+      "loss": 0.6384,
+      "step": 2934
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969638520462635e-05,
+      "loss": 0.6948,
+      "step": 2935
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969609827274674e-05,
+      "loss": 0.6588,
+      "step": 2936
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96958112074403e-05,
+      "loss": 0.7543,
+      "step": 2937
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695524008710972e-05,
+      "loss": 0.7392,
+      "step": 2938
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695236676562717e-05,
+      "loss": 0.6158,
+      "step": 2939
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969494921099949e-05,
+      "loss": 0.6122,
+      "step": 2940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969466161202524e-05,
+      "loss": 0.7654,
+      "step": 2941
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694373879643928e-05,
+      "loss": 0.8,
+      "step": 2942
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694086013859516e-05,
+      "loss": 0.763,
+      "step": 2943
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969379801467596e-05,
+      "loss": 0.6181,
+      "step": 2944
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693509882097226e-05,
+      "loss": 0.7629,
+      "step": 2945
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693221616127277e-05,
+      "loss": 0.6771,
+      "step": 2946
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692933216770083e-05,
+      "loss": 0.7487,
+      "step": 2947
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969264468402961e-05,
+      "loss": 0.7457,
+      "step": 2948
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969235601790983e-05,
+      "loss": 0.6926,
+      "step": 2949
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692067218414717e-05,
+      "loss": 0.7405,
+      "step": 2950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969177828554824e-05,
+      "loss": 0.7336,
+      "step": 2951
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969148921931438e-05,
+      "loss": 0.7637,
+      "step": 2952
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691200019717113e-05,
+      "loss": 0.7323,
+      "step": 2953
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690910686760414e-05,
+      "loss": 0.6226,
+      "step": 2954
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969062122044827e-05,
+      "loss": 0.6574,
+      "step": 2955
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690331620784665e-05,
+      "loss": 0.8114,
+      "step": 2956
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690041887773584e-05,
+      "loss": 0.7607,
+      "step": 2957
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689752021419008e-05,
+      "loss": 0.6257,
+      "step": 2958
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689462021724933e-05,
+      "loss": 0.569,
+      "step": 2959
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689171888695345e-05,
+      "loss": 0.7162,
+      "step": 2960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968888162233424e-05,
+      "loss": 0.6013,
+      "step": 2961
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688591222645607e-05,
+      "loss": 0.6985,
+      "step": 2962
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688300689633452e-05,
+      "loss": 0.583,
+      "step": 2963
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968801002330176e-05,
+      "loss": 0.6079,
+      "step": 2964
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968771922365454e-05,
+      "loss": 0.663,
+      "step": 2965
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687428290695792e-05,
+      "loss": 0.7656,
+      "step": 2966
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687137224429515e-05,
+      "loss": 0.7024,
+      "step": 2967
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968684602485972e-05,
+      "loss": 0.7181,
+      "step": 2968
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686554691990417e-05,
+      "loss": 0.6929,
+      "step": 2969
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686263225825604e-05,
+      "loss": 0.8166,
+      "step": 2970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96859716263693e-05,
+      "loss": 0.6817,
+      "step": 2971
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685679893625512e-05,
+      "loss": 0.7483,
+      "step": 2972
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685388027598265e-05,
+      "loss": 0.7118,
+      "step": 2973
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685096028291562e-05,
+      "loss": 0.6941,
+      "step": 2974
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684803895709434e-05,
+      "loss": 0.6234,
+      "step": 2975
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968451162985589e-05,
+      "loss": 0.6046,
+      "step": 2976
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684219230734955e-05,
+      "loss": 0.7312,
+      "step": 2977
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683926698350656e-05,
+      "loss": 0.7254,
+      "step": 2978
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968363403270702e-05,
+      "loss": 0.6853,
+      "step": 2979
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683341233808067e-05,
+      "loss": 0.72,
+      "step": 2980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968304830165783e-05,
+      "loss": 0.6115,
+      "step": 2981
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968275523626034e-05,
+      "loss": 0.719,
+      "step": 2982
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682462037619633e-05,
+      "loss": 0.7877,
+      "step": 2983
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968216870573974e-05,
+      "loss": 0.7483,
+      "step": 2984
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681875240624696e-05,
+      "loss": 0.6692,
+      "step": 2985
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968158164227854e-05,
+      "loss": 0.7268,
+      "step": 2986
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681287910705317e-05,
+      "loss": 0.7431,
+      "step": 2987
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680994045909064e-05,
+      "loss": 0.6587,
+      "step": 2988
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680700047893827e-05,
+      "loss": 0.6646,
+      "step": 2989
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680405916663652e-05,
+      "loss": 0.6545,
+      "step": 2990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680111652222584e-05,
+      "loss": 0.7604,
+      "step": 2991
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679817254574676e-05,
+      "loss": 0.6774,
+      "step": 2992
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679522723723975e-05,
+      "loss": 0.7138,
+      "step": 2993
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679228059674534e-05,
+      "loss": 0.6103,
+      "step": 2994
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678933262430417e-05,
+      "loss": 0.6103,
+      "step": 2995
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678638331995667e-05,
+      "loss": 0.7326,
+      "step": 2996
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678343268374355e-05,
+      "loss": 0.6865,
+      "step": 2997
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967804807157053e-05,
+      "loss": 0.624,
+      "step": 2998
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967775274158826e-05,
+      "loss": 0.7292,
+      "step": 2999
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677457278431614e-05,
+      "loss": 0.7398,
+      "step": 3000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677161682104647e-05,
+      "loss": 0.6438,
+      "step": 3001
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676865952611436e-05,
+      "loss": 0.6946,
+      "step": 3002
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676570089956043e-05,
+      "loss": 0.6947,
+      "step": 3003
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676274094142544e-05,
+      "loss": 0.6156,
+      "step": 3004
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675977965175015e-05,
+      "loss": 0.6725,
+      "step": 3005
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675681703057523e-05,
+      "loss": 0.7925,
+      "step": 3006
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675385307794152e-05,
+      "loss": 0.7236,
+      "step": 3007
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675088779388972e-05,
+      "loss": 0.6825,
+      "step": 3008
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674792117846073e-05,
+      "loss": 0.6336,
+      "step": 3009
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674495323169537e-05,
+      "loss": 0.6269,
+      "step": 3010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674198395363436e-05,
+      "loss": 0.6633,
+      "step": 3011
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967390133443187e-05,
+      "loss": 0.714,
+      "step": 3012
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673604140378925e-05,
+      "loss": 0.64,
+      "step": 3013
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673306813208682e-05,
+      "loss": 0.752,
+      "step": 3014
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967300935292524e-05,
+      "loss": 0.6739,
+      "step": 3015
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672711759532684e-05,
+      "loss": 0.7407,
+      "step": 3016
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967241403303512e-05,
+      "loss": 0.757,
+      "step": 3017
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967211617343664e-05,
+      "loss": 0.7222,
+      "step": 3018
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671818180741343e-05,
+      "loss": 0.7044,
+      "step": 3019
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671520054953327e-05,
+      "loss": 0.648,
+      "step": 3020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671221796076702e-05,
+      "loss": 0.695,
+      "step": 3021
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967092340411556e-05,
+      "loss": 0.7578,
+      "step": 3022
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670624879074018e-05,
+      "loss": 0.7379,
+      "step": 3023
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967032622095618e-05,
+      "loss": 0.7698,
+      "step": 3024
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670027429766157e-05,
+      "loss": 0.6647,
+      "step": 3025
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669728505508063e-05,
+      "loss": 0.7856,
+      "step": 3026
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669429448186e-05,
+      "loss": 0.6798,
+      "step": 3027
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669130257804096e-05,
+      "loss": 0.7143,
+      "step": 3028
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966883093436646e-05,
+      "loss": 0.6992,
+      "step": 3029
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668531477877223e-05,
+      "loss": 0.656,
+      "step": 3030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966823188834049e-05,
+      "loss": 0.7182,
+      "step": 3031
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667932165760397e-05,
+      "loss": 0.7019,
+      "step": 3032
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667632310141055e-05,
+      "loss": 0.7298,
+      "step": 3033
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667332321486603e-05,
+      "loss": 0.7361,
+      "step": 3034
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966703219980116e-05,
+      "loss": 0.7261,
+      "step": 3035
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666731945088863e-05,
+      "loss": 0.7588,
+      "step": 3036
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966643155735384e-05,
+      "loss": 0.7303,
+      "step": 3037
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666131036600225e-05,
+      "loss": 0.6873,
+      "step": 3038
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665830382832152e-05,
+      "loss": 0.7795,
+      "step": 3039
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665529596053764e-05,
+      "loss": 0.7411,
+      "step": 3040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966522867626919e-05,
+      "loss": 0.7242,
+      "step": 3041
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966492762348258e-05,
+      "loss": 0.7458,
+      "step": 3042
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664626437698074e-05,
+      "loss": 0.7147,
+      "step": 3043
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664325118919813e-05,
+      "loss": 0.6641,
+      "step": 3044
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966402366715195e-05,
+      "loss": 0.7328,
+      "step": 3045
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966372208239863e-05,
+      "loss": 0.7283,
+      "step": 3046
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663420364663998e-05,
+      "loss": 0.8292,
+      "step": 3047
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663118513952217e-05,
+      "loss": 0.6394,
+      "step": 3048
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966281653026743e-05,
+      "loss": 0.7136,
+      "step": 3049
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662514413613796e-05,
+      "loss": 0.7095,
+      "step": 3050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662212163995478e-05,
+      "loss": 0.7068,
+      "step": 3051
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966190978141663e-05,
+      "loss": 0.6757,
+      "step": 3052
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661607265881414e-05,
+      "loss": 0.6804,
+      "step": 3053
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966130461739399e-05,
+      "loss": 0.7278,
+      "step": 3054
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661001835958526e-05,
+      "loss": 0.7018,
+      "step": 3055
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966069892157919e-05,
+      "loss": 0.7268,
+      "step": 3056
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660395874260146e-05,
+      "loss": 0.7112,
+      "step": 3057
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660092694005566e-05,
+      "loss": 0.615,
+      "step": 3058
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659789380819624e-05,
+      "loss": 0.6841,
+      "step": 3059
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965948593470649e-05,
+      "loss": 0.7089,
+      "step": 3060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659182355670345e-05,
+      "loss": 0.6595,
+      "step": 3061
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658878643715362e-05,
+      "loss": 0.7501,
+      "step": 3062
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965857479884572e-05,
+      "loss": 0.6922,
+      "step": 3063
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658270821065603e-05,
+      "loss": 0.7129,
+      "step": 3064
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657966710379195e-05,
+      "loss": 0.7506,
+      "step": 3065
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657662466790678e-05,
+      "loss": 0.8001,
+      "step": 3066
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965735809030424e-05,
+      "loss": 0.6819,
+      "step": 3067
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965705358092407e-05,
+      "loss": 0.729,
+      "step": 3068
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9656748938654352e-05,
+      "loss": 0.7501,
+      "step": 3069
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9656444163499288e-05,
+      "loss": 0.7403,
+      "step": 3070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9656139255463066e-05,
+      "loss": 0.7235,
+      "step": 3071
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655834214549883e-05,
+      "loss": 0.7596,
+      "step": 3072
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655529040763936e-05,
+      "loss": 0.6237,
+      "step": 3073
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965522373410943e-05,
+      "loss": 0.6783,
+      "step": 3074
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654918294590554e-05,
+      "loss": 0.7333,
+      "step": 3075
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654612722211525e-05,
+      "loss": 0.6688,
+      "step": 3076
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965430701697654e-05,
+      "loss": 0.6817,
+      "step": 3077
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965400117888981e-05,
+      "loss": 0.715,
+      "step": 3078
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653695207955534e-05,
+      "loss": 0.7678,
+      "step": 3079
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653389104177935e-05,
+      "loss": 0.7488,
+      "step": 3080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965308286756122e-05,
+      "loss": 0.7616,
+      "step": 3081
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9652776498109597e-05,
+      "loss": 0.7739,
+      "step": 3082
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9652469995827292e-05,
+      "loss": 0.6193,
+      "step": 3083
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965216336071852e-05,
+      "loss": 0.6637,
+      "step": 3084
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651856592787497e-05,
+      "loss": 0.7666,
+      "step": 3085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651549692038444e-05,
+      "loss": 0.8126,
+      "step": 3086
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965124265847559e-05,
+      "loss": 0.758,
+      "step": 3087
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650935492103158e-05,
+      "loss": 0.7475,
+      "step": 3088
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650628192925372e-05,
+      "loss": 0.6982,
+      "step": 3089
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650320760946463e-05,
+      "loss": 0.6535,
+      "step": 3090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650013196170658e-05,
+      "loss": 0.7286,
+      "step": 3091
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649705498602193e-05,
+      "loss": 0.6579,
+      "step": 3092
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649397668245306e-05,
+      "loss": 0.7064,
+      "step": 3093
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649089705104223e-05,
+      "loss": 0.5847,
+      "step": 3094
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648781609183187e-05,
+      "loss": 0.7414,
+      "step": 3095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964847338048644e-05,
+      "loss": 0.7447,
+      "step": 3096
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964816501901822e-05,
+      "loss": 0.7511,
+      "step": 3097
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647856524782774e-05,
+      "loss": 0.6359,
+      "step": 3098
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647547897784346e-05,
+      "loss": 0.6042,
+      "step": 3099
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964723913802718e-05,
+      "loss": 0.6629,
+      "step": 3100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9646930245515524e-05,
+      "loss": 0.6451,
+      "step": 3101
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9646621220253635e-05,
+      "loss": 0.7129,
+      "step": 3102
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964631206224576e-05,
+      "loss": 0.6063,
+      "step": 3103
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964600277149615e-05,
+      "loss": 0.7659,
+      "step": 3104
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9645693348009075e-05,
+      "loss": 0.6705,
+      "step": 3105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9645383791788778e-05,
+      "loss": 0.6118,
+      "step": 3106
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964507410283953e-05,
+      "loss": 0.745,
+      "step": 3107
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644764281165584e-05,
+      "loss": 0.7111,
+      "step": 3108
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964445432677121e-05,
+      "loss": 0.7949,
+      "step": 3109
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644144239660665e-05,
+      "loss": 0.7112,
+      "step": 3110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643834019838226e-05,
+      "loss": 0.7016,
+      "step": 3111
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643523667308157e-05,
+      "loss": 0.7593,
+      "step": 3112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643213182074724e-05,
+      "loss": 0.5537,
+      "step": 3113
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964290256414221e-05,
+      "loss": 0.7665,
+      "step": 3114
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642591813514877e-05,
+      "loss": 0.6095,
+      "step": 3115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642280930197012e-05,
+      "loss": 0.6139,
+      "step": 3116
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964196991419289e-05,
+      "loss": 0.7532,
+      "step": 3117
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964165876550679e-05,
+      "loss": 0.753,
+      "step": 3118
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964134748414299e-05,
+      "loss": 0.6171,
+      "step": 3119
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964103607010578e-05,
+      "loss": 0.7316,
+      "step": 3120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964072452339944e-05,
+      "loss": 0.7005,
+      "step": 3121
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964041284402826e-05,
+      "loss": 0.6219,
+      "step": 3122
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964010103199653e-05,
+      "loss": 0.7415,
+      "step": 3123
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963978908730854e-05,
+      "loss": 0.6303,
+      "step": 3124
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963947700996858e-05,
+      "loss": 0.7015,
+      "step": 3125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639164799980946e-05,
+      "loss": 0.707,
+      "step": 3126
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638852457349936e-05,
+      "loss": 0.7436,
+      "step": 3127
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638539982079844e-05,
+      "loss": 0.6016,
+      "step": 3128
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638227374174977e-05,
+      "loss": 0.6669,
+      "step": 3129
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963791463363963e-05,
+      "loss": 0.783,
+      "step": 3130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963760176047811e-05,
+      "loss": 0.7047,
+      "step": 3131
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637288754694717e-05,
+      "loss": 0.7223,
+      "step": 3132
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636975616293766e-05,
+      "loss": 0.7737,
+      "step": 3133
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636662345279568e-05,
+      "loss": 0.6527,
+      "step": 3134
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636348941656422e-05,
+      "loss": 0.7621,
+      "step": 3135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963603540542865e-05,
+      "loss": 0.7389,
+      "step": 3136
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635721736600562e-05,
+      "loss": 0.7589,
+      "step": 3137
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635407935176477e-05,
+      "loss": 0.6004,
+      "step": 3138
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635094001160714e-05,
+      "loss": 0.6183,
+      "step": 3139
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963477993455759e-05,
+      "loss": 0.7414,
+      "step": 3140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634465735371427e-05,
+      "loss": 0.7476,
+      "step": 3141
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634151403606553e-05,
+      "loss": 0.6379,
+      "step": 3142
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633836939267287e-05,
+      "loss": 0.7044,
+      "step": 3143
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633522342357962e-05,
+      "loss": 0.6937,
+      "step": 3144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633207612882903e-05,
+      "loss": 0.5771,
+      "step": 3145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632892750846445e-05,
+      "loss": 0.6551,
+      "step": 3146
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963257775625292e-05,
+      "loss": 0.6174,
+      "step": 3147
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632262629106655e-05,
+      "loss": 0.6789,
+      "step": 3148
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631947369411993e-05,
+      "loss": 0.7756,
+      "step": 3149
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631631977173276e-05,
+      "loss": 0.6308,
+      "step": 3150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963131645239484e-05,
+      "loss": 0.722,
+      "step": 3151
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631000795081024e-05,
+      "loss": 0.7017,
+      "step": 3152
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630685005236175e-05,
+      "loss": 0.612,
+      "step": 3153
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630369082864643e-05,
+      "loss": 0.6385,
+      "step": 3154
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630053027970764e-05,
+      "loss": 0.7646,
+      "step": 3155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629736840558896e-05,
+      "loss": 0.7519,
+      "step": 3156
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629420520633392e-05,
+      "loss": 0.7396,
+      "step": 3157
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629104068198593e-05,
+      "loss": 0.7182,
+      "step": 3158
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628787483258868e-05,
+      "loss": 0.6083,
+      "step": 3159
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628470765818562e-05,
+      "loss": 0.7052,
+      "step": 3160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962815391588204e-05,
+      "loss": 0.726,
+      "step": 3161
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9627836933453656e-05,
+      "loss": 0.8092,
+      "step": 3162
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962751981853778e-05,
+      "loss": 0.5826,
+      "step": 3163
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962720257113877e-05,
+      "loss": 0.7618,
+      "step": 3164
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626885191260997e-05,
+      "loss": 0.7173,
+      "step": 3165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626567678908822e-05,
+      "loss": 0.7062,
+      "step": 3166
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626250034086617e-05,
+      "loss": 0.7514,
+      "step": 3167
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625932256798755e-05,
+      "loss": 0.6223,
+      "step": 3168
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96256143470496e-05,
+      "loss": 0.7493,
+      "step": 3169
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962529630484354e-05,
+      "loss": 0.7276,
+      "step": 3170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624978130184948e-05,
+      "loss": 0.6826,
+      "step": 3171
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624659823078195e-05,
+      "loss": 0.6676,
+      "step": 3172
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962434138352767e-05,
+      "loss": 0.7385,
+      "step": 3173
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624022811537748e-05,
+      "loss": 0.7168,
+      "step": 3174
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9623704107112816e-05,
+      "loss": 0.7213,
+      "step": 3175
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9623385270257258e-05,
+      "loss": 0.6993,
+      "step": 3176
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962306630097546e-05,
+      "loss": 0.7075,
+      "step": 3177
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622747199271822e-05,
+      "loss": 0.6488,
+      "step": 3178
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622427965150724e-05,
+      "loss": 0.7512,
+      "step": 3179
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962210859861656e-05,
+      "loss": 0.7407,
+      "step": 3180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962178909967373e-05,
+      "loss": 0.7213,
+      "step": 3181
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621469468326624e-05,
+      "loss": 0.7291,
+      "step": 3182
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621149704579646e-05,
+      "loss": 0.6009,
+      "step": 3183
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620829808437193e-05,
+      "loss": 0.6096,
+      "step": 3184
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620509779903673e-05,
+      "loss": 0.7394,
+      "step": 3185
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962018961898348e-05,
+      "loss": 0.7277,
+      "step": 3186
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619869325681028e-05,
+      "loss": 0.7545,
+      "step": 3187
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961954890000072e-05,
+      "loss": 0.7121,
+      "step": 3188
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619228341946964e-05,
+      "loss": 0.632,
+      "step": 3189
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618907651524176e-05,
+      "loss": 0.6312,
+      "step": 3190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961858682873677e-05,
+      "loss": 0.7984,
+      "step": 3191
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618265873589154e-05,
+      "loss": 0.7645,
+      "step": 3192
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617944786085746e-05,
+      "loss": 0.6472,
+      "step": 3193
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961762356623097e-05,
+      "loss": 0.7336,
+      "step": 3194
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617302214029242e-05,
+      "loss": 0.7151,
+      "step": 3195
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961698072948499e-05,
+      "loss": 0.7383,
+      "step": 3196
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616659112602625e-05,
+      "loss": 0.7053,
+      "step": 3197
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616337363386586e-05,
+      "loss": 0.7139,
+      "step": 3198
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616015481841293e-05,
+      "loss": 0.7074,
+      "step": 3199
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961569346797118e-05,
+      "loss": 0.744,
+      "step": 3200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961537132178067e-05,
+      "loss": 0.6984,
+      "step": 3201
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615049043274207e-05,
+      "loss": 0.6156,
+      "step": 3202
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9614726632456217e-05,
+      "loss": 0.6461,
+      "step": 3203
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961440408933114e-05,
+      "loss": 0.7138,
+      "step": 3204
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961408141390342e-05,
+      "loss": 0.6714,
+      "step": 3205
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613758606177492e-05,
+      "loss": 0.762,
+      "step": 3206
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613435666157793e-05,
+      "loss": 0.7301,
+      "step": 3207
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613112593848775e-05,
+      "loss": 0.627,
+      "step": 3208
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961278938925488e-05,
+      "loss": 0.689,
+      "step": 3209
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9612466052380557e-05,
+      "loss": 0.6808,
+      "step": 3210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9612142583230255e-05,
+      "loss": 0.6805,
+      "step": 3211
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611818981808425e-05,
+      "loss": 0.767,
+      "step": 3212
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961149524811952e-05,
+      "loss": 0.6761,
+      "step": 3213
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611171382167998e-05,
+      "loss": 0.7292,
+      "step": 3214
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9610847383958308e-05,
+      "loss": 0.6864,
+      "step": 3215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9610523253494915e-05,
+      "loss": 0.7043,
+      "step": 3216
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9610198990782277e-05,
+      "loss": 0.6534,
+      "step": 3217
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609874595824858e-05,
+      "loss": 0.6137,
+      "step": 3218
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960955006862712e-05,
+      "loss": 0.7558,
+      "step": 3219
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609225409193526e-05,
+      "loss": 0.6535,
+      "step": 3220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960890061752855e-05,
+      "loss": 0.6548,
+      "step": 3221
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960857569363666e-05,
+      "loss": 0.7509,
+      "step": 3222
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608250637522322e-05,
+      "loss": 0.7233,
+      "step": 3223
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607925449190016e-05,
+      "loss": 0.7676,
+      "step": 3224
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960760012864421e-05,
+      "loss": 0.6885,
+      "step": 3225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607274675889388e-05,
+      "loss": 0.7269,
+      "step": 3226
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606949090930025e-05,
+      "loss": 0.7909,
+      "step": 3227
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96066233737706e-05,
+      "loss": 0.7423,
+      "step": 3228
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606297524415595e-05,
+      "loss": 0.6033,
+      "step": 3229
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605971542869497e-05,
+      "loss": 0.6641,
+      "step": 3230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960564542913679e-05,
+      "loss": 0.7906,
+      "step": 3231
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605319183221956e-05,
+      "loss": 0.6587,
+      "step": 3232
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604992805129493e-05,
+      "loss": 0.6709,
+      "step": 3233
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960466629486389e-05,
+      "loss": 0.7482,
+      "step": 3234
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604339652429646e-05,
+      "loss": 0.742,
+      "step": 3235
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960401287783124e-05,
+      "loss": 0.7598,
+      "step": 3236
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603685971073183e-05,
+      "loss": 0.7404,
+      "step": 3237
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603358932159966e-05,
+      "loss": 0.8338,
+      "step": 3238
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603031761096095e-05,
+      "loss": 0.7049,
+      "step": 3239
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602704457886068e-05,
+      "loss": 0.6338,
+      "step": 3240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602377022534384e-05,
+      "loss": 0.6704,
+      "step": 3241
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960204945504556e-05,
+      "loss": 0.5998,
+      "step": 3242
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601721755424098e-05,
+      "loss": 0.7731,
+      "step": 3243
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960139392367451e-05,
+      "loss": 0.5861,
+      "step": 3244
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601065959801302e-05,
+      "loss": 0.6428,
+      "step": 3245
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600737863808992e-05,
+      "loss": 0.6108,
+      "step": 3246
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960040963570209e-05,
+      "loss": 0.6154,
+      "step": 3247
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960008127548512e-05,
+      "loss": 0.5861,
+      "step": 3248
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599752783162594e-05,
+      "loss": 0.6576,
+      "step": 3249
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599424158739033e-05,
+      "loss": 0.6075,
+      "step": 3250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599095402218965e-05,
+      "loss": 0.6465,
+      "step": 3251
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598766513606904e-05,
+      "loss": 0.7486,
+      "step": 3252
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598437492907388e-05,
+      "loss": 0.7436,
+      "step": 3253
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598108340124935e-05,
+      "loss": 0.7317,
+      "step": 3254
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959777905526408e-05,
+      "loss": 0.7023,
+      "step": 3255
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597449638329346e-05,
+      "loss": 0.6854,
+      "step": 3256
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597120089325273e-05,
+      "loss": 0.71,
+      "step": 3257
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95967904082564e-05,
+      "loss": 0.7966,
+      "step": 3258
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9596460595127255e-05,
+      "loss": 0.7505,
+      "step": 3259
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9596130649942383e-05,
+      "loss": 0.6912,
+      "step": 3260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595800572706318e-05,
+      "loss": 0.6895,
+      "step": 3261
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595470363423606e-05,
+      "loss": 0.7326,
+      "step": 3262
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595140022098792e-05,
+      "loss": 0.7023,
+      "step": 3263
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959480954873642e-05,
+      "loss": 0.7249,
+      "step": 3264
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594478943341034e-05,
+      "loss": 0.7656,
+      "step": 3265
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594148205917193e-05,
+      "loss": 0.734,
+      "step": 3266
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593817336469437e-05,
+      "loss": 0.704,
+      "step": 3267
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959348633500233e-05,
+      "loss": 0.7334,
+      "step": 3268
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593155201520418e-05,
+      "loss": 0.7525,
+      "step": 3269
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592823936028262e-05,
+      "loss": 0.768,
+      "step": 3270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959249253853042e-05,
+      "loss": 0.6058,
+      "step": 3271
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592161009031453e-05,
+      "loss": 0.7122,
+      "step": 3272
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959182934753592e-05,
+      "loss": 0.7082,
+      "step": 3273
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959149755404839e-05,
+      "loss": 0.7451,
+      "step": 3274
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959116562857342e-05,
+      "loss": 0.7001,
+      "step": 3275
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959083357111559e-05,
+      "loss": 0.7667,
+      "step": 3276
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959050138167946e-05,
+      "loss": 0.6375,
+      "step": 3277
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590169060269602e-05,
+      "loss": 0.6265,
+      "step": 3278
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9589836606890594e-05,
+      "loss": 0.7489,
+      "step": 3279
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9589504021547007e-05,
+      "loss": 0.745,
+      "step": 3280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958917130424342e-05,
+      "loss": 0.6634,
+      "step": 3281
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958883845498441e-05,
+      "loss": 0.6997,
+      "step": 3282
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588505473774554e-05,
+      "loss": 0.6235,
+      "step": 3283
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958817236061844e-05,
+      "loss": 0.7438,
+      "step": 3284
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958783911552065e-05,
+      "loss": 0.6184,
+      "step": 3285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587505738485765e-05,
+      "loss": 0.6183,
+      "step": 3286
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958717222951838e-05,
+      "loss": 0.7712,
+      "step": 3287
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586838588623084e-05,
+      "loss": 0.6014,
+      "step": 3288
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586504815804463e-05,
+      "loss": 0.6947,
+      "step": 3289
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958617091106711e-05,
+      "loss": 0.7252,
+      "step": 3290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585836874415625e-05,
+      "loss": 0.8021,
+      "step": 3291
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585502705854598e-05,
+      "loss": 0.7508,
+      "step": 3292
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585168405388636e-05,
+      "loss": 0.7361,
+      "step": 3293
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958483397302233e-05,
+      "loss": 0.6562,
+      "step": 3294
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584499408760288e-05,
+      "loss": 0.6086,
+      "step": 3295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584164712607114e-05,
+      "loss": 0.6749,
+      "step": 3296
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583829884567413e-05,
+      "loss": 0.6339,
+      "step": 3297
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583494924645787e-05,
+      "loss": 0.7824,
+      "step": 3298
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583159832846853e-05,
+      "loss": 0.8094,
+      "step": 3299
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958282460917522e-05,
+      "loss": 0.6739,
+      "step": 3300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9582489253635503e-05,
+      "loss": 0.7521,
+      "step": 3301
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9582153766232304e-05,
+      "loss": 0.7262,
+      "step": 3302
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958181814697026e-05,
+      "loss": 0.7247,
+      "step": 3303
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581482395853973e-05,
+      "loss": 0.7227,
+      "step": 3304
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581146512888072e-05,
+      "loss": 0.7605,
+      "step": 3305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580810498077177e-05,
+      "loss": 0.6883,
+      "step": 3306
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580474351425905e-05,
+      "loss": 0.6918,
+      "step": 3307
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580138072938895e-05,
+      "loss": 0.7409,
+      "step": 3308
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579801662620764e-05,
+      "loss": 0.7873,
+      "step": 3309
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579465120476144e-05,
+      "loss": 0.696,
+      "step": 3310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957912844650967e-05,
+      "loss": 0.7658,
+      "step": 3311
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957879164072597e-05,
+      "loss": 0.6602,
+      "step": 3312
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957845470312968e-05,
+      "loss": 0.7337,
+      "step": 3313
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578117633725434e-05,
+      "loss": 0.7768,
+      "step": 3314
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957778043251788e-05,
+      "loss": 0.6917,
+      "step": 3315
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577443099511644e-05,
+      "loss": 0.6919,
+      "step": 3316
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957710563471138e-05,
+      "loss": 0.6263,
+      "step": 3317
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957676803812173e-05,
+      "loss": 0.7084,
+      "step": 3318
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576430309747328e-05,
+      "loss": 0.7386,
+      "step": 3319
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576092449592835e-05,
+      "loss": 0.748,
+      "step": 3320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575754457662897e-05,
+      "loss": 0.7308,
+      "step": 3321
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575416333962163e-05,
+      "loss": 0.692,
+      "step": 3322
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575078078495287e-05,
+      "loss": 0.6754,
+      "step": 3323
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574739691266923e-05,
+      "loss": 0.6749,
+      "step": 3324
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574401172281727e-05,
+      "loss": 0.7482,
+      "step": 3325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574062521544358e-05,
+      "loss": 0.74,
+      "step": 3326
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573723739059477e-05,
+      "loss": 0.7462,
+      "step": 3327
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573384824831744e-05,
+      "loss": 0.7162,
+      "step": 3328
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573045778865823e-05,
+      "loss": 0.6572,
+      "step": 3329
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9572706601166384e-05,
+      "loss": 0.7023,
+      "step": 3330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957236729173809e-05,
+      "loss": 0.727,
+      "step": 3331
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9572027850585606e-05,
+      "loss": 0.716,
+      "step": 3332
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571688277713614e-05,
+      "loss": 0.7463,
+      "step": 3333
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957134857312678e-05,
+      "loss": 0.7674,
+      "step": 3334
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571008736829778e-05,
+      "loss": 0.5687,
+      "step": 3335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570668768827284e-05,
+      "loss": 0.7117,
+      "step": 3336
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957032866912398e-05,
+      "loss": 0.682,
+      "step": 3337
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569988437724548e-05,
+      "loss": 0.725,
+      "step": 3338
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569648074633663e-05,
+      "loss": 0.7249,
+      "step": 3339
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956930757985601e-05,
+      "loss": 0.6495,
+      "step": 3340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956896695339628e-05,
+      "loss": 0.6267,
+      "step": 3341
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568626195259152e-05,
+      "loss": 0.7718,
+      "step": 3342
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568285305449323e-05,
+      "loss": 0.7549,
+      "step": 3343
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956794428397148e-05,
+      "loss": 0.6966,
+      "step": 3344
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9567603130830317e-05,
+      "loss": 0.7507,
+      "step": 3345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956726184603053e-05,
+      "loss": 0.726,
+      "step": 3346
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956692042957681e-05,
+      "loss": 0.6658,
+      "step": 3347
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566578881473862e-05,
+      "loss": 0.7055,
+      "step": 3348
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566237201726382e-05,
+      "loss": 0.6354,
+      "step": 3349
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956589539033907e-05,
+      "loss": 0.7182,
+      "step": 3350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956555344731663e-05,
+      "loss": 0.6843,
+      "step": 3351
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565211372663774e-05,
+      "loss": 0.6942,
+      "step": 3352
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564869166385204e-05,
+      "loss": 0.6843,
+      "step": 3353
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564526828485633e-05,
+      "loss": 0.5889,
+      "step": 3354
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564184358969762e-05,
+      "loss": 0.7033,
+      "step": 3355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9563841757842315e-05,
+      "loss": 0.7011,
+      "step": 3356
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9563499025108e-05,
+      "loss": 0.7801,
+      "step": 3357
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9563156160771538e-05,
+      "loss": 0.7205,
+      "step": 3358
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956281316483764e-05,
+      "loss": 0.568,
+      "step": 3359
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562470037311035e-05,
+      "loss": 0.7262,
+      "step": 3360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562126778196434e-05,
+      "loss": 0.731,
+      "step": 3361
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561783387498573e-05,
+      "loss": 0.6234,
+      "step": 3362
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561439865222165e-05,
+      "loss": 0.7087,
+      "step": 3363
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561096211371946e-05,
+      "loss": 0.7241,
+      "step": 3364
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956075242595264e-05,
+      "loss": 0.7545,
+      "step": 3365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956040850896898e-05,
+      "loss": 0.6384,
+      "step": 3366
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560064460425705e-05,
+      "loss": 0.6965,
+      "step": 3367
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559720280327536e-05,
+      "loss": 0.7352,
+      "step": 3368
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559375968679217e-05,
+      "loss": 0.6741,
+      "step": 3369
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559031525485486e-05,
+      "loss": 0.7234,
+      "step": 3370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558686950751083e-05,
+      "loss": 0.7104,
+      "step": 3371
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955834224448075e-05,
+      "loss": 0.6002,
+      "step": 3372
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557997406679224e-05,
+      "loss": 0.7544,
+      "step": 3373
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557652437351255e-05,
+      "loss": 0.7779,
+      "step": 3374
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557307336501595e-05,
+      "loss": 0.6169,
+      "step": 3375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556962104134985e-05,
+      "loss": 0.6354,
+      "step": 3376
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556616740256183e-05,
+      "loss": 0.7096,
+      "step": 3377
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556271244869932e-05,
+      "loss": 0.6798,
+      "step": 3378
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555925617980997e-05,
+      "loss": 0.6171,
+      "step": 3379
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555579859594126e-05,
+      "loss": 0.7316,
+      "step": 3380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555233969714078e-05,
+      "loss": 0.6976,
+      "step": 3381
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955488794834562e-05,
+      "loss": 0.6698,
+      "step": 3382
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554541795493502e-05,
+      "loss": 0.7381,
+      "step": 3383
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554195511162496e-05,
+      "loss": 0.7503,
+      "step": 3384
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553849095357366e-05,
+      "loss": 0.6402,
+      "step": 3385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553502548082878e-05,
+      "loss": 0.6856,
+      "step": 3386
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95531558693438e-05,
+      "loss": 0.7402,
+      "step": 3387
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95528090591449e-05,
+      "loss": 0.6136,
+      "step": 3388
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552462117490956e-05,
+      "loss": 0.7275,
+      "step": 3389
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552115044386742e-05,
+      "loss": 0.7389,
+      "step": 3390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9551767839837027e-05,
+      "loss": 0.7101,
+      "step": 3391
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95514205038466e-05,
+      "loss": 0.633,
+      "step": 3392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955107303642023e-05,
+      "loss": 0.7197,
+      "step": 3393
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550725437562703e-05,
+      "loss": 0.6762,
+      "step": 3394
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95503777072788e-05,
+      "loss": 0.7222,
+      "step": 3395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550029845573313e-05,
+      "loss": 0.7678,
+      "step": 3396
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549681852451022e-05,
+      "loss": 0.6328,
+      "step": 3397
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549333727916716e-05,
+      "loss": 0.7463,
+      "step": 3398
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548985471975186e-05,
+      "loss": 0.7208,
+      "step": 3399
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548637084631228e-05,
+      "loss": 0.7402,
+      "step": 3400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954828856588963e-05,
+      "loss": 0.6608,
+      "step": 3401
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547939915755196e-05,
+      "loss": 0.6162,
+      "step": 3402
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954759113423272e-05,
+      "loss": 0.7888,
+      "step": 3403
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547242221326995e-05,
+      "loss": 0.7129,
+      "step": 3404
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546893177042834e-05,
+      "loss": 0.7262,
+      "step": 3405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954654400138503e-05,
+      "loss": 0.7333,
+      "step": 3406
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546194694358392e-05,
+      "loss": 0.6297,
+      "step": 3407
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954584525596773e-05,
+      "loss": 0.7371,
+      "step": 3408
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954549568621785e-05,
+      "loss": 0.6021,
+      "step": 3409
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954514598511356e-05,
+      "loss": 0.775,
+      "step": 3410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544796152659676e-05,
+      "loss": 0.6411,
+      "step": 3411
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954444618886101e-05,
+      "loss": 0.7415,
+      "step": 3412
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544096093722373e-05,
+      "loss": 0.7421,
+      "step": 3413
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954374586724859e-05,
+      "loss": 0.6228,
+      "step": 3414
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954339550944448e-05,
+      "loss": 0.7201,
+      "step": 3415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9543045020314864e-05,
+      "loss": 0.6484,
+      "step": 3416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9542694399864558e-05,
+      "loss": 0.6864,
+      "step": 3417
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9542343648098395e-05,
+      "loss": 0.7471,
+      "step": 3418
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95419927650212e-05,
+      "loss": 0.7099,
+      "step": 3419
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541641750637797e-05,
+      "loss": 0.7414,
+      "step": 3420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541290604953023e-05,
+      "loss": 0.7373,
+      "step": 3421
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540939327971707e-05,
+      "loss": 0.7683,
+      "step": 3422
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954058791969868e-05,
+      "loss": 0.7041,
+      "step": 3423
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540236380138776e-05,
+      "loss": 0.7031,
+      "step": 3424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539884709296844e-05,
+      "loss": 0.6288,
+      "step": 3425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539532907177713e-05,
+      "loss": 0.705,
+      "step": 3426
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539180973786223e-05,
+      "loss": 0.6873,
+      "step": 3427
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953882890912723e-05,
+      "loss": 0.7575,
+      "step": 3428
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538476713205564e-05,
+      "loss": 0.7307,
+      "step": 3429
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538124386026078e-05,
+      "loss": 0.7062,
+      "step": 3430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953777192759362e-05,
+      "loss": 0.6277,
+      "step": 3431
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9537419337913036e-05,
+      "loss": 0.6064,
+      "step": 3432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9537066616989183e-05,
+      "loss": 0.6279,
+      "step": 3433
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536713764826917e-05,
+      "loss": 0.7416,
+      "step": 3434
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953636078143109e-05,
+      "loss": 0.7088,
+      "step": 3435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536007666806555e-05,
+      "loss": 0.739,
+      "step": 3436
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9535654420958177e-05,
+      "loss": 0.7189,
+      "step": 3437
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9535301043890817e-05,
+      "loss": 0.638,
+      "step": 3438
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534947535609334e-05,
+      "loss": 0.6667,
+      "step": 3439
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534593896118596e-05,
+      "loss": 0.7003,
+      "step": 3440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534240125423468e-05,
+      "loss": 0.6741,
+      "step": 3441
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533886223528818e-05,
+      "loss": 0.7247,
+      "step": 3442
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533532190439518e-05,
+      "loss": 0.7141,
+      "step": 3443
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533178026160438e-05,
+      "loss": 0.7451,
+      "step": 3444
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532823730696448e-05,
+      "loss": 0.5939,
+      "step": 3445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532469304052432e-05,
+      "loss": 0.7083,
+      "step": 3446
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953211474623326e-05,
+      "loss": 0.7152,
+      "step": 3447
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953176005724381e-05,
+      "loss": 0.8104,
+      "step": 3448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953140523708897e-05,
+      "loss": 0.7407,
+      "step": 3449
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9531050285773622e-05,
+      "loss": 0.7199,
+      "step": 3450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9530695203302645e-05,
+      "loss": 0.6823,
+      "step": 3451
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9530339989680927e-05,
+      "loss": 0.7115,
+      "step": 3452
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529984644913356e-05,
+      "loss": 0.712,
+      "step": 3453
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952962916900482e-05,
+      "loss": 0.6792,
+      "step": 3454
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952927356196022e-05,
+      "loss": 0.7133,
+      "step": 3455
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528917823784436e-05,
+      "loss": 0.7028,
+      "step": 3456
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528561954482375e-05,
+      "loss": 0.6804,
+      "step": 3457
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952820595405893e-05,
+      "loss": 0.7163,
+      "step": 3458
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952784982251899e-05,
+      "loss": 0.6931,
+      "step": 3459
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9527493559867473e-05,
+      "loss": 0.745,
+      "step": 3460
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952713716610927e-05,
+      "loss": 0.686,
+      "step": 3461
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526780641249287e-05,
+      "loss": 0.6675,
+      "step": 3462
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526423985292434e-05,
+      "loss": 0.7338,
+      "step": 3463
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526067198243615e-05,
+      "loss": 0.6223,
+      "step": 3464
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952571028010774e-05,
+      "loss": 0.7231,
+      "step": 3465
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9525353230889722e-05,
+      "loss": 0.7603,
+      "step": 3466
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524996050594477e-05,
+      "loss": 0.7328,
+      "step": 3467
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524638739226914e-05,
+      "loss": 0.6783,
+      "step": 3468
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524281296791952e-05,
+      "loss": 0.7342,
+      "step": 3469
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523923723294513e-05,
+      "loss": 0.6244,
+      "step": 3470
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523566018739514e-05,
+      "loss": 0.7416,
+      "step": 3471
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952320818313188e-05,
+      "loss": 0.6322,
+      "step": 3472
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952285021647653e-05,
+      "loss": 0.7466,
+      "step": 3473
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522492118778396e-05,
+      "loss": 0.7159,
+      "step": 3474
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95221338900424e-05,
+      "loss": 0.7606,
+      "step": 3475
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952177553027348e-05,
+      "loss": 0.7806,
+      "step": 3476
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952141703947656e-05,
+      "loss": 0.7415,
+      "step": 3477
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9521058417656575e-05,
+      "loss": 0.7149,
+      "step": 3478
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520699664818462e-05,
+      "loss": 0.6941,
+      "step": 3479
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520340780967152e-05,
+      "loss": 0.7919,
+      "step": 3480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519981766107593e-05,
+      "loss": 0.7029,
+      "step": 3481
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519622620244716e-05,
+      "loss": 0.5837,
+      "step": 3482
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519263343383467e-05,
+      "loss": 0.6104,
+      "step": 3483
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951890393552879e-05,
+      "loss": 0.5928,
+      "step": 3484
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951854439668563e-05,
+      "loss": 0.6192,
+      "step": 3485
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951818472685894e-05,
+      "loss": 0.7293,
+      "step": 3486
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951782492605366e-05,
+      "loss": 0.5978,
+      "step": 3487
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951746499427475e-05,
+      "loss": 0.7682,
+      "step": 3488
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517104931527154e-05,
+      "loss": 0.7098,
+      "step": 3489
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516744737815837e-05,
+      "loss": 0.6407,
+      "step": 3490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516384413145746e-05,
+      "loss": 0.7772,
+      "step": 3491
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516023957521846e-05,
+      "loss": 0.7319,
+      "step": 3492
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9515663370949093e-05,
+      "loss": 0.7898,
+      "step": 3493
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9515302653432452e-05,
+      "loss": 0.7448,
+      "step": 3494
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514941804976885e-05,
+      "loss": 0.7627,
+      "step": 3495
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951458082558736e-05,
+      "loss": 0.7079,
+      "step": 3496
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951421971526884e-05,
+      "loss": 0.7327,
+      "step": 3497
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95138584740263e-05,
+      "loss": 0.7522,
+      "step": 3498
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513497101864705e-05,
+      "loss": 0.7029,
+      "step": 3499
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951313559878903e-05,
+      "loss": 0.6723,
+      "step": 3500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512773964804252e-05,
+      "loss": 0.6522,
+      "step": 3501
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512412199915343e-05,
+      "loss": 0.6943,
+      "step": 3502
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512050304127287e-05,
+      "loss": 0.761,
+      "step": 3503
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951168827744506e-05,
+      "loss": 0.7094,
+      "step": 3504
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951132611987365e-05,
+      "loss": 0.6762,
+      "step": 3505
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510963831418026e-05,
+      "loss": 0.6285,
+      "step": 3506
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510601412083184e-05,
+      "loss": 0.7592,
+      "step": 3507
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510238861874112e-05,
+      "loss": 0.7365,
+      "step": 3508
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95098761807958e-05,
+      "loss": 0.6635,
+      "step": 3509
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509513368853232e-05,
+      "loss": 0.7389,
+      "step": 3510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509150426051405e-05,
+      "loss": 0.5786,
+      "step": 3511
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950878735239531e-05,
+      "loss": 0.7245,
+      "step": 3512
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950842414788995e-05,
+      "loss": 0.7184,
+      "step": 3513
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508060812540317e-05,
+      "loss": 0.6208,
+      "step": 3514
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507697346351414e-05,
+      "loss": 0.6731,
+      "step": 3515
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950733374932824e-05,
+      "loss": 0.5887,
+      "step": 3516
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9506970021475798e-05,
+      "loss": 0.5912,
+      "step": 3517
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95066061627991e-05,
+      "loss": 0.6773,
+      "step": 3518
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9506242173303142e-05,
+      "loss": 0.7126,
+      "step": 3519
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505878052992945e-05,
+      "loss": 0.728,
+      "step": 3520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950551380187351e-05,
+      "loss": 0.6621,
+      "step": 3521
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505149419949853e-05,
+      "loss": 0.7477,
+      "step": 3522
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504784907226984e-05,
+      "loss": 0.5605,
+      "step": 3523
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950442026370993e-05,
+      "loss": 0.6188,
+      "step": 3524
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504055489403696e-05,
+      "loss": 0.7059,
+      "step": 3525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9503690584313312e-05,
+      "loss": 0.763,
+      "step": 3526
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9503325548443793e-05,
+      "loss": 0.6812,
+      "step": 3527
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502960381800165e-05,
+      "loss": 0.7758,
+      "step": 3528
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950259508438745e-05,
+      "loss": 0.6865,
+      "step": 3529
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950222965621068e-05,
+      "loss": 0.6097,
+      "step": 3530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950186409727488e-05,
+      "loss": 0.6829,
+      "step": 3531
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950149840758508e-05,
+      "loss": 0.7132,
+      "step": 3532
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950113258714631e-05,
+      "loss": 0.7336,
+      "step": 3533
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500766635963614e-05,
+      "loss": 0.6182,
+      "step": 3534
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500400554042015e-05,
+      "loss": 0.7979,
+      "step": 3535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950003434138656e-05,
+      "loss": 0.7426,
+      "step": 3536
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499667998002284e-05,
+      "loss": 0.711,
+      "step": 3537
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499301523894226e-05,
+      "loss": 0.7221,
+      "step": 3538
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498934919067435e-05,
+      "loss": 0.7053,
+      "step": 3539
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498568183526956e-05,
+      "loss": 0.7189,
+      "step": 3540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949820131727783e-05,
+      "loss": 0.7234,
+      "step": 3541
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497834320325107e-05,
+      "loss": 0.7319,
+      "step": 3542
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497467192673836e-05,
+      "loss": 0.6486,
+      "step": 3543
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497099934329075e-05,
+      "loss": 0.7488,
+      "step": 3544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496732545295874e-05,
+      "loss": 0.7308,
+      "step": 3545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949636502557929e-05,
+      "loss": 0.7866,
+      "step": 3546
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495997375184377e-05,
+      "loss": 0.688,
+      "step": 3547
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495629594116193e-05,
+      "loss": 0.7333,
+      "step": 3548
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495261682379808e-05,
+      "loss": 0.7051,
+      "step": 3549
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494893639980276e-05,
+      "loss": 0.6321,
+      "step": 3550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494525466922663e-05,
+      "loss": 0.72,
+      "step": 3551
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949415716321204e-05,
+      "loss": 0.7324,
+      "step": 3552
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949378872885347e-05,
+      "loss": 0.6417,
+      "step": 3553
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949342016385203e-05,
+      "loss": 0.7439,
+      "step": 3554
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949305146821278e-05,
+      "loss": 0.7185,
+      "step": 3555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492682641940803e-05,
+      "loss": 0.6769,
+      "step": 3556
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492313685041174e-05,
+      "loss": 0.7094,
+      "step": 3557
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491944597518965e-05,
+      "loss": 0.6872,
+      "step": 3558
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949157537937926e-05,
+      "loss": 0.7231,
+      "step": 3559
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491206030627136e-05,
+      "loss": 0.6541,
+      "step": 3560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949083655126768e-05,
+      "loss": 0.7347,
+      "step": 3561
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490466941305972e-05,
+      "loss": 0.6303,
+      "step": 3562
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94900972007471e-05,
+      "loss": 0.7276,
+      "step": 3563
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489727329596147e-05,
+      "loss": 0.7025,
+      "step": 3564
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489357327858216e-05,
+      "loss": 0.6049,
+      "step": 3565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488987195538386e-05,
+      "loss": 0.6441,
+      "step": 3566
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488616932641756e-05,
+      "loss": 0.7051,
+      "step": 3567
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488246539173415e-05,
+      "loss": 0.7111,
+      "step": 3568
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948787601513847e-05,
+      "loss": 0.6931,
+      "step": 3569
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487505360542013e-05,
+      "loss": 0.7028,
+      "step": 3570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948713457538914e-05,
+      "loss": 0.7862,
+      "step": 3571
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948676365968497e-05,
+      "loss": 0.7261,
+      "step": 3572
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9486392613434584e-05,
+      "loss": 0.6938,
+      "step": 3573
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948602143664311e-05,
+      "loss": 0.6519,
+      "step": 3574
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485650129315645e-05,
+      "loss": 0.6784,
+      "step": 3575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485278691457292e-05,
+      "loss": 0.6546,
+      "step": 3576
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484907123073175e-05,
+      "loss": 0.7519,
+      "step": 3577
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484535424168403e-05,
+      "loss": 0.7038,
+      "step": 3578
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484163594748088e-05,
+      "loss": 0.6901,
+      "step": 3579
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483791634817347e-05,
+      "loss": 0.737,
+      "step": 3580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94834195443813e-05,
+      "loss": 0.7572,
+      "step": 3581
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948304732344507e-05,
+      "loss": 0.6921,
+      "step": 3582
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9482674972013775e-05,
+      "loss": 0.6059,
+      "step": 3583
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948230249009254e-05,
+      "loss": 0.7663,
+      "step": 3584
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948192987768649e-05,
+      "loss": 0.6531,
+      "step": 3585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481557134800756e-05,
+      "loss": 0.5992,
+      "step": 3586
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481184261440462e-05,
+      "loss": 0.7629,
+      "step": 3587
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480811257610744e-05,
+      "loss": 0.743,
+      "step": 3588
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948043812331673e-05,
+      "loss": 0.6402,
+      "step": 3589
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480064858563558e-05,
+      "loss": 0.6557,
+      "step": 3590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947969146335636e-05,
+      "loss": 0.6729,
+      "step": 3591
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479317937700284e-05,
+      "loss": 0.5997,
+      "step": 3592
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947894428160046e-05,
+      "loss": 0.7627,
+      "step": 3593
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478570495062038e-05,
+      "loss": 0.757,
+      "step": 3594
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478196578090152e-05,
+      "loss": 0.5824,
+      "step": 3595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947782253068996e-05,
+      "loss": 0.5427,
+      "step": 3596
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477448352866596e-05,
+      "loss": 0.7067,
+      "step": 3597
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477074044625215e-05,
+      "loss": 0.6954,
+      "step": 3598
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947669960597097e-05,
+      "loss": 0.7786,
+      "step": 3599
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947632503690901e-05,
+      "loss": 0.7826,
+      "step": 3600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947595033744449e-05,
+      "loss": 0.6843,
+      "step": 3601
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947557550758257e-05,
+      "loss": 0.5907,
+      "step": 3602
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475200547328407e-05,
+      "loss": 0.7413,
+      "step": 3603
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947482545668715e-05,
+      "loss": 0.6457,
+      "step": 3604
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474450235663978e-05,
+      "loss": 0.6415,
+      "step": 3605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474074884264043e-05,
+      "loss": 0.6194,
+      "step": 3606
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473699402492513e-05,
+      "loss": 0.6246,
+      "step": 3607
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473323790354556e-05,
+      "loss": 0.6878,
+      "step": 3608
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472948047855338e-05,
+      "loss": 0.7371,
+      "step": 3609
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947257217500003e-05,
+      "loss": 0.6782,
+      "step": 3610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472196171793808e-05,
+      "loss": 0.7502,
+      "step": 3611
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471820038241844e-05,
+      "loss": 0.7238,
+      "step": 3612
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947144377434931e-05,
+      "loss": 0.661,
+      "step": 3613
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947106738012139e-05,
+      "loss": 0.7215,
+      "step": 3614
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470690855563263e-05,
+      "loss": 0.6657,
+      "step": 3615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470314200680104e-05,
+      "loss": 0.601,
+      "step": 3616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94699374154771e-05,
+      "loss": 0.7363,
+      "step": 3617
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946956049995944e-05,
+      "loss": 0.6446,
+      "step": 3618
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9469183454132303e-05,
+      "loss": 0.5699,
+      "step": 3619
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468806278000884e-05,
+      "loss": 0.6062,
+      "step": 3620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946842897157037e-05,
+      "loss": 0.7333,
+      "step": 3621
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468051534845954e-05,
+      "loss": 0.5838,
+      "step": 3622
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9467673967832828e-05,
+      "loss": 0.6866,
+      "step": 3623
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9467296270536192e-05,
+      "loss": 0.7296,
+      "step": 3624
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466918442961237e-05,
+      "loss": 0.6427,
+      "step": 3625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466540485113167e-05,
+      "loss": 0.7741,
+      "step": 3626
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466162396997183e-05,
+      "loss": 0.6639,
+      "step": 3627
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9465784178618485e-05,
+      "loss": 0.7077,
+      "step": 3628
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946540582998228e-05,
+      "loss": 0.6359,
+      "step": 3629
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9465027351093776e-05,
+      "loss": 0.6808,
+      "step": 3630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946464874195818e-05,
+      "loss": 0.7229,
+      "step": 3631
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9464270002580697e-05,
+      "loss": 0.7723,
+      "step": 3632
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463891132966545e-05,
+      "loss": 0.6301,
+      "step": 3633
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463512133120935e-05,
+      "loss": 0.6783,
+      "step": 3634
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946313300304908e-05,
+      "loss": 0.7531,
+      "step": 3635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94627537427562e-05,
+      "loss": 0.7504,
+      "step": 3636
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946237435224752e-05,
+      "loss": 0.6786,
+      "step": 3637
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461994831528244e-05,
+      "loss": 0.7098,
+      "step": 3638
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461615180603614e-05,
+      "loss": 0.7014,
+      "step": 3639
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461235399478842e-05,
+      "loss": 0.7654,
+      "step": 3640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946085548815916e-05,
+      "loss": 0.7773,
+      "step": 3641
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946047544664979e-05,
+      "loss": 0.6755,
+      "step": 3642
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9460095274955968e-05,
+      "loss": 0.6706,
+      "step": 3643
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459714973082923e-05,
+      "loss": 0.6379,
+      "step": 3644
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459334541035883e-05,
+      "loss": 0.72,
+      "step": 3645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458953978820092e-05,
+      "loss": 0.6838,
+      "step": 3646
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458573286440783e-05,
+      "loss": 0.7351,
+      "step": 3647
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458192463903195e-05,
+      "loss": 0.7545,
+      "step": 3648
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457811511212565e-05,
+      "loss": 0.7691,
+      "step": 3649
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945743042837414e-05,
+      "loss": 0.6696,
+      "step": 3650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945704921539316e-05,
+      "loss": 0.6609,
+      "step": 3651
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945666787227488e-05,
+      "loss": 0.5737,
+      "step": 3652
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9456286399024537e-05,
+      "loss": 0.7251,
+      "step": 3653
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945590479564738e-05,
+      "loss": 0.6717,
+      "step": 3654
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455523062148667e-05,
+      "loss": 0.7123,
+      "step": 3655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945514119853365e-05,
+      "loss": 0.7236,
+      "step": 3656
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945475920480758e-05,
+      "loss": 0.6428,
+      "step": 3657
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9454377080975718e-05,
+      "loss": 0.719,
+      "step": 3658
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945399482704332e-05,
+      "loss": 0.7527,
+      "step": 3659
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453612443015642e-05,
+      "loss": 0.7006,
+      "step": 3660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453229928897955e-05,
+      "loss": 0.7727,
+      "step": 3661
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452847284695515e-05,
+      "loss": 0.656,
+      "step": 3662
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452464510413593e-05,
+      "loss": 0.7178,
+      "step": 3663
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945208160605745e-05,
+      "loss": 0.6977,
+      "step": 3664
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945169857163236e-05,
+      "loss": 0.6775,
+      "step": 3665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9451315407143593e-05,
+      "loss": 0.7527,
+      "step": 3666
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450932112596423e-05,
+      "loss": 0.6853,
+      "step": 3667
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450548687996122e-05,
+      "loss": 0.7376,
+      "step": 3668
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450165133347963e-05,
+      "loss": 0.6152,
+      "step": 3669
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449781448657234e-05,
+      "loss": 0.7172,
+      "step": 3670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449397633929208e-05,
+      "loss": 0.6776,
+      "step": 3671
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449013689169166e-05,
+      "loss": 0.7401,
+      "step": 3672
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9448629614382394e-05,
+      "loss": 0.6659,
+      "step": 3673
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9448245409574173e-05,
+      "loss": 0.7379,
+      "step": 3674
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447861074749798e-05,
+      "loss": 0.6621,
+      "step": 3675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944747660991455e-05,
+      "loss": 0.6333,
+      "step": 3676
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447092015073724e-05,
+      "loss": 0.6863,
+      "step": 3677
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446707290232608e-05,
+      "loss": 0.7487,
+      "step": 3678
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446322435396504e-05,
+      "loss": 0.6464,
+      "step": 3679
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9445937450570697e-05,
+      "loss": 0.637,
+      "step": 3680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9445552335760497e-05,
+      "loss": 0.6881,
+      "step": 3681
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944516709097119e-05,
+      "loss": 0.701,
+      "step": 3682
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944478171620809e-05,
+      "loss": 0.6909,
+      "step": 3683
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9444396211476495e-05,
+      "loss": 0.7491,
+      "step": 3684
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9444010576781708e-05,
+      "loss": 0.7064,
+      "step": 3685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443624812129037e-05,
+      "loss": 0.7461,
+      "step": 3686
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944323891752379e-05,
+      "loss": 0.6871,
+      "step": 3687
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944285289297128e-05,
+      "loss": 0.7308,
+      "step": 3688
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9442466738476814e-05,
+      "loss": 0.7931,
+      "step": 3689
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944208045404571e-05,
+      "loss": 0.628,
+      "step": 3690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9441694039683284e-05,
+      "loss": 0.6009,
+      "step": 3691
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944130749539485e-05,
+      "loss": 0.7285,
+      "step": 3692
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440920821185734e-05,
+      "loss": 0.7595,
+      "step": 3693
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440534017061247e-05,
+      "loss": 0.7618,
+      "step": 3694
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440147083026717e-05,
+      "loss": 0.7364,
+      "step": 3695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439760019087468e-05,
+      "loss": 0.7474,
+      "step": 3696
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439372825248828e-05,
+      "loss": 0.7338,
+      "step": 3697
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438985501516123e-05,
+      "loss": 0.6827,
+      "step": 3698
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438598047894684e-05,
+      "loss": 0.7756,
+      "step": 3699
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438210464389842e-05,
+      "loss": 0.7137,
+      "step": 3700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437822751006932e-05,
+      "loss": 0.7069,
+      "step": 3701
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437434907751285e-05,
+      "loss": 0.7839,
+      "step": 3702
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437046934628246e-05,
+      "loss": 0.7284,
+      "step": 3703
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9436658831643146e-05,
+      "loss": 0.6465,
+      "step": 3704
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943627059880133e-05,
+      "loss": 0.7498,
+      "step": 3705
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943588223610814e-05,
+      "loss": 0.7563,
+      "step": 3706
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435493743568918e-05,
+      "loss": 0.7051,
+      "step": 3707
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435105121189012e-05,
+      "loss": 0.7174,
+      "step": 3708
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434716368973765e-05,
+      "loss": 0.6128,
+      "step": 3709
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434327486928537e-05,
+      "loss": 0.6055,
+      "step": 3710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943393847505867e-05,
+      "loss": 0.7571,
+      "step": 3711
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943354933336952e-05,
+      "loss": 0.7219,
+      "step": 3712
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433160061866442e-05,
+      "loss": 0.7081,
+      "step": 3713
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943277066055479e-05,
+      "loss": 0.6796,
+      "step": 3714
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943238112943993e-05,
+      "loss": 0.7108,
+      "step": 3715
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9431991468527217e-05,
+      "loss": 0.6258,
+      "step": 3716
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9431601677822014e-05,
+      "loss": 0.7048,
+      "step": 3717
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9431211757329683e-05,
+      "loss": 0.6951,
+      "step": 3718
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943082170705559e-05,
+      "loss": 0.6646,
+      "step": 3719
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430431527005107e-05,
+      "loss": 0.615,
+      "step": 3720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94300412171836e-05,
+      "loss": 0.7168,
+      "step": 3721
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9429650777596437e-05,
+      "loss": 0.7025,
+      "step": 3722
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9429260208248994e-05,
+      "loss": 0.7148,
+      "step": 3723
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428869509146648e-05,
+      "loss": 0.7218,
+      "step": 3724
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428478680294768e-05,
+      "loss": 0.5834,
+      "step": 3725
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942808772169874e-05,
+      "loss": 0.5961,
+      "step": 3726
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9427696633363942e-05,
+      "loss": 0.72,
+      "step": 3727
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942730541529575e-05,
+      "loss": 0.7058,
+      "step": 3728
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9426914067499554e-05,
+      "loss": 0.6807,
+      "step": 3729
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942652258998074e-05,
+      "loss": 0.6758,
+      "step": 3730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942613098274469e-05,
+      "loss": 0.7181,
+      "step": 3731
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9425739245796794e-05,
+      "loss": 0.7331,
+      "step": 3732
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9425347379142447e-05,
+      "loss": 0.7568,
+      "step": 3733
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424955382787035e-05,
+      "loss": 0.7483,
+      "step": 3734
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424563256735956e-05,
+      "loss": 0.6739,
+      "step": 3735
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424171000994607e-05,
+      "loss": 0.7337,
+      "step": 3736
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423778615568386e-05,
+      "loss": 0.7253,
+      "step": 3737
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423386100462688e-05,
+      "loss": 0.6568,
+      "step": 3738
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422993455682917e-05,
+      "loss": 0.6455,
+      "step": 3739
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422600681234477e-05,
+      "loss": 0.7691,
+      "step": 3740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422207777122772e-05,
+      "loss": 0.7,
+      "step": 3741
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421814743353207e-05,
+      "loss": 0.5834,
+      "step": 3742
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421421579931198e-05,
+      "loss": 0.6398,
+      "step": 3743
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421028286862144e-05,
+      "loss": 0.6187,
+      "step": 3744
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942063486415146e-05,
+      "loss": 0.7219,
+      "step": 3745
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942024131180457e-05,
+      "loss": 0.7523,
+      "step": 3746
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941984762982688e-05,
+      "loss": 0.6805,
+      "step": 3747
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941945381822381e-05,
+      "loss": 0.6384,
+      "step": 3748
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419059877000776e-05,
+      "loss": 0.6352,
+      "step": 3749
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418665806163203e-05,
+      "loss": 0.6588,
+      "step": 3750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418271605716514e-05,
+      "loss": 0.7436,
+      "step": 3751
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941787727566613e-05,
+      "loss": 0.6776,
+      "step": 3752
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417482816017483e-05,
+      "loss": 0.7341,
+      "step": 3753
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417088226775993e-05,
+      "loss": 0.7173,
+      "step": 3754
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94166935079471e-05,
+      "loss": 0.7068,
+      "step": 3755
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416298659536225e-05,
+      "loss": 0.6882,
+      "step": 3756
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415903681548808e-05,
+      "loss": 0.6408,
+      "step": 3757
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415508573990284e-05,
+      "loss": 0.704,
+      "step": 3758
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415113336866086e-05,
+      "loss": 0.6101,
+      "step": 3759
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414717970181657e-05,
+      "loss": 0.7217,
+      "step": 3760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414322473942437e-05,
+      "loss": 0.735,
+      "step": 3761
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941392684815387e-05,
+      "loss": 0.6328,
+      "step": 3762
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9413531092821395e-05,
+      "loss": 0.6776,
+      "step": 3763
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941313520795046e-05,
+      "loss": 0.5988,
+      "step": 3764
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941273919354651e-05,
+      "loss": 0.6649,
+      "step": 3765
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9412343049615007e-05,
+      "loss": 0.7574,
+      "step": 3766
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411946776161388e-05,
+      "loss": 0.7048,
+      "step": 3767
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941155037319111e-05,
+      "loss": 0.6754,
+      "step": 3768
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411153840709632e-05,
+      "loss": 0.6504,
+      "step": 3769
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9410757178722406e-05,
+      "loss": 0.6034,
+      "step": 3770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9410360387234895e-05,
+      "loss": 0.6955,
+      "step": 3771
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940996346625255e-05,
+      "loss": 0.7241,
+      "step": 3772
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9409566415780845e-05,
+      "loss": 0.6269,
+      "step": 3773
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9409169235825237e-05,
+      "loss": 0.6822,
+      "step": 3774
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9408771926391193e-05,
+      "loss": 0.7308,
+      "step": 3775
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940837448748418e-05,
+      "loss": 0.7334,
+      "step": 3776
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407976919109666e-05,
+      "loss": 0.6773,
+      "step": 3777
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407579221273123e-05,
+      "loss": 0.7071,
+      "step": 3778
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407181393980026e-05,
+      "loss": 0.5574,
+      "step": 3779
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406783437235845e-05,
+      "loss": 0.7308,
+      "step": 3780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940638535104606e-05,
+      "loss": 0.6574,
+      "step": 3781
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405987135416145e-05,
+      "loss": 0.743,
+      "step": 3782
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405588790351583e-05,
+      "loss": 0.7389,
+      "step": 3783
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405190315857855e-05,
+      "loss": 0.7672,
+      "step": 3784
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404791711940445e-05,
+      "loss": 0.7087,
+      "step": 3785
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404392978604836e-05,
+      "loss": 0.753,
+      "step": 3786
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403994115856515e-05,
+      "loss": 0.7361,
+      "step": 3787
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403595123700974e-05,
+      "loss": 0.7162,
+      "step": 3788
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94031960021437e-05,
+      "loss": 0.6392,
+      "step": 3789
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940279675119019e-05,
+      "loss": 0.6175,
+      "step": 3790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9402397370845933e-05,
+      "loss": 0.7318,
+      "step": 3791
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401997861116425e-05,
+      "loss": 0.5845,
+      "step": 3792
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401598222007168e-05,
+      "loss": 0.6952,
+      "step": 3793
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401198453523654e-05,
+      "loss": 0.6575,
+      "step": 3794
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9400798555671395e-05,
+      "loss": 0.6938,
+      "step": 3795
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9400398528455883e-05,
+      "loss": 0.7563,
+      "step": 3796
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9399998371882628e-05,
+      "loss": 0.6352,
+      "step": 3797
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9399598085957135e-05,
+      "loss": 0.6353,
+      "step": 3798
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9399197670684913e-05,
+      "loss": 0.7023,
+      "step": 3799
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398797126071472e-05,
+      "loss": 0.6771,
+      "step": 3800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398396452122328e-05,
+      "loss": 0.7355,
+      "step": 3801
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9397995648842988e-05,
+      "loss": 0.6846,
+      "step": 3802
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939759471623897e-05,
+      "loss": 0.7251,
+      "step": 3803
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939719365431579e-05,
+      "loss": 0.6606,
+      "step": 3804
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396792463078972e-05,
+      "loss": 0.7733,
+      "step": 3805
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939639114253403e-05,
+      "loss": 0.7293,
+      "step": 3806
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395989692686487e-05,
+      "loss": 0.6413,
+      "step": 3807
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395588113541875e-05,
+      "loss": 0.6534,
+      "step": 3808
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395186405105713e-05,
+      "loss": 0.5913,
+      "step": 3809
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394784567383528e-05,
+      "loss": 0.7244,
+      "step": 3810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394382600380855e-05,
+      "loss": 0.669,
+      "step": 3811
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393980504103225e-05,
+      "loss": 0.7271,
+      "step": 3812
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393578278556163e-05,
+      "loss": 0.8082,
+      "step": 3813
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393175923745215e-05,
+      "loss": 0.6711,
+      "step": 3814
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9392773439675912e-05,
+      "loss": 0.6734,
+      "step": 3815
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9392370826353792e-05,
+      "loss": 0.7762,
+      "step": 3816
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93919680837844e-05,
+      "loss": 0.7594,
+      "step": 3817
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391565211973268e-05,
+      "loss": 0.6121,
+      "step": 3818
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939116221092595e-05,
+      "loss": 0.7805,
+      "step": 3819
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939075908064799e-05,
+      "loss": 0.6786,
+      "step": 3820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390355821144934e-05,
+      "loss": 0.6568,
+      "step": 3821
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9389952432422328e-05,
+      "loss": 0.6612,
+      "step": 3822
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9389548914485724e-05,
+      "loss": 0.7882,
+      "step": 3823
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938914526734068e-05,
+      "loss": 0.5807,
+      "step": 3824
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388741490992746e-05,
+      "loss": 0.6361,
+      "step": 3825
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388337585447482e-05,
+      "loss": 0.7321,
+      "step": 3826
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938793355071044e-05,
+      "loss": 0.6407,
+      "step": 3827
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9387529386787187e-05,
+      "loss": 0.6128,
+      "step": 3828
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938712509368328e-05,
+      "loss": 0.789,
+      "step": 3829
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9386720671404277e-05,
+      "loss": 0.6902,
+      "step": 3830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9386316119955757e-05,
+      "loss": 0.7351,
+      "step": 3831
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9385911439343278e-05,
+      "loss": 0.7189,
+      "step": 3832
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938550662957241e-05,
+      "loss": 0.6703,
+      "step": 3833
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938510169064872e-05,
+      "loss": 0.6226,
+      "step": 3834
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938469662257779e-05,
+      "loss": 0.5996,
+      "step": 3835
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9384291425365186e-05,
+      "loss": 0.6599,
+      "step": 3836
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9383886099016485e-05,
+      "loss": 0.7182,
+      "step": 3837
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938348064353727e-05,
+      "loss": 0.6297,
+      "step": 3838
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9383075058933113e-05,
+      "loss": 0.6755,
+      "step": 3839
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93826693452096e-05,
+      "loss": 0.7076,
+      "step": 3840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938226350237231e-05,
+      "loss": 0.7423,
+      "step": 3841
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938185753042683e-05,
+      "loss": 0.7095,
+      "step": 3842
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938145142937875e-05,
+      "loss": 0.6387,
+      "step": 3843
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9381045199233655e-05,
+      "loss": 0.5896,
+      "step": 3844
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9380638839997135e-05,
+      "loss": 0.7835,
+      "step": 3845
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938023235167478e-05,
+      "loss": 0.6068,
+      "step": 3846
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9379825734272186e-05,
+      "loss": 0.685,
+      "step": 3847
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9379418987794948e-05,
+      "loss": 0.6616,
+      "step": 3848
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937901211224867e-05,
+      "loss": 0.7338,
+      "step": 3849
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9378605107638932e-05,
+      "loss": 0.7662,
+      "step": 3850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9378197973971357e-05,
+      "loss": 0.6876,
+      "step": 3851
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937779071125153e-05,
+      "loss": 0.6325,
+      "step": 3852
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377383319485068e-05,
+      "loss": 0.7114,
+      "step": 3853
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937697579867757e-05,
+      "loss": 0.6031,
+      "step": 3854
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376568148834645e-05,
+      "loss": 0.6349,
+      "step": 3855
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376160369961904e-05,
+      "loss": 0.6005,
+      "step": 3856
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375752462064957e-05,
+      "loss": 0.6565,
+      "step": 3857
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375344425149422e-05,
+      "loss": 0.6715,
+      "step": 3858
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374936259220903e-05,
+      "loss": 0.6754,
+      "step": 3859
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374527964285027e-05,
+      "loss": 0.5959,
+      "step": 3860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374119540347413e-05,
+      "loss": 0.5996,
+      "step": 3861
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937371098741367e-05,
+      "loss": 0.6151,
+      "step": 3862
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937330230548943e-05,
+      "loss": 0.7283,
+      "step": 3863
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937289349458032e-05,
+      "loss": 0.7256,
+      "step": 3864
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9372484554691954e-05,
+      "loss": 0.6853,
+      "step": 3865
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9372075485829966e-05,
+      "loss": 0.7769,
+      "step": 3866
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937166628799998e-05,
+      "loss": 0.6148,
+      "step": 3867
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371256961207637e-05,
+      "loss": 0.6805,
+      "step": 3868
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9370847505458562e-05,
+      "loss": 0.7529,
+      "step": 3869
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937043792075839e-05,
+      "loss": 0.7216,
+      "step": 3870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937002820711276e-05,
+      "loss": 0.7168,
+      "step": 3871
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9369618364527308e-05,
+      "loss": 0.7529,
+      "step": 3872
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9369208393007678e-05,
+      "loss": 0.6152,
+      "step": 3873
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9368798292559506e-05,
+      "loss": 0.6389,
+      "step": 3874
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9368388063188437e-05,
+      "loss": 0.7033,
+      "step": 3875
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9367977704900113e-05,
+      "loss": 0.6567,
+      "step": 3876
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936756721770019e-05,
+      "loss": 0.5425,
+      "step": 3877
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9367156601594308e-05,
+      "loss": 0.6146,
+      "step": 3878
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936674585658812e-05,
+      "loss": 0.7305,
+      "step": 3879
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936633498268728e-05,
+      "loss": 0.7465,
+      "step": 3880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936592397989744e-05,
+      "loss": 0.7362,
+      "step": 3881
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9365512848224257e-05,
+      "loss": 0.6071,
+      "step": 3882
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9365101587673388e-05,
+      "loss": 0.7091,
+      "step": 3883
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936469019825049e-05,
+      "loss": 0.7129,
+      "step": 3884
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936427867996123e-05,
+      "loss": 0.6438,
+      "step": 3885
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363867032811267e-05,
+      "loss": 0.6967,
+      "step": 3886
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363455256806263e-05,
+      "loss": 0.722,
+      "step": 3887
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363043351951888e-05,
+      "loss": 0.6577,
+      "step": 3888
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362631318253807e-05,
+      "loss": 0.6478,
+      "step": 3889
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362219155717698e-05,
+      "loss": 0.5921,
+      "step": 3890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9361806864349223e-05,
+      "loss": 0.7156,
+      "step": 3891
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9361394444154063e-05,
+      "loss": 0.6204,
+      "step": 3892
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9360981895137888e-05,
+      "loss": 0.6164,
+      "step": 3893
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9360569217306375e-05,
+      "loss": 0.6186,
+      "step": 3894
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9360156410665206e-05,
+      "loss": 0.6939,
+      "step": 3895
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9359743475220057e-05,
+      "loss": 0.6989,
+      "step": 3896
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935933041097662e-05,
+      "loss": 0.6324,
+      "step": 3897
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935891721794057e-05,
+      "loss": 0.6699,
+      "step": 3898
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358503896117594e-05,
+      "loss": 0.6937,
+      "step": 3899
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358090445513383e-05,
+      "loss": 0.7411,
+      "step": 3900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9357676866133625e-05,
+      "loss": 0.6776,
+      "step": 3901
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935726315798401e-05,
+      "loss": 0.691,
+      "step": 3902
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935684932107023e-05,
+      "loss": 0.7071,
+      "step": 3903
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356435355397988e-05,
+      "loss": 0.721,
+      "step": 3904
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356021260972965e-05,
+      "loss": 0.7422,
+      "step": 3905
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355607037800875e-05,
+      "loss": 0.7272,
+      "step": 3906
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355192685887408e-05,
+      "loss": 0.7329,
+      "step": 3907
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9354778205238275e-05,
+      "loss": 0.7338,
+      "step": 3908
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9354363595859168e-05,
+      "loss": 0.6665,
+      "step": 3909
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93539488577558e-05,
+      "loss": 0.6914,
+      "step": 3910
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935353399093388e-05,
+      "loss": 0.7637,
+      "step": 3911
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935311899539911e-05,
+      "loss": 0.6545,
+      "step": 3912
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352703871157208e-05,
+      "loss": 0.6921,
+      "step": 3913
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352288618213883e-05,
+      "loss": 0.6766,
+      "step": 3914
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935187323657484e-05,
+      "loss": 0.5637,
+      "step": 3915
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351457726245814e-05,
+      "loss": 0.7502,
+      "step": 3916
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351042087232508e-05,
+      "loss": 0.7742,
+      "step": 3917
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9350626319540647e-05,
+      "loss": 0.6884,
+      "step": 3918
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9350210423175953e-05,
+      "loss": 0.6685,
+      "step": 3919
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9349794398144146e-05,
+      "loss": 0.7258,
+      "step": 3920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9349378244450955e-05,
+      "loss": 0.6103,
+      "step": 3921
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348961962102103e-05,
+      "loss": 0.7538,
+      "step": 3922
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348545551103318e-05,
+      "loss": 0.591,
+      "step": 3923
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348129011460333e-05,
+      "loss": 0.6999,
+      "step": 3924
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934771234317888e-05,
+      "loss": 0.6935,
+      "step": 3925
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934729554626469e-05,
+      "loss": 0.5682,
+      "step": 3926
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93468786207235e-05,
+      "loss": 0.6923,
+      "step": 3927
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934646156656105e-05,
+      "loss": 0.7585,
+      "step": 3928
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9346044383783074e-05,
+      "loss": 0.6661,
+      "step": 3929
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345627072395314e-05,
+      "loss": 0.629,
+      "step": 3930
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345209632403516e-05,
+      "loss": 0.7183,
+      "step": 3931
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934479206381342e-05,
+      "loss": 0.7729,
+      "step": 3932
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344374366630777e-05,
+      "loss": 0.7042,
+      "step": 3933
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9343956540861328e-05,
+      "loss": 0.6739,
+      "step": 3934
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934353858651083e-05,
+      "loss": 0.6887,
+      "step": 3935
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934312050358503e-05,
+      "loss": 0.6681,
+      "step": 3936
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9342702292089682e-05,
+      "loss": 0.699,
+      "step": 3937
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934228395203054e-05,
+      "loss": 0.6162,
+      "step": 3938
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341865483413364e-05,
+      "loss": 0.6504,
+      "step": 3939
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934144688624391e-05,
+      "loss": 0.6853,
+      "step": 3940
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341028160527943e-05,
+      "loss": 0.6349,
+      "step": 3941
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340609306271214e-05,
+      "loss": 0.7231,
+      "step": 3942
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340190323479495e-05,
+      "loss": 0.7009,
+      "step": 3943
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339771212158553e-05,
+      "loss": 0.7168,
+      "step": 3944
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339351972314147e-05,
+      "loss": 0.6718,
+      "step": 3945
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933893260395206e-05,
+      "loss": 0.6011,
+      "step": 3946
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9338513107078046e-05,
+      "loss": 0.6942,
+      "step": 3947
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9338093481697894e-05,
+      "loss": 0.7719,
+      "step": 3948
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337673727817362e-05,
+      "loss": 0.5919,
+      "step": 3949
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933725384544224e-05,
+      "loss": 0.6703,
+      "step": 3950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9336833834578298e-05,
+      "loss": 0.7113,
+      "step": 3951
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9336413695231323e-05,
+      "loss": 0.7206,
+      "step": 3952
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933599342740709e-05,
+      "loss": 0.6823,
+      "step": 3953
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335573031111382e-05,
+      "loss": 0.7119,
+      "step": 3954
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335152506349987e-05,
+      "loss": 0.7743,
+      "step": 3955
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933473185312869e-05,
+      "loss": 0.6759,
+      "step": 3956
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334311071453283e-05,
+      "loss": 0.6434,
+      "step": 3957
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933389016132955e-05,
+      "loss": 0.6669,
+      "step": 3958
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933346912276329e-05,
+      "loss": 0.7399,
+      "step": 3959
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9333047955760293e-05,
+      "loss": 0.6221,
+      "step": 3960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332626660326357e-05,
+      "loss": 0.6853,
+      "step": 3961
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332205236467277e-05,
+      "loss": 0.6549,
+      "step": 3962
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933178368418885e-05,
+      "loss": 0.6346,
+      "step": 3963
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9331362003496883e-05,
+      "loss": 0.6721,
+      "step": 3964
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9330940194397176e-05,
+      "loss": 0.7185,
+      "step": 3965
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933051825689553e-05,
+      "loss": 0.7242,
+      "step": 3966
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933009619099776e-05,
+      "loss": 0.7386,
+      "step": 3967
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932967399670966e-05,
+      "loss": 0.7418,
+      "step": 3968
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329251674037053e-05,
+      "loss": 0.7046,
+      "step": 3969
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9328829222985746e-05,
+      "loss": 0.6746,
+      "step": 3970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932840664356155e-05,
+      "loss": 0.7154,
+      "step": 3971
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327983935770286e-05,
+      "loss": 0.7113,
+      "step": 3972
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932756109961776e-05,
+      "loss": 0.619,
+      "step": 3973
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327138135109803e-05,
+      "loss": 0.734,
+      "step": 3974
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9326715042252226e-05,
+      "loss": 0.5943,
+      "step": 3975
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932629182105086e-05,
+      "loss": 0.7264,
+      "step": 3976
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932586847151152e-05,
+      "loss": 0.7623,
+      "step": 3977
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325444993640032e-05,
+      "loss": 0.6505,
+      "step": 3978
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932502138744223e-05,
+      "loss": 0.7377,
+      "step": 3979
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9324597652923943e-05,
+      "loss": 0.739,
+      "step": 3980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9324173790090995e-05,
+      "loss": 0.727,
+      "step": 3981
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323749798949225e-05,
+      "loss": 0.6961,
+      "step": 3982
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323325679504464e-05,
+      "loss": 0.7485,
+      "step": 3983
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932290143176255e-05,
+      "loss": 0.6914,
+      "step": 3984
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322477055729317e-05,
+      "loss": 0.7573,
+      "step": 3985
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322052551410617e-05,
+      "loss": 0.6564,
+      "step": 3986
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9321627918812273e-05,
+      "loss": 0.7123,
+      "step": 3987
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9321203157940144e-05,
+      "loss": 0.6689,
+      "step": 3988
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9320778268800068e-05,
+      "loss": 0.7501,
+      "step": 3989
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932035325139789e-05,
+      "loss": 0.6215,
+      "step": 3990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9319928105739464e-05,
+      "loss": 0.5894,
+      "step": 3991
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931950283183064e-05,
+      "loss": 0.6449,
+      "step": 3992
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931907742967727e-05,
+      "loss": 0.7123,
+      "step": 3993
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9318651899285198e-05,
+      "loss": 0.7635,
+      "step": 3994
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9318226240660297e-05,
+      "loss": 0.7055,
+      "step": 3995
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317800453808408e-05,
+      "loss": 0.7712,
+      "step": 3996
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317374538735402e-05,
+      "loss": 0.5993,
+      "step": 3997
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931694849544713e-05,
+      "loss": 0.6816,
+      "step": 3998
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9316522323949466e-05,
+      "loss": 0.6815,
+      "step": 3999
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9316096024248265e-05,
+      "loss": 0.7051,
+      "step": 4000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315669596349397e-05,
+      "loss": 0.701,
+      "step": 4001
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315243040258733e-05,
+      "loss": 0.5934,
+      "step": 4002
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9314816355982138e-05,
+      "loss": 0.6923,
+      "step": 4003
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9314389543525487e-05,
+      "loss": 0.6935,
+      "step": 4004
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9313962602894648e-05,
+      "loss": 0.6887,
+      "step": 4005
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.93135355340955e-05,
+      "loss": 0.683,
+      "step": 4006
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931310833713392e-05,
+      "loss": 0.6577,
+      "step": 4007
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931268101201579e-05,
+      "loss": 0.7387,
+      "step": 4008
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312253558746984e-05,
+      "loss": 0.6278,
+      "step": 4009
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9311825977333383e-05,
+      "loss": 0.7309,
+      "step": 4010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9311398267780878e-05,
+      "loss": 0.6856,
+      "step": 4011
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931097043009535e-05,
+      "loss": 0.6043,
+      "step": 4012
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9310542464282688e-05,
+      "loss": 0.6354,
+      "step": 4013
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931011437034878e-05,
+      "loss": 0.7342,
+      "step": 4014
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930968614829952e-05,
+      "loss": 0.6843,
+      "step": 4015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9309257798140793e-05,
+      "loss": 0.6484,
+      "step": 4016
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308829319878503e-05,
+      "loss": 0.7504,
+      "step": 4017
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308400713518542e-05,
+      "loss": 0.6653,
+      "step": 4018
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930797197906681e-05,
+      "loss": 0.6876,
+      "step": 4019
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.93075431165292e-05,
+      "loss": 0.6463,
+      "step": 4020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9307114125911622e-05,
+      "loss": 0.6943,
+      "step": 4021
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9306685007219975e-05,
+      "loss": 0.5978,
+      "step": 4022
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9306255760460165e-05,
+      "loss": 0.606,
+      "step": 4023
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9305826385638097e-05,
+      "loss": 0.7098,
+      "step": 4024
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9305396882759685e-05,
+      "loss": 0.5874,
+      "step": 4025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930496725183083e-05,
+      "loss": 0.606,
+      "step": 4026
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9304537492857453e-05,
+      "loss": 0.5483,
+      "step": 4027
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9304107605845465e-05,
+      "loss": 0.6135,
+      "step": 4028
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9303677590800775e-05,
+      "loss": 0.7376,
+      "step": 4029
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930324744772931e-05,
+      "loss": 0.6389,
+      "step": 4030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930281717663699e-05,
+      "loss": 0.7205,
+      "step": 4031
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930238677752973e-05,
+      "loss": 0.7598,
+      "step": 4032
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301956250413448e-05,
+      "loss": 0.7302,
+      "step": 4033
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930152559529408e-05,
+      "loss": 0.6314,
+      "step": 4034
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301094812177546e-05,
+      "loss": 0.6635,
+      "step": 4035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9300663901069775e-05,
+      "loss": 0.7053,
+      "step": 4036
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9300232861976695e-05,
+      "loss": 0.7334,
+      "step": 4037
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9299801694904238e-05,
+      "loss": 0.656,
+      "step": 4038
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929937039985834e-05,
+      "loss": 0.681,
+      "step": 4039
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298938976844934e-05,
+      "loss": 0.7731,
+      "step": 4040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298507425869955e-05,
+      "loss": 0.634,
+      "step": 4041
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298075746939348e-05,
+      "loss": 0.6584,
+      "step": 4042
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9297643940059046e-05,
+      "loss": 0.702,
+      "step": 4043
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9297212005234994e-05,
+      "loss": 0.7131,
+      "step": 4044
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296779942473134e-05,
+      "loss": 0.6991,
+      "step": 4045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296347751779415e-05,
+      "loss": 0.6905,
+      "step": 4046
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929591543315978e-05,
+      "loss": 0.7303,
+      "step": 4047
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295482986620185e-05,
+      "loss": 0.6899,
+      "step": 4048
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295050412166572e-05,
+      "loss": 0.7118,
+      "step": 4049
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9294617709804902e-05,
+      "loss": 0.721,
+      "step": 4050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929418487954112e-05,
+      "loss": 0.6874,
+      "step": 4051
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929375192138119e-05,
+      "loss": 0.6795,
+      "step": 4052
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9293318835331068e-05,
+      "loss": 0.5925,
+      "step": 4053
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9292885621396714e-05,
+      "loss": 0.7167,
+      "step": 4054
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929245227958409e-05,
+      "loss": 0.747,
+      "step": 4055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9292018809899152e-05,
+      "loss": 0.6814,
+      "step": 4056
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9291585212347876e-05,
+      "loss": 0.6182,
+      "step": 4057
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929115148693622e-05,
+      "loss": 0.7176,
+      "step": 4058
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9290717633670154e-05,
+      "loss": 0.6533,
+      "step": 4059
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9290283652555654e-05,
+      "loss": 0.6968,
+      "step": 4060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9289849543598685e-05,
+      "loss": 0.6604,
+      "step": 4061
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9289415306805223e-05,
+      "loss": 0.6549,
+      "step": 4062
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928898094218125e-05,
+      "loss": 0.6465,
+      "step": 4063
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928854644973273e-05,
+      "loss": 0.6476,
+      "step": 4064
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9288111829465653e-05,
+      "loss": 0.6939,
+      "step": 4065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928767708138599e-05,
+      "loss": 0.7205,
+      "step": 4066
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287242205499736e-05,
+      "loss": 0.6497,
+      "step": 4067
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9286807201812866e-05,
+      "loss": 0.7713,
+      "step": 4068
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928637207033137e-05,
+      "loss": 0.6606,
+      "step": 4069
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285936811061233e-05,
+      "loss": 0.6876,
+      "step": 4070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928550142400845e-05,
+      "loss": 0.6662,
+      "step": 4071
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285065909179006e-05,
+      "loss": 0.7025,
+      "step": 4072
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284630266578897e-05,
+      "loss": 0.616,
+      "step": 4073
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284194496214118e-05,
+      "loss": 0.6846,
+      "step": 4074
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9283758598090664e-05,
+      "loss": 0.5792,
+      "step": 4075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9283322572214533e-05,
+      "loss": 0.7492,
+      "step": 4076
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282886418591725e-05,
+      "loss": 0.7711,
+      "step": 4077
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928245013722825e-05,
+      "loss": 0.7315,
+      "step": 4078
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282013728130096e-05,
+      "loss": 0.6467,
+      "step": 4079
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928157719130328e-05,
+      "loss": 0.6761,
+      "step": 4080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281140526753808e-05,
+      "loss": 0.6239,
+      "step": 4081
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9280703734487688e-05,
+      "loss": 0.7698,
+      "step": 4082
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928026681451093e-05,
+      "loss": 0.6373,
+      "step": 4083
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279829766829546e-05,
+      "loss": 0.7335,
+      "step": 4084
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927939259144955e-05,
+      "loss": 0.6694,
+      "step": 4085
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278955288376958e-05,
+      "loss": 0.7234,
+      "step": 4086
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278517857617788e-05,
+      "loss": 0.6961,
+      "step": 4087
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927808029917806e-05,
+      "loss": 0.7162,
+      "step": 4088
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9277642613063794e-05,
+      "loss": 0.6385,
+      "step": 4089
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9277204799281014e-05,
+      "loss": 0.6969,
+      "step": 4090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276766857835746e-05,
+      "loss": 0.713,
+      "step": 4091
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927632878873401e-05,
+      "loss": 0.6338,
+      "step": 4092
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9275890591981842e-05,
+      "loss": 0.6585,
+      "step": 4093
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927545226758527e-05,
+      "loss": 0.7177,
+      "step": 4094
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927501381555032e-05,
+      "loss": 0.6558,
+      "step": 4095
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9274575235883038e-05,
+      "loss": 0.7454,
+      "step": 4096
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9274136528589444e-05,
+      "loss": 0.5733,
+      "step": 4097
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927369769367559e-05,
+      "loss": 0.7231,
+      "step": 4098
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92732587311475e-05,
+      "loss": 0.6431,
+      "step": 4099
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9272819641011226e-05,
+      "loss": 0.7812,
+      "step": 4100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92723804232728e-05,
+      "loss": 0.7259,
+      "step": 4101
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927194107793828e-05,
+      "loss": 0.7263,
+      "step": 4102
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92715016050137e-05,
+      "loss": 0.6615,
+      "step": 4103
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927106200450511e-05,
+      "loss": 0.6402,
+      "step": 4104
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270622276418562e-05,
+      "loss": 0.6799,
+      "step": 4105
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270182420760104e-05,
+      "loss": 0.6465,
+      "step": 4106
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926974243753579e-05,
+      "loss": 0.712,
+      "step": 4107
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9269302326751678e-05,
+      "loss": 0.6952,
+      "step": 4108
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926886208841382e-05,
+      "loss": 0.699,
+      "step": 4109
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9268421722528273e-05,
+      "loss": 0.7378,
+      "step": 4110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267981229101105e-05,
+      "loss": 0.6269,
+      "step": 4111
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267540608138368e-05,
+      "loss": 0.6052,
+      "step": 4112
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267099859646126e-05,
+      "loss": 0.7217,
+      "step": 4113
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9266658983630452e-05,
+      "loss": 0.6809,
+      "step": 4114
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9266217980097407e-05,
+      "loss": 0.7524,
+      "step": 4115
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926577684905306e-05,
+      "loss": 0.6676,
+      "step": 4116
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926533559050348e-05,
+      "loss": 0.6363,
+      "step": 4117
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9264894204454742e-05,
+      "loss": 0.7515,
+      "step": 4118
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9264452690912924e-05,
+      "loss": 0.6133,
+      "step": 4119
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9264011049884092e-05,
+      "loss": 0.6868,
+      "step": 4120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9263569281374325e-05,
+      "loss": 0.7404,
+      "step": 4121
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9263127385389707e-05,
+      "loss": 0.7201,
+      "step": 4122
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926268536193632e-05,
+      "loss": 0.7158,
+      "step": 4123
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926224321102024e-05,
+      "loss": 0.6726,
+      "step": 4124
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9261800932647557e-05,
+      "loss": 0.7279,
+      "step": 4125
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9261358526824356e-05,
+      "loss": 0.6744,
+      "step": 4126
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260915993556723e-05,
+      "loss": 0.6904,
+      "step": 4127
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260473332850747e-05,
+      "loss": 0.7197,
+      "step": 4128
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260030544712523e-05,
+      "loss": 0.7142,
+      "step": 4129
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259587629148137e-05,
+      "loss": 0.7629,
+      "step": 4130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259144586163696e-05,
+      "loss": 0.7118,
+      "step": 4131
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9258701415765287e-05,
+      "loss": 0.7303,
+      "step": 4132
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925825811795901e-05,
+      "loss": 0.7561,
+      "step": 4133
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9257814692750967e-05,
+      "loss": 0.7372,
+      "step": 4134
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925737114014726e-05,
+      "loss": 0.7753,
+      "step": 4135
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256927460153992e-05,
+      "loss": 0.7332,
+      "step": 4136
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925648365277727e-05,
+      "loss": 0.6687,
+      "step": 4137
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256039718023195e-05,
+      "loss": 0.6974,
+      "step": 4138
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925559565589788e-05,
+      "loss": 0.607,
+      "step": 4139
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925515146640744e-05,
+      "loss": 0.6874,
+      "step": 4140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925470714955798e-05,
+      "loss": 0.7137,
+      "step": 4141
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925426270535562e-05,
+      "loss": 0.6857,
+      "step": 4142
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253818133806473e-05,
+      "loss": 0.7425,
+      "step": 4143
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253373434916657e-05,
+      "loss": 0.7211,
+      "step": 4144
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9252928608692296e-05,
+      "loss": 0.6889,
+      "step": 4145
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9252483655139503e-05,
+      "loss": 0.7364,
+      "step": 4146
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9252038574264403e-05,
+      "loss": 0.7551,
+      "step": 4147
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251593366073126e-05,
+      "loss": 0.7425,
+      "step": 4148
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251148030571795e-05,
+      "loss": 0.7182,
+      "step": 4149
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9250702567766538e-05,
+      "loss": 0.6244,
+      "step": 4150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9250256977663487e-05,
+      "loss": 0.6709,
+      "step": 4151
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924981126026877e-05,
+      "loss": 0.6593,
+      "step": 4152
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924936541558852e-05,
+      "loss": 0.6817,
+      "step": 4153
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9248919443628878e-05,
+      "loss": 0.7574,
+      "step": 4154
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9248473344395976e-05,
+      "loss": 0.6072,
+      "step": 4155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9248027117895954e-05,
+      "loss": 0.5834,
+      "step": 4156
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9247580764134955e-05,
+      "loss": 0.775,
+      "step": 4157
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9247134283119116e-05,
+      "loss": 0.7054,
+      "step": 4158
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924668767485459e-05,
+      "loss": 0.6761,
+      "step": 4159
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9246240939347512e-05,
+      "loss": 0.5624,
+      "step": 4160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9245794076604035e-05,
+      "loss": 0.7657,
+      "step": 4161
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9245347086630307e-05,
+      "loss": 0.7375,
+      "step": 4162
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924489996943248e-05,
+      "loss": 0.6571,
+      "step": 4163
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244452725016707e-05,
+      "loss": 0.748,
+      "step": 4164
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924400535338914e-05,
+      "loss": 0.6766,
+      "step": 4165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9243557854555937e-05,
+      "loss": 0.641,
+      "step": 4166
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9243110228523254e-05,
+      "loss": 0.6879,
+      "step": 4167
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9242662475297255e-05,
+      "loss": 0.5757,
+      "step": 4168
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92422145948841e-05,
+      "loss": 0.6911,
+      "step": 4169
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924176658728995e-05,
+      "loss": 0.6925,
+      "step": 4170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9241318452520973e-05,
+      "loss": 0.7604,
+      "step": 4171
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924087019058333e-05,
+      "loss": 0.62,
+      "step": 4172
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9240421801483195e-05,
+      "loss": 0.7364,
+      "step": 4173
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923997328522674e-05,
+      "loss": 0.6984,
+      "step": 4174
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9239524641820127e-05,
+      "loss": 0.6932,
+      "step": 4175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923907587126954e-05,
+      "loss": 0.6737,
+      "step": 4176
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9238626973581157e-05,
+      "loss": 0.7176,
+      "step": 4177
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9238177948761142e-05,
+      "loss": 0.7212,
+      "step": 4178
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923772879681568e-05,
+      "loss": 0.6966,
+      "step": 4179
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9237279517750958e-05,
+      "loss": 0.7483,
+      "step": 4180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923683011157315e-05,
+      "loss": 0.6914,
+      "step": 4181
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236380578288444e-05,
+      "loss": 0.8004,
+      "step": 4182
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235930917903024e-05,
+      "loss": 0.6142,
+      "step": 4183
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923548113042308e-05,
+      "loss": 0.6658,
+      "step": 4184
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235031215854803e-05,
+      "loss": 0.7495,
+      "step": 4185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923458117420438e-05,
+      "loss": 0.7608,
+      "step": 4186
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234131005478003e-05,
+      "loss": 0.723,
+      "step": 4187
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9233680709681873e-05,
+      "loss": 0.7086,
+      "step": 4188
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9233230286822184e-05,
+      "loss": 0.7867,
+      "step": 4189
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923277973690513e-05,
+      "loss": 0.6615,
+      "step": 4190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9232329059936917e-05,
+      "loss": 0.6589,
+      "step": 4191
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9231878255923742e-05,
+      "loss": 0.7301,
+      "step": 4192
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9231427324871814e-05,
+      "loss": 0.6712,
+      "step": 4193
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923097626678733e-05,
+      "loss": 0.5926,
+      "step": 4194
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92305250816765e-05,
+      "loss": 0.6807,
+      "step": 4195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923007376954554e-05,
+      "loss": 0.6753,
+      "step": 4196
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922962233040065e-05,
+      "loss": 0.6043,
+      "step": 4197
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922917076424805e-05,
+      "loss": 0.7454,
+      "step": 4198
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922871907109395e-05,
+      "loss": 0.7102,
+      "step": 4199
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9228267250944566e-05,
+      "loss": 0.7104,
+      "step": 4200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9227815303806117e-05,
+      "loss": 0.6854,
+      "step": 4201
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922736322968482e-05,
+      "loss": 0.6244,
+      "step": 4202
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9226911028586895e-05,
+      "loss": 0.6813,
+      "step": 4203
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922645870051857e-05,
+      "loss": 0.7264,
+      "step": 4204
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9226006245486067e-05,
+      "loss": 0.6815,
+      "step": 4205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922555366349561e-05,
+      "loss": 0.6518,
+      "step": 4206
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922510095455343e-05,
+      "loss": 0.6102,
+      "step": 4207
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9224648118665752e-05,
+      "loss": 0.7325,
+      "step": 4208
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922419515583881e-05,
+      "loss": 0.7089,
+      "step": 4209
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223742066078838e-05,
+      "loss": 0.694,
+      "step": 4210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223288849392073e-05,
+      "loss": 0.596,
+      "step": 4211
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222835505784746e-05,
+      "loss": 0.7253,
+      "step": 4212
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222382035263105e-05,
+      "loss": 0.5823,
+      "step": 4213
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9221928437833373e-05,
+      "loss": 0.5909,
+      "step": 4214
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9221474713501813e-05,
+      "loss": 0.7093,
+      "step": 4215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922102086227465e-05,
+      "loss": 0.7094,
+      "step": 4216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220566884158144e-05,
+      "loss": 0.6419,
+      "step": 4217
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220112779158532e-05,
+      "loss": 0.6982,
+      "step": 4218
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921965854728207e-05,
+      "loss": 0.7567,
+      "step": 4219
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9219204188535e-05,
+      "loss": 0.7004,
+      "step": 4220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9218749702923583e-05,
+      "loss": 0.7335,
+      "step": 4221
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921829509045407e-05,
+      "loss": 0.748,
+      "step": 4222
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217840351132718e-05,
+      "loss": 0.7032,
+      "step": 4223
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921738548496578e-05,
+      "loss": 0.7371,
+      "step": 4224
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9216930491959522e-05,
+      "loss": 0.6844,
+      "step": 4225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9216475372120198e-05,
+      "loss": 0.6877,
+      "step": 4226
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9216020125454075e-05,
+      "loss": 0.6525,
+      "step": 4227
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921556475196742e-05,
+      "loss": 0.7317,
+      "step": 4228
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92151092516665e-05,
+      "loss": 0.5999,
+      "step": 4229
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921465362455757e-05,
+      "loss": 0.6937,
+      "step": 4230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9214197870646916e-05,
+      "loss": 0.6369,
+      "step": 4231
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92137419899408e-05,
+      "loss": 0.6052,
+      "step": 4232
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92132859824455e-05,
+      "loss": 0.7111,
+      "step": 4233
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921282984816729e-05,
+      "loss": 0.7831,
+      "step": 4234
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9212373587112443e-05,
+      "loss": 0.7216,
+      "step": 4235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9211917199287243e-05,
+      "loss": 0.7931,
+      "step": 4236
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921146068469797e-05,
+      "loss": 0.605,
+      "step": 4237
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92110040433509e-05,
+      "loss": 0.7382,
+      "step": 4238
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921054727525232e-05,
+      "loss": 0.7008,
+      "step": 4239
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921009038040852e-05,
+      "loss": 0.6452,
+      "step": 4240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209633358825783e-05,
+      "loss": 0.7266,
+      "step": 4241
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209176210510398e-05,
+      "loss": 0.7048,
+      "step": 4242
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920871893546866e-05,
+      "loss": 0.7234,
+      "step": 4243
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208261533706855e-05,
+      "loss": 0.7244,
+      "step": 4244
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920780400523128e-05,
+      "loss": 0.774,
+      "step": 4245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920734635004823e-05,
+      "loss": 0.6856,
+      "step": 4246
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9206888568164003e-05,
+      "loss": 0.6972,
+      "step": 4247
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9206430659584903e-05,
+      "loss": 0.696,
+      "step": 4248
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9205972624317227e-05,
+      "loss": 0.6834,
+      "step": 4249
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920551446236728e-05,
+      "loss": 0.7889,
+      "step": 4250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920505617374136e-05,
+      "loss": 0.7706,
+      "step": 4251
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9204597758445784e-05,
+      "loss": 0.6791,
+      "step": 4252
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9204139216486854e-05,
+      "loss": 0.623,
+      "step": 4253
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203680547870877e-05,
+      "loss": 0.7582,
+      "step": 4254
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203221752604173e-05,
+      "loss": 0.7158,
+      "step": 4255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920276283069305e-05,
+      "loss": 0.6728,
+      "step": 4256
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9202303782143822e-05,
+      "loss": 0.7124,
+      "step": 4257
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920184460696281e-05,
+      "loss": 0.5951,
+      "step": 4258
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920138530515633e-05,
+      "loss": 0.6876,
+      "step": 4259
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9200925876730705e-05,
+      "loss": 0.6916,
+      "step": 4260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920046632169225e-05,
+      "loss": 0.6411,
+      "step": 4261
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9200006640047303e-05,
+      "loss": 0.6405,
+      "step": 4262
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9199546831802177e-05,
+      "loss": 0.6931,
+      "step": 4263
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.91990868969632e-05,
+      "loss": 0.8014,
+      "step": 4264
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919862683553671e-05,
+      "loss": 0.6367,
+      "step": 4265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9198166647529026e-05,
+      "loss": 0.7305,
+      "step": 4266
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197706332946487e-05,
+      "loss": 0.6576,
+      "step": 4267
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919724589179543e-05,
+      "loss": 0.5699,
+      "step": 4268
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9196785324082186e-05,
+      "loss": 0.651,
+      "step": 4269
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.91963246298131e-05,
+      "loss": 0.7037,
+      "step": 4270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9195863808994503e-05,
+      "loss": 0.6725,
+      "step": 4271
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919540286163274e-05,
+      "loss": 0.6727,
+      "step": 4272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9194941787734153e-05,
+      "loss": 0.5912,
+      "step": 4273
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9194480587305086e-05,
+      "loss": 0.7182,
+      "step": 4274
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9194019260351892e-05,
+      "loss": 0.7924,
+      "step": 4275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919355780688091e-05,
+      "loss": 0.7374,
+      "step": 4276
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9193096226898497e-05,
+      "loss": 0.7248,
+      "step": 4277
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9192634520411e-05,
+      "loss": 0.6493,
+      "step": 4278
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9192172687424776e-05,
+      "loss": 0.6174,
+      "step": 4279
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.919171072794618e-05,
+      "loss": 0.7459,
+      "step": 4280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9191248641981566e-05,
+      "loss": 0.6339,
+      "step": 4281
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190786429537293e-05,
+      "loss": 0.7385,
+      "step": 4282
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.919032409061973e-05,
+      "loss": 0.7091,
+      "step": 4283
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9189861625235226e-05,
+      "loss": 0.7091,
+      "step": 4284
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9189399033390154e-05,
+      "loss": 0.7162,
+      "step": 4285
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9188936315090873e-05,
+      "loss": 0.7183,
+      "step": 4286
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9188473470343755e-05,
+      "loss": 0.6736,
+      "step": 4287
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918801049915517e-05,
+      "loss": 0.7018,
+      "step": 4288
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187547401531487e-05,
+      "loss": 0.6832,
+      "step": 4289
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187084177479082e-05,
+      "loss": 0.7067,
+      "step": 4290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918662082700432e-05,
+      "loss": 0.7139,
+      "step": 4291
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9186157350113586e-05,
+      "loss": 0.6732,
+      "step": 4292
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185693746813258e-05,
+      "loss": 0.6707,
+      "step": 4293
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918523001710971e-05,
+      "loss": 0.6562,
+      "step": 4294
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918476616100933e-05,
+      "loss": 0.7674,
+      "step": 4295
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918430217851849e-05,
+      "loss": 0.7313,
+      "step": 4296
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9183838069643586e-05,
+      "loss": 0.7314,
+      "step": 4297
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9183373834391e-05,
+      "loss": 0.7475,
+      "step": 4298
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9182909472767123e-05,
+      "loss": 0.7229,
+      "step": 4299
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9182444984778338e-05,
+      "loss": 0.6702,
+      "step": 4300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181980370431047e-05,
+      "loss": 0.6342,
+      "step": 4301
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181515629731632e-05,
+      "loss": 0.6684,
+      "step": 4302
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91810507626865e-05,
+      "loss": 0.7411,
+      "step": 4303
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918058576930204e-05,
+      "loss": 0.6027,
+      "step": 4304
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9180120649584655e-05,
+      "loss": 0.7171,
+      "step": 4305
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179655403540742e-05,
+      "loss": 0.7304,
+      "step": 4306
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179190031176705e-05,
+      "loss": 0.6814,
+      "step": 4307
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9178724532498947e-05,
+      "loss": 0.7244,
+      "step": 4308
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9178258907513872e-05,
+      "loss": 0.714,
+      "step": 4309
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177793156227894e-05,
+      "loss": 0.7238,
+      "step": 4310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177327278647416e-05,
+      "loss": 0.685,
+      "step": 4311
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917686127477885e-05,
+      "loss": 0.7483,
+      "step": 4312
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9176395144628614e-05,
+      "loss": 0.598,
+      "step": 4313
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9175928888203113e-05,
+      "loss": 0.6957,
+      "step": 4314
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9175462505508767e-05,
+      "loss": 0.7484,
+      "step": 4315
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9174995996551997e-05,
+      "loss": 0.7457,
+      "step": 4316
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9174529361339217e-05,
+      "loss": 0.6234,
+      "step": 4317
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9174062599876854e-05,
+      "loss": 0.7097,
+      "step": 4318
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917359571217133e-05,
+      "loss": 0.6652,
+      "step": 4319
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9173128698229066e-05,
+      "loss": 0.7437,
+      "step": 4320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172661558056494e-05,
+      "loss": 0.6767,
+      "step": 4321
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172194291660037e-05,
+      "loss": 0.7442,
+      "step": 4322
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9171726899046126e-05,
+      "loss": 0.6593,
+      "step": 4323
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9171259380221196e-05,
+      "loss": 0.6216,
+      "step": 4324
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9170791735191677e-05,
+      "loss": 0.6411,
+      "step": 4325
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9170323963964007e-05,
+      "loss": 0.683,
+      "step": 4326
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916985606654462e-05,
+      "loss": 0.7541,
+      "step": 4327
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9169388042939956e-05,
+      "loss": 0.6368,
+      "step": 4328
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9168919893156457e-05,
+      "loss": 0.7025,
+      "step": 4329
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9168451617200566e-05,
+      "loss": 0.6635,
+      "step": 4330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916798321507872e-05,
+      "loss": 0.623,
+      "step": 4331
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916751468679737e-05,
+      "loss": 0.6782,
+      "step": 4332
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9167046032362964e-05,
+      "loss": 0.7314,
+      "step": 4333
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9166577251781947e-05,
+      "loss": 0.7113,
+      "step": 4334
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9166108345060777e-05,
+      "loss": 0.7621,
+      "step": 4335
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91656393122059e-05,
+      "loss": 0.7071,
+      "step": 4336
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9165170153223773e-05,
+      "loss": 0.7064,
+      "step": 4337
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9164700868120852e-05,
+      "loss": 0.7103,
+      "step": 4338
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9164231456903592e-05,
+      "loss": 0.7415,
+      "step": 4339
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163761919578457e-05,
+      "loss": 0.7751,
+      "step": 4340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163292256151902e-05,
+      "loss": 0.6759,
+      "step": 4341
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91628224666304e-05,
+      "loss": 0.7445,
+      "step": 4342
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9162352551020408e-05,
+      "loss": 0.7109,
+      "step": 4343
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916188250932839e-05,
+      "loss": 0.688,
+      "step": 4344
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9161412341560824e-05,
+      "loss": 0.7081,
+      "step": 4345
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916094204772417e-05,
+      "loss": 0.6088,
+      "step": 4346
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9160471627824904e-05,
+      "loss": 0.6848,
+      "step": 4347
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91600010818695e-05,
+      "loss": 0.6706,
+      "step": 4348
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9159530409864434e-05,
+      "loss": 0.7513,
+      "step": 4349
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915905961181618e-05,
+      "loss": 0.6984,
+      "step": 4350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9158588687731217e-05,
+      "loss": 0.7067,
+      "step": 4351
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915811763761603e-05,
+      "loss": 0.7212,
+      "step": 4352
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9157646461477096e-05,
+      "loss": 0.8104,
+      "step": 4353
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91571751593209e-05,
+      "loss": 0.7463,
+      "step": 4354
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9156703731153925e-05,
+      "loss": 0.7156,
+      "step": 4355
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915623217698266e-05,
+      "loss": 0.7165,
+      "step": 4356
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9155760496813597e-05,
+      "loss": 0.7207,
+      "step": 4357
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9155288690653226e-05,
+      "loss": 0.7433,
+      "step": 4358
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915481675850804e-05,
+      "loss": 0.7048,
+      "step": 4359
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154344700384525e-05,
+      "loss": 0.6304,
+      "step": 4360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9153872516289187e-05,
+      "loss": 0.7308,
+      "step": 4361
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915340020622852e-05,
+      "loss": 0.6894,
+      "step": 4362
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915292777020902e-05,
+      "loss": 0.7586,
+      "step": 4363
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9152455208237196e-05,
+      "loss": 0.5551,
+      "step": 4364
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9151982520319542e-05,
+      "loss": 0.7333,
+      "step": 4365
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915150970646257e-05,
+      "loss": 0.7127,
+      "step": 4366
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9151036766672782e-05,
+      "loss": 0.7261,
+      "step": 4367
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915056370095669e-05,
+      "loss": 0.8141,
+      "step": 4368
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9150090509320797e-05,
+      "loss": 0.7245,
+      "step": 4369
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914961719177162e-05,
+      "loss": 0.7455,
+      "step": 4370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914914374831567e-05,
+      "loss": 0.7284,
+      "step": 4371
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9148670178959464e-05,
+      "loss": 0.5965,
+      "step": 4372
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9148196483709518e-05,
+      "loss": 0.6627,
+      "step": 4373
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9147722662572353e-05,
+      "loss": 0.775,
+      "step": 4374
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914724871555448e-05,
+      "loss": 0.7656,
+      "step": 4375
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914677464266243e-05,
+      "loss": 0.592,
+      "step": 4376
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146300443902724e-05,
+      "loss": 0.7421,
+      "step": 4377
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914582611928189e-05,
+      "loss": 0.632,
+      "step": 4378
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914535166880645e-05,
+      "loss": 0.6917,
+      "step": 4379
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144877092482934e-05,
+      "loss": 0.7045,
+      "step": 4380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144402390317872e-05,
+      "loss": 0.713,
+      "step": 4381
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9143927562317803e-05,
+      "loss": 0.7511,
+      "step": 4382
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914345260848925e-05,
+      "loss": 0.6148,
+      "step": 4383
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142977528838763e-05,
+      "loss": 0.7414,
+      "step": 4384
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142502323372867e-05,
+      "loss": 0.6927,
+      "step": 4385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142026992098105e-05,
+      "loss": 0.7289,
+      "step": 4386
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914155153502102e-05,
+      "loss": 0.6278,
+      "step": 4387
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9141075952148154e-05,
+      "loss": 0.7008,
+      "step": 4388
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9140600243486052e-05,
+      "loss": 0.6163,
+      "step": 4389
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9140124409041255e-05,
+      "loss": 0.6473,
+      "step": 4390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913964844882032e-05,
+      "loss": 0.6921,
+      "step": 4391
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9139172362829786e-05,
+      "loss": 0.6587,
+      "step": 4392
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138696151076213e-05,
+      "loss": 0.6067,
+      "step": 4393
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913821981356615e-05,
+      "loss": 0.7466,
+      "step": 4394
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9137743350306155e-05,
+      "loss": 0.6434,
+      "step": 4395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913726676130278e-05,
+      "loss": 0.7137,
+      "step": 4396
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913679004656259e-05,
+      "loss": 0.7246,
+      "step": 4397
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9136313206092133e-05,
+      "loss": 0.753,
+      "step": 4398
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913583623989798e-05,
+      "loss": 0.7531,
+      "step": 4399
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91353591479867e-05,
+      "loss": 0.6947,
+      "step": 4400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9134881930364845e-05,
+      "loss": 0.6168,
+      "step": 4401
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913440458703899e-05,
+      "loss": 0.7227,
+      "step": 4402
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91339271180157e-05,
+      "loss": 0.7333,
+      "step": 4403
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913344952330155e-05,
+      "loss": 0.7602,
+      "step": 4404
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9132971802903105e-05,
+      "loss": 0.7343,
+      "step": 4405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9132493956826948e-05,
+      "loss": 0.6644,
+      "step": 4406
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9132015985079646e-05,
+      "loss": 0.7447,
+      "step": 4407
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913153788766778e-05,
+      "loss": 0.6897,
+      "step": 4408
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9131059664597932e-05,
+      "loss": 0.7026,
+      "step": 4409
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9130581315876676e-05,
+      "loss": 0.6511,
+      "step": 4410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91301028415106e-05,
+      "loss": 0.6352,
+      "step": 4411
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129624241506288e-05,
+      "loss": 0.7539,
+      "step": 4412
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129145515870325e-05,
+      "loss": 0.6552,
+      "step": 4413
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9128666664609293e-05,
+      "loss": 0.7413,
+      "step": 4414
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9128187687729793e-05,
+      "loss": 0.7009,
+      "step": 4415
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9127708585238406e-05,
+      "loss": 0.632,
+      "step": 4416
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9127229357141726e-05,
+      "loss": 0.6695,
+      "step": 4417
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9126750003446355e-05,
+      "loss": 0.697,
+      "step": 4418
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9126270524158883e-05,
+      "loss": 0.7674,
+      "step": 4419
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125790919285906e-05,
+      "loss": 0.6622,
+      "step": 4420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125311188834034e-05,
+      "loss": 0.5914,
+      "step": 4421
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912483133280986e-05,
+      "loss": 0.7524,
+      "step": 4422
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9124351351219984e-05,
+      "loss": 0.7099,
+      "step": 4423
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9123871244071018e-05,
+      "loss": 0.7463,
+      "step": 4424
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912339101136957e-05,
+      "loss": 0.6559,
+      "step": 4425
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9122910653122244e-05,
+      "loss": 0.6385,
+      "step": 4426
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912243016933565e-05,
+      "loss": 0.6988,
+      "step": 4427
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91219495600164e-05,
+      "loss": 0.6261,
+      "step": 4428
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9121468825171114e-05,
+      "loss": 0.6822,
+      "step": 4429
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9120987964806397e-05,
+      "loss": 0.7215,
+      "step": 4430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9120506978928873e-05,
+      "loss": 0.6883,
+      "step": 4431
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912002586754516e-05,
+      "loss": 0.728,
+      "step": 4432
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9119544630661877e-05,
+      "loss": 0.681,
+      "step": 4433
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9119063268285645e-05,
+      "loss": 0.6554,
+      "step": 4434
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911858178042309e-05,
+      "loss": 0.6297,
+      "step": 4435
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911810016708084e-05,
+      "loss": 0.6347,
+      "step": 4436
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911761842826552e-05,
+      "loss": 0.7585,
+      "step": 4437
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9117136563983754e-05,
+      "loss": 0.7418,
+      "step": 4438
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9116654574242185e-05,
+      "loss": 0.7324,
+      "step": 4439
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9116172459047434e-05,
+      "loss": 0.7472,
+      "step": 4440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911569021840614e-05,
+      "loss": 0.7543,
+      "step": 4441
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911520785232494e-05,
+      "loss": 0.7349,
+      "step": 4442
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911472536081047e-05,
+      "loss": 0.6315,
+      "step": 4443
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9114242743869368e-05,
+      "loss": 0.6948,
+      "step": 4444
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911376000150828e-05,
+      "loss": 0.5885,
+      "step": 4445
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9113277133733847e-05,
+      "loss": 0.6852,
+      "step": 4446
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911279414055271e-05,
+      "loss": 0.6064,
+      "step": 4447
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911231102197152e-05,
+      "loss": 0.7658,
+      "step": 4448
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9111827777996926e-05,
+      "loss": 0.744,
+      "step": 4449
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9111344408635573e-05,
+      "loss": 0.6778,
+      "step": 4450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110860913894114e-05,
+      "loss": 0.7203,
+      "step": 4451
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110377293779205e-05,
+      "loss": 0.7704,
+      "step": 4452
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91098935482975e-05,
+      "loss": 0.6935,
+      "step": 4453
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9109409677455652e-05,
+      "loss": 0.6415,
+      "step": 4454
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9108925681260327e-05,
+      "loss": 0.6083,
+      "step": 4455
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910844155971818e-05,
+      "loss": 0.6176,
+      "step": 4456
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910795731283587e-05,
+      "loss": 0.7678,
+      "step": 4457
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9107472940620065e-05,
+      "loss": 0.6136,
+      "step": 4458
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910698844307743e-05,
+      "loss": 0.6999,
+      "step": 4459
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106503820214636e-05,
+      "loss": 0.5997,
+      "step": 4460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106019072038345e-05,
+      "loss": 0.7151,
+      "step": 4461
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910553419855523e-05,
+      "loss": 0.5795,
+      "step": 4462
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9105049199771963e-05,
+      "loss": 0.74,
+      "step": 4463
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910456407569522e-05,
+      "loss": 0.8002,
+      "step": 4464
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9104078826331674e-05,
+      "loss": 0.565,
+      "step": 4465
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9103593451688005e-05,
+      "loss": 0.7278,
+      "step": 4466
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910310795177089e-05,
+      "loss": 0.7468,
+      "step": 4467
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102622326587013e-05,
+      "loss": 0.7698,
+      "step": 4468
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102136576143052e-05,
+      "loss": 0.6972,
+      "step": 4469
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9101650700445697e-05,
+      "loss": 0.6618,
+      "step": 4470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9101164699501628e-05,
+      "loss": 0.6539,
+      "step": 4471
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9100678573317538e-05,
+      "loss": 0.7,
+      "step": 4472
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910019232190011e-05,
+      "loss": 0.6058,
+      "step": 4473
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9099705945256043e-05,
+      "loss": 0.7116,
+      "step": 4474
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9099219443392028e-05,
+      "loss": 0.7379,
+      "step": 4475
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098732816314757e-05,
+      "loss": 0.6147,
+      "step": 4476
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098246064030926e-05,
+      "loss": 0.7115,
+      "step": 4477
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097759186547236e-05,
+      "loss": 0.7257,
+      "step": 4478
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097272183870386e-05,
+      "loss": 0.674,
+      "step": 4479
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096785056007074e-05,
+      "loss": 0.7004,
+      "step": 4480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909629780296401e-05,
+      "loss": 0.74,
+      "step": 4481
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095810424747895e-05,
+      "loss": 0.6871,
+      "step": 4482
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095322921365435e-05,
+      "loss": 0.6851,
+      "step": 4483
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909483529282334e-05,
+      "loss": 0.7662,
+      "step": 4484
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909434753912832e-05,
+      "loss": 0.6796,
+      "step": 4485
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9093859660287087e-05,
+      "loss": 0.7374,
+      "step": 4486
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9093371656306355e-05,
+      "loss": 0.6951,
+      "step": 4487
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9092883527192837e-05,
+      "loss": 0.7009,
+      "step": 4488
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9092395272953257e-05,
+      "loss": 0.7756,
+      "step": 4489
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9091906893594324e-05,
+      "loss": 0.7097,
+      "step": 4490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9091418389122766e-05,
+      "loss": 0.726,
+      "step": 4491
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90909297595453e-05,
+      "loss": 0.6121,
+      "step": 4492
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9090441004868655e-05,
+      "loss": 0.7196,
+      "step": 4493
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908995212509955e-05,
+      "loss": 0.7111,
+      "step": 4494
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908946312024472e-05,
+      "loss": 0.5922,
+      "step": 4495
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9088973990310894e-05,
+      "loss": 0.6133,
+      "step": 4496
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9088484735304793e-05,
+      "loss": 0.5696,
+      "step": 4497
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9087995355233157e-05,
+      "loss": 0.6807,
+      "step": 4498
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9087505850102724e-05,
+      "loss": 0.666,
+      "step": 4499
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9087016219920222e-05,
+      "loss": 0.5999,
+      "step": 4500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9086526464692394e-05,
+      "loss": 0.6844,
+      "step": 4501
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908603658442598e-05,
+      "loss": 0.7002,
+      "step": 4502
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9085546579127713e-05,
+      "loss": 0.607,
+      "step": 4503
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9085056448804345e-05,
+      "loss": 0.7041,
+      "step": 4504
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9084566193462617e-05,
+      "loss": 0.7004,
+      "step": 4505
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9084075813109275e-05,
+      "loss": 0.6906,
+      "step": 4506
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9083585307751068e-05,
+      "loss": 0.6822,
+      "step": 4507
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908309467739475e-05,
+      "loss": 0.7396,
+      "step": 4508
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9082603922047063e-05,
+      "loss": 0.6998,
+      "step": 4509
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9082113041714767e-05,
+      "loss": 0.5672,
+      "step": 4510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9081622036404616e-05,
+      "loss": 0.7216,
+      "step": 4511
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9081130906123366e-05,
+      "loss": 0.6898,
+      "step": 4512
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9080639650877774e-05,
+      "loss": 0.7036,
+      "step": 4513
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90801482706746e-05,
+      "loss": 0.6249,
+      "step": 4514
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907965676552061e-05,
+      "loss": 0.6866,
+      "step": 4515
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9079165135422567e-05,
+      "loss": 0.6267,
+      "step": 4516
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907867338038723e-05,
+      "loss": 0.7283,
+      "step": 4517
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907818150042137e-05,
+      "loss": 0.6668,
+      "step": 4518
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9077689495531755e-05,
+      "loss": 0.6002,
+      "step": 4519
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9077197365725163e-05,
+      "loss": 0.7077,
+      "step": 4520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907670511100835e-05,
+      "loss": 0.7623,
+      "step": 4521
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9076212731388106e-05,
+      "loss": 0.6457,
+      "step": 4522
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90757202268712e-05,
+      "loss": 0.6946,
+      "step": 4523
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9075227597464404e-05,
+      "loss": 0.7623,
+      "step": 4524
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9074734843174504e-05,
+      "loss": 0.6439,
+      "step": 4525
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907424196400828e-05,
+      "loss": 0.6005,
+      "step": 4526
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9073748959972514e-05,
+      "loss": 0.7019,
+      "step": 4527
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9073255831073985e-05,
+      "loss": 0.7275,
+      "step": 4528
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9072762577319487e-05,
+      "loss": 0.7367,
+      "step": 4529
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9072269198715805e-05,
+      "loss": 0.7506,
+      "step": 4530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9071775695269726e-05,
+      "loss": 0.6029,
+      "step": 4531
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907128206698804e-05,
+      "loss": 0.6689,
+      "step": 4532
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9070788313877546e-05,
+      "loss": 0.681,
+      "step": 4533
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9070294435945032e-05,
+      "loss": 0.7212,
+      "step": 4534
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9069800433197296e-05,
+      "loss": 0.7655,
+      "step": 4535
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906930630564114e-05,
+      "loss": 0.7693,
+      "step": 4536
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9068812053283355e-05,
+      "loss": 0.717,
+      "step": 4537
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9068317676130753e-05,
+      "loss": 0.6484,
+      "step": 4538
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9067823174190127e-05,
+      "loss": 0.7253,
+      "step": 4539
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906732854746829e-05,
+      "loss": 0.6855,
+      "step": 4540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906683379597204e-05,
+      "loss": 0.7493,
+      "step": 4541
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90663389197082e-05,
+      "loss": 0.6928,
+      "step": 4542
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065843918683563e-05,
+      "loss": 0.6737,
+      "step": 4543
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065348792904946e-05,
+      "loss": 0.7579,
+      "step": 4544
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9064853542379166e-05,
+      "loss": 0.6294,
+      "step": 4545
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9064358167113036e-05,
+      "loss": 0.6109,
+      "step": 4546
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9063862667113376e-05,
+      "loss": 0.692,
+      "step": 4547
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9063367042387e-05,
+      "loss": 0.7111,
+      "step": 4548
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9062871292940728e-05,
+      "loss": 0.6959,
+      "step": 4549
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9062375418781384e-05,
+      "loss": 0.6429,
+      "step": 4550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906187941991579e-05,
+      "loss": 0.6051,
+      "step": 4551
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9061383296350777e-05,
+      "loss": 0.6991,
+      "step": 4552
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9060887048093166e-05,
+      "loss": 0.6788,
+      "step": 4553
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9060390675149787e-05,
+      "loss": 0.6468,
+      "step": 4554
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.905989417752747e-05,
+      "loss": 0.7175,
+      "step": 4555
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905939755523305e-05,
+      "loss": 0.727,
+      "step": 4556
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9058900808273358e-05,
+      "loss": 0.6814,
+      "step": 4557
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9058403936655235e-05,
+      "loss": 0.7288,
+      "step": 4558
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9057906940385508e-05,
+      "loss": 0.761,
+      "step": 4559
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905740981947103e-05,
+      "loss": 0.6256,
+      "step": 4560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905691257391863e-05,
+      "loss": 0.7608,
+      "step": 4561
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905641520373515e-05,
+      "loss": 0.7026,
+      "step": 4562
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905591770892745e-05,
+      "loss": 0.6585,
+      "step": 4563
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905542008950236e-05,
+      "loss": 0.6978,
+      "step": 4564
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905492234546673e-05,
+      "loss": 0.7468,
+      "step": 4565
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9054424476827417e-05,
+      "loss": 0.7599,
+      "step": 4566
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905392648359126e-05,
+      "loss": 0.6078,
+      "step": 4567
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9053428365765126e-05,
+      "loss": 0.7502,
+      "step": 4568
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9052930123355858e-05,
+      "loss": 0.6651,
+      "step": 4569
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905243175637032e-05,
+      "loss": 0.5784,
+      "step": 4570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9051933264815368e-05,
+      "loss": 0.6889,
+      "step": 4571
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9051434648697857e-05,
+      "loss": 0.6893,
+      "step": 4572
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9050935908024656e-05,
+      "loss": 0.6166,
+      "step": 4573
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9050437042802622e-05,
+      "loss": 0.6284,
+      "step": 4574
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904993805303862e-05,
+      "loss": 0.7135,
+      "step": 4575
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9049438938739522e-05,
+      "loss": 0.7154,
+      "step": 4576
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904893969991219e-05,
+      "loss": 0.7277,
+      "step": 4577
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9048440336563495e-05,
+      "loss": 0.7284,
+      "step": 4578
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9047940848700315e-05,
+      "loss": 0.7026,
+      "step": 4579
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9047441236329516e-05,
+      "loss": 0.6871,
+      "step": 4580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904694149945798e-05,
+      "loss": 0.6912,
+      "step": 4581
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9046441638092572e-05,
+      "loss": 0.7687,
+      "step": 4582
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9045941652240183e-05,
+      "loss": 0.5756,
+      "step": 4583
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9045441541907686e-05,
+      "loss": 0.6675,
+      "step": 4584
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9044941307101966e-05,
+      "loss": 0.752,
+      "step": 4585
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9044440947829908e-05,
+      "loss": 0.6284,
+      "step": 4586
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9043940464098397e-05,
+      "loss": 0.6115,
+      "step": 4587
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9043439855914313e-05,
+      "loss": 0.5952,
+      "step": 4588
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9042939123284555e-05,
+      "loss": 0.5932,
+      "step": 4589
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9042438266216004e-05,
+      "loss": 0.7234,
+      "step": 4590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9041937284715562e-05,
+      "loss": 0.6029,
+      "step": 4591
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9041436178790117e-05,
+      "loss": 0.6729,
+      "step": 4592
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9040934948446564e-05,
+      "loss": 0.7001,
+      "step": 4593
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90404335936918e-05,
+      "loss": 0.7166,
+      "step": 4594
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903993211453273e-05,
+      "loss": 0.5789,
+      "step": 4595
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9039430510976252e-05,
+      "loss": 0.5839,
+      "step": 4596
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9038928783029266e-05,
+      "loss": 0.7834,
+      "step": 4597
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9038426930698676e-05,
+      "loss": 0.6786,
+      "step": 4598
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903792495399139e-05,
+      "loss": 0.627,
+      "step": 4599
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9037422852914316e-05,
+      "loss": 0.6316,
+      "step": 4600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9036920627474362e-05,
+      "loss": 0.6238,
+      "step": 4601
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903641827767844e-05,
+      "loss": 0.7718,
+      "step": 4602
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903591580353346e-05,
+      "loss": 0.7039,
+      "step": 4603
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9035413205046344e-05,
+      "loss": 0.6843,
+      "step": 4604
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034910482224e-05,
+      "loss": 0.593,
+      "step": 4605
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034407635073348e-05,
+      "loss": 0.6317,
+      "step": 4606
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903390466360131e-05,
+      "loss": 0.6804,
+      "step": 4607
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9033401567814807e-05,
+      "loss": 0.6549,
+      "step": 4608
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903289834772076e-05,
+      "loss": 0.7881,
+      "step": 4609
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9032395003326096e-05,
+      "loss": 0.6818,
+      "step": 4610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9031891534637735e-05,
+      "loss": 0.6383,
+      "step": 4611
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9031387941662614e-05,
+      "loss": 0.6828,
+      "step": 4612
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030884224407658e-05,
+      "loss": 0.7521,
+      "step": 4613
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030380382879806e-05,
+      "loss": 0.7147,
+      "step": 4614
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902987641708598e-05,
+      "loss": 0.6543,
+      "step": 4615
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902937232703312e-05,
+      "loss": 0.688,
+      "step": 4616
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902886811272816e-05,
+      "loss": 0.6592,
+      "step": 4617
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9028363774178045e-05,
+      "loss": 0.7544,
+      "step": 4618
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902785931138971e-05,
+      "loss": 0.6355,
+      "step": 4619
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9027354724370102e-05,
+      "loss": 0.6421,
+      "step": 4620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902685001312616e-05,
+      "loss": 0.7186,
+      "step": 4621
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9026345177664826e-05,
+      "loss": 0.7333,
+      "step": 4622
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9025840217993056e-05,
+      "loss": 0.6976,
+      "step": 4623
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9025335134117792e-05,
+      "loss": 0.6394,
+      "step": 4624
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024829926045984e-05,
+      "loss": 0.6379,
+      "step": 4625
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024324593784586e-05,
+      "loss": 0.7412,
+      "step": 4626
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9023819137340552e-05,
+      "loss": 0.6983,
+      "step": 4627
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902331355672084e-05,
+      "loss": 0.726,
+      "step": 4628
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902280785193241e-05,
+      "loss": 0.619,
+      "step": 4629
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9022302022982207e-05,
+      "loss": 0.6047,
+      "step": 4630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9021796069877207e-05,
+      "loss": 0.6524,
+      "step": 4631
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902128999262436e-05,
+      "loss": 0.6566,
+      "step": 4632
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9020783791230636e-05,
+      "loss": 0.5854,
+      "step": 4633
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9020277465703006e-05,
+      "loss": 0.7101,
+      "step": 4634
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901977101604843e-05,
+      "loss": 0.7273,
+      "step": 4635
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901926444227388e-05,
+      "loss": 0.6636,
+      "step": 4636
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9018757744386324e-05,
+      "loss": 0.6031,
+      "step": 4637
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9018250922392737e-05,
+      "loss": 0.554,
+      "step": 4638
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9017743976300092e-05,
+      "loss": 0.6361,
+      "step": 4639
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901723690611537e-05,
+      "loss": 0.7373,
+      "step": 4640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9016729711845543e-05,
+      "loss": 0.6639,
+      "step": 4641
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901622239349759e-05,
+      "loss": 0.6476,
+      "step": 4642
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90157149510785e-05,
+      "loss": 0.7134,
+      "step": 4643
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901520738459525e-05,
+      "loss": 0.7043,
+      "step": 4644
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901469969405482e-05,
+      "loss": 0.7274,
+      "step": 4645
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9014191879464203e-05,
+      "loss": 0.7289,
+      "step": 4646
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9013683940830385e-05,
+      "loss": 0.6301,
+      "step": 4647
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9013175878160355e-05,
+      "loss": 0.7172,
+      "step": 4648
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901266769146111e-05,
+      "loss": 0.7483,
+      "step": 4649
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012159380739635e-05,
+      "loss": 0.6399,
+      "step": 4650
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901165094600293e-05,
+      "loss": 0.7387,
+      "step": 4651
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9011142387257985e-05,
+      "loss": 0.7408,
+      "step": 4652
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9010633704511808e-05,
+      "loss": 0.7066,
+      "step": 4653
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901012489777139e-05,
+      "loss": 0.7168,
+      "step": 4654
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9009615967043736e-05,
+      "loss": 0.6317,
+      "step": 4655
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9009106912335854e-05,
+      "loss": 0.7543,
+      "step": 4656
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9008597733654743e-05,
+      "loss": 0.6915,
+      "step": 4657
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900808843100741e-05,
+      "loss": 0.5929,
+      "step": 4658
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9007579004400868e-05,
+      "loss": 0.7441,
+      "step": 4659
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900706945384212e-05,
+      "loss": 0.7027,
+      "step": 4660
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006559779338186e-05,
+      "loss": 0.7278,
+      "step": 4661
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006049980896074e-05,
+      "loss": 0.7074,
+      "step": 4662
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90055400585228e-05,
+      "loss": 0.7171,
+      "step": 4663
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9005030012225383e-05,
+      "loss": 0.6269,
+      "step": 4664
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900451984201084e-05,
+      "loss": 0.5883,
+      "step": 4665
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9004009547886193e-05,
+      "loss": 0.6596,
+      "step": 4666
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900349912985846e-05,
+      "loss": 0.701,
+      "step": 4667
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900298858793467e-05,
+      "loss": 0.6538,
+      "step": 4668
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9002477922121846e-05,
+      "loss": 0.7185,
+      "step": 4669
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001967132427017e-05,
+      "loss": 0.6697,
+      "step": 4670
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001456218857207e-05,
+      "loss": 0.607,
+      "step": 4671
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900094518141945e-05,
+      "loss": 0.7232,
+      "step": 4672
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9000434020120785e-05,
+      "loss": 0.7053,
+      "step": 4673
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8999922734968236e-05,
+      "loss": 0.6024,
+      "step": 4674
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899941132596884e-05,
+      "loss": 0.6163,
+      "step": 4675
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998899793129637e-05,
+      "loss": 0.73,
+      "step": 4676
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998388136457667e-05,
+      "loss": 0.6638,
+      "step": 4677
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899787635595997e-05,
+      "loss": 0.7053,
+      "step": 4678
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8997364451643585e-05,
+      "loss": 0.6619,
+      "step": 4679
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8996852423515562e-05,
+      "loss": 0.7912,
+      "step": 4680
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8996340271582943e-05,
+      "loss": 0.6829,
+      "step": 4681
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899582799585278e-05,
+      "loss": 0.611,
+      "step": 4682
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8995315596332115e-05,
+      "loss": 0.7001,
+      "step": 4683
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8994803073028004e-05,
+      "loss": 0.6315,
+      "step": 4684
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89942904259475e-05,
+      "loss": 0.6646,
+      "step": 4685
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8993777655097658e-05,
+      "loss": 0.6894,
+      "step": 4686
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8993264760485535e-05,
+      "loss": 0.5974,
+      "step": 4687
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8992751742118184e-05,
+      "loss": 0.7213,
+      "step": 4688
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8992238600002668e-05,
+      "loss": 0.6551,
+      "step": 4689
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8991725334146046e-05,
+      "loss": 0.6589,
+      "step": 4690
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899121194455538e-05,
+      "loss": 0.7171,
+      "step": 4691
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8990698431237744e-05,
+      "loss": 0.6549,
+      "step": 4692
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899018479420019e-05,
+      "loss": 0.7542,
+      "step": 4693
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89896710334498e-05,
+      "loss": 0.6398,
+      "step": 4694
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8989157148993634e-05,
+      "loss": 0.7309,
+      "step": 4695
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898864314083877e-05,
+      "loss": 0.6969,
+      "step": 4696
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8988129008992276e-05,
+      "loss": 0.7053,
+      "step": 4697
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8987614753461234e-05,
+      "loss": 0.6975,
+      "step": 4698
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898710037425271e-05,
+      "loss": 0.6152,
+      "step": 4699
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8986585871373792e-05,
+      "loss": 0.6125,
+      "step": 4700
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898607124483155e-05,
+      "loss": 0.6598,
+      "step": 4701
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8985556494633078e-05,
+      "loss": 0.6432,
+      "step": 4702
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898504162078545e-05,
+      "loss": 0.6948,
+      "step": 4703
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898452662329576e-05,
+      "loss": 0.6782,
+      "step": 4704
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8984011502171082e-05,
+      "loss": 0.662,
+      "step": 4705
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8983496257418515e-05,
+      "loss": 0.7142,
+      "step": 4706
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8982980889045147e-05,
+      "loss": 0.7274,
+      "step": 4707
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8982465397058065e-05,
+      "loss": 0.7481,
+      "step": 4708
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8981949781464367e-05,
+      "loss": 0.762,
+      "step": 4709
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8981434042271153e-05,
+      "loss": 0.6875,
+      "step": 4710
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8980918179485508e-05,
+      "loss": 0.6747,
+      "step": 4711
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898040219311454e-05,
+      "loss": 0.6422,
+      "step": 4712
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897988608316535e-05,
+      "loss": 0.6944,
+      "step": 4713
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8979369849645036e-05,
+      "loss": 0.7651,
+      "step": 4714
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89788534925607e-05,
+      "loss": 0.7025,
+      "step": 4715
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897833701191945e-05,
+      "loss": 0.6301,
+      "step": 4716
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977820407728397e-05,
+      "loss": 0.7379,
+      "step": 4717
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977303679994645e-05,
+      "loss": 0.6968,
+      "step": 4718
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8976786828725305e-05,
+      "loss": 0.704,
+      "step": 4719
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8976269853927492e-05,
+      "loss": 0.7113,
+      "step": 4720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897575275560832e-05,
+      "loss": 0.6099,
+      "step": 4721
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89752355337749e-05,
+      "loss": 0.75,
+      "step": 4722
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974718188434357e-05,
+      "loss": 0.7441,
+      "step": 4723
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974200719593803e-05,
+      "loss": 0.6569,
+      "step": 4724
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8973683127260362e-05,
+      "loss": 0.7366,
+      "step": 4725
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8973165411441156e-05,
+      "loss": 0.7,
+      "step": 4726
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8972647572143312e-05,
+      "loss": 0.6865,
+      "step": 4727
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897212960937395e-05,
+      "loss": 0.6419,
+      "step": 4728
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8971611523140203e-05,
+      "loss": 0.7129,
+      "step": 4729
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8971093313449197e-05,
+      "loss": 0.64,
+      "step": 4730
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8970574980308062e-05,
+      "loss": 0.6532,
+      "step": 4731
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8970056523723937e-05,
+      "loss": 0.7039,
+      "step": 4732
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8969537943703953e-05,
+      "loss": 0.6252,
+      "step": 4733
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8969019240255244e-05,
+      "loss": 0.7022,
+      "step": 4734
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8968500413384946e-05,
+      "loss": 0.7259,
+      "step": 4735
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8967981463100207e-05,
+      "loss": 0.6229,
+      "step": 4736
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896746238940816e-05,
+      "loss": 0.6036,
+      "step": 4737
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896694319231595e-05,
+      "loss": 0.7089,
+      "step": 4738
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896642387183073e-05,
+      "loss": 0.6796,
+      "step": 4739
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896590442795963e-05,
+      "loss": 0.6918,
+      "step": 4740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896538486070981e-05,
+      "loss": 0.6285,
+      "step": 4741
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896486517008842e-05,
+      "loss": 0.666,
+      "step": 4742
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8964345356102606e-05,
+      "loss": 0.705,
+      "step": 4743
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8963825418759524e-05,
+      "loss": 0.6989,
+      "step": 4744
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8963305358066325e-05,
+      "loss": 0.6515,
+      "step": 4745
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8962785174030166e-05,
+      "loss": 0.6381,
+      "step": 4746
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8962264866658214e-05,
+      "loss": 0.7339,
+      "step": 4747
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8961744435957618e-05,
+      "loss": 0.6776,
+      "step": 4748
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896122388193555e-05,
+      "loss": 0.7079,
+      "step": 4749
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896070320459916e-05,
+      "loss": 0.634,
+      "step": 4750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8960182403955626e-05,
+      "loss": 0.6667,
+      "step": 4751
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8959661480012105e-05,
+      "loss": 0.6866,
+      "step": 4752
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8959140432775772e-05,
+      "loss": 0.5828,
+      "step": 4753
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895861926225379e-05,
+      "loss": 0.8079,
+      "step": 4754
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895809796845334e-05,
+      "loss": 0.5929,
+      "step": 4755
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895757655138159e-05,
+      "loss": 0.7204,
+      "step": 4756
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8957055011045718e-05,
+      "loss": 0.7078,
+      "step": 4757
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8956533347452895e-05,
+      "loss": 0.6586,
+      "step": 4758
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8956011560610304e-05,
+      "loss": 0.7095,
+      "step": 4759
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8955489650525124e-05,
+      "loss": 0.6487,
+      "step": 4760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895496761720454e-05,
+      "loss": 0.6812,
+      "step": 4761
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895444546065573e-05,
+      "loss": 0.734,
+      "step": 4762
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8953923180885883e-05,
+      "loss": 0.6532,
+      "step": 4763
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8953400777902188e-05,
+      "loss": 0.7221,
+      "step": 4764
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8952878251711828e-05,
+      "loss": 0.7768,
+      "step": 4765
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8952355602322e-05,
+      "loss": 0.7395,
+      "step": 4766
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8951832829739887e-05,
+      "loss": 0.7235,
+      "step": 4767
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8951309933972694e-05,
+      "loss": 0.6679,
+      "step": 4768
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895078691502761e-05,
+      "loss": 0.6834,
+      "step": 4769
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895026377291183e-05,
+      "loss": 0.7579,
+      "step": 4770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894974050763256e-05,
+      "loss": 0.7059,
+      "step": 4771
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8949217119196995e-05,
+      "loss": 0.6107,
+      "step": 4772
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8948693607612337e-05,
+      "loss": 0.7102,
+      "step": 4773
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8948169972885795e-05,
+      "loss": 0.7686,
+      "step": 4774
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8947646215024573e-05,
+      "loss": 0.6364,
+      "step": 4775
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894712233403587e-05,
+      "loss": 0.7377,
+      "step": 4776
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894659832992691e-05,
+      "loss": 0.7679,
+      "step": 4777
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8946074202704888e-05,
+      "loss": 0.7082,
+      "step": 4778
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894554995237703e-05,
+      "loss": 0.7168,
+      "step": 4779
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8945025578950544e-05,
+      "loss": 0.6561,
+      "step": 4780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8944501082432644e-05,
+      "loss": 0.5986,
+      "step": 4781
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894397646283055e-05,
+      "loss": 0.6667,
+      "step": 4782
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8943451720151486e-05,
+      "loss": 0.6871,
+      "step": 4783
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894292685440266e-05,
+      "loss": 0.6778,
+      "step": 4784
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894240186559131e-05,
+      "loss": 0.7044,
+      "step": 4785
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8941876753724648e-05,
+      "loss": 0.6821,
+      "step": 4786
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8941351518809903e-05,
+      "loss": 0.6399,
+      "step": 4787
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894082616085431e-05,
+      "loss": 0.76,
+      "step": 4788
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8940300679865092e-05,
+      "loss": 0.6718,
+      "step": 4789
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893977507584948e-05,
+      "loss": 0.6415,
+      "step": 4790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8939249348814704e-05,
+      "loss": 0.7027,
+      "step": 4791
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8938723498768007e-05,
+      "loss": 0.7015,
+      "step": 4792
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893819752571662e-05,
+      "loss": 0.6255,
+      "step": 4793
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893767142966778e-05,
+      "loss": 0.6118,
+      "step": 4794
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8937145210628727e-05,
+      "loss": 0.7371,
+      "step": 4795
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8936618868606707e-05,
+      "loss": 0.7937,
+      "step": 4796
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8936092403608954e-05,
+      "loss": 0.6821,
+      "step": 4797
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893556581564272e-05,
+      "loss": 0.6733,
+      "step": 4798
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893503910471525e-05,
+      "loss": 0.7214,
+      "step": 4799
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8934512270833787e-05,
+      "loss": 0.7512,
+      "step": 4800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8933985314005587e-05,
+      "loss": 0.7863,
+      "step": 4801
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89334582342379e-05,
+      "loss": 0.7178,
+      "step": 4802
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893293103153798e-05,
+      "loss": 0.6045,
+      "step": 4803
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8932403705913077e-05,
+      "loss": 0.6241,
+      "step": 4804
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8931876257370448e-05,
+      "loss": 0.6772,
+      "step": 4805
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8931348685917354e-05,
+      "loss": 0.7091,
+      "step": 4806
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8930820991561057e-05,
+      "loss": 0.7089,
+      "step": 4807
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8930293174308815e-05,
+      "loss": 0.6056,
+      "step": 4808
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8929765234167893e-05,
+      "loss": 0.6877,
+      "step": 4809
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8929237171145554e-05,
+      "loss": 0.7368,
+      "step": 4810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8928708985249065e-05,
+      "loss": 0.595,
+      "step": 4811
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8928180676485697e-05,
+      "loss": 0.7132,
+      "step": 4812
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8927652244862718e-05,
+      "loss": 0.7782,
+      "step": 4813
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8927123690387397e-05,
+      "loss": 0.7605,
+      "step": 4814
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892659501306701e-05,
+      "loss": 0.5952,
+      "step": 4815
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8926066212908835e-05,
+      "loss": 0.6806,
+      "step": 4816
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892553728992014e-05,
+      "loss": 0.641,
+      "step": 4817
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8925008244108215e-05,
+      "loss": 0.6955,
+      "step": 4818
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892447907548033e-05,
+      "loss": 0.6176,
+      "step": 4819
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8923949784043777e-05,
+      "loss": 0.7422,
+      "step": 4820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892342036980583e-05,
+      "loss": 0.6453,
+      "step": 4821
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892289083277378e-05,
+      "loss": 0.5883,
+      "step": 4822
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892236117295491e-05,
+      "loss": 0.7167,
+      "step": 4823
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892183139035651e-05,
+      "loss": 0.6791,
+      "step": 4824
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892130148498587e-05,
+      "loss": 0.6452,
+      "step": 4825
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892077145685028e-05,
+      "loss": 0.7093,
+      "step": 4826
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8920241305957043e-05,
+      "loss": 0.6907,
+      "step": 4827
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919711032313444e-05,
+      "loss": 0.6185,
+      "step": 4828
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919180635926782e-05,
+      "loss": 0.7159,
+      "step": 4829
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891865011680436e-05,
+      "loss": 0.6838,
+      "step": 4830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8918119474953477e-05,
+      "loss": 0.6488,
+      "step": 4831
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8917588710381434e-05,
+      "loss": 0.6793,
+      "step": 4832
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8917057823095533e-05,
+      "loss": 0.6247,
+      "step": 4833
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891652681310308e-05,
+      "loss": 0.7122,
+      "step": 4834
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8915995680411385e-05,
+      "loss": 0.7189,
+      "step": 4835
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8915464425027753e-05,
+      "loss": 0.6242,
+      "step": 4836
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89149330469595e-05,
+      "loss": 0.7356,
+      "step": 4837
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8914401546213935e-05,
+      "loss": 0.7069,
+      "step": 4838
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8913869922798368e-05,
+      "loss": 0.646,
+      "step": 4839
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891333817672012e-05,
+      "loss": 0.6667,
+      "step": 4840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891280630798651e-05,
+      "loss": 0.7299,
+      "step": 4841
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891227431660485e-05,
+      "loss": 0.7402,
+      "step": 4842
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8911742202582463e-05,
+      "loss": 0.7658,
+      "step": 4843
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8911209965926677e-05,
+      "loss": 0.6543,
+      "step": 4844
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891067760664481e-05,
+      "loss": 0.7679,
+      "step": 4845
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891014512474419e-05,
+      "loss": 0.7085,
+      "step": 4846
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890961252023214e-05,
+      "loss": 0.5965,
+      "step": 4847
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8909079793115998e-05,
+      "loss": 0.6214,
+      "step": 4848
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8908546943403087e-05,
+      "loss": 0.6856,
+      "step": 4849
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8908013971100744e-05,
+      "loss": 0.5994,
+      "step": 4850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89074808762163e-05,
+      "loss": 0.5903,
+      "step": 4851
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8906947658757095e-05,
+      "loss": 0.6172,
+      "step": 4852
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8906414318730462e-05,
+      "loss": 0.7249,
+      "step": 4853
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905880856143742e-05,
+      "loss": 0.7584,
+      "step": 4854
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905347271004277e-05,
+      "loss": 0.6906,
+      "step": 4855
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890481356331941e-05,
+      "loss": 0.7501,
+      "step": 4856
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890427973309648e-05,
+      "loss": 0.7319,
+      "step": 4857
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903745780342838e-05,
+      "loss": 0.5878,
+      "step": 4858
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903211705065833e-05,
+      "loss": 0.7436,
+      "step": 4859
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890267750727281e-05,
+      "loss": 0.7169,
+      "step": 4860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8902143186971123e-05,
+      "loss": 0.6664,
+      "step": 4861
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890160874416812e-05,
+      "loss": 0.5479,
+      "step": 4862
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8901074178871163e-05,
+      "loss": 0.5799,
+      "step": 4863
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8900539491087605e-05,
+      "loss": 0.6873,
+      "step": 4864
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8900004680824798e-05,
+      "loss": 0.6661,
+      "step": 4865
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889946974809011e-05,
+      "loss": 0.6374,
+      "step": 4866
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8898934692890897e-05,
+      "loss": 0.5825,
+      "step": 4867
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8898399515234525e-05,
+      "loss": 0.6446,
+      "step": 4868
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8897864215128356e-05,
+      "loss": 0.7662,
+      "step": 4869
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889732879257976e-05,
+      "loss": 0.6283,
+      "step": 4870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88967932475961e-05,
+      "loss": 0.7644,
+      "step": 4871
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8896257580184744e-05,
+      "loss": 0.7272,
+      "step": 4872
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889572179035307e-05,
+      "loss": 0.5971,
+      "step": 4873
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8895185878108445e-05,
+      "loss": 0.7149,
+      "step": 4874
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889464984345825e-05,
+      "loss": 0.622,
+      "step": 4875
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8894113686409854e-05,
+      "loss": 0.7225,
+      "step": 4876
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8893577406970642e-05,
+      "loss": 0.7574,
+      "step": 4877
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889304100514799e-05,
+      "loss": 0.7116,
+      "step": 4878
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8892504480949275e-05,
+      "loss": 0.7127,
+      "step": 4879
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889196783438189e-05,
+      "loss": 0.6585,
+      "step": 4880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8891431065453216e-05,
+      "loss": 0.6817,
+      "step": 4881
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889089417417063e-05,
+      "loss": 0.75,
+      "step": 4882
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8890357160541533e-05,
+      "loss": 0.7708,
+      "step": 4883
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8889820024573308e-05,
+      "loss": 0.6164,
+      "step": 4884
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888928276627335e-05,
+      "loss": 0.6745,
+      "step": 4885
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888874538564905e-05,
+      "loss": 0.6616,
+      "step": 4886
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8888207882707802e-05,
+      "loss": 0.6264,
+      "step": 4887
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8887670257457006e-05,
+      "loss": 0.772,
+      "step": 4888
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8887132509904054e-05,
+      "loss": 0.7373,
+      "step": 4889
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8886594640056354e-05,
+      "loss": 0.7179,
+      "step": 4890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88860566479213e-05,
+      "loss": 0.7081,
+      "step": 4891
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88855185335063e-05,
+      "loss": 0.7317,
+      "step": 4892
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888498029681876e-05,
+      "loss": 0.5935,
+      "step": 4893
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888444193786608e-05,
+      "loss": 0.658,
+      "step": 4894
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888390345665567e-05,
+      "loss": 0.6816,
+      "step": 4895
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888336485319495e-05,
+      "loss": 0.7154,
+      "step": 4896
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888282612749132e-05,
+      "loss": 0.6464,
+      "step": 4897
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8882287279552197e-05,
+      "loss": 0.621,
+      "step": 4898
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8881748309384998e-05,
+      "loss": 0.747,
+      "step": 4899
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888120921699714e-05,
+      "loss": 0.7675,
+      "step": 4900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8880670002396034e-05,
+      "loss": 0.6208,
+      "step": 4901
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8880130665589106e-05,
+      "loss": 0.7447,
+      "step": 4902
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887959120658378e-05,
+      "loss": 0.7666,
+      "step": 4903
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8879051625387477e-05,
+      "loss": 0.6815,
+      "step": 4904
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887851192200762e-05,
+      "loss": 0.6398,
+      "step": 4905
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887797209645164e-05,
+      "loss": 0.6953,
+      "step": 4906
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8877432148726962e-05,
+      "loss": 0.6936,
+      "step": 4907
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8876892078841017e-05,
+      "loss": 0.7253,
+      "step": 4908
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8876351886801236e-05,
+      "loss": 0.6411,
+      "step": 4909
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887581157261505e-05,
+      "loss": 0.6534,
+      "step": 4910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8875271136289904e-05,
+      "loss": 0.5908,
+      "step": 4911
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8874730577833228e-05,
+      "loss": 0.7715,
+      "step": 4912
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8874189897252463e-05,
+      "loss": 0.7333,
+      "step": 4913
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8873649094555042e-05,
+      "loss": 0.7333,
+      "step": 4914
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8873108169748414e-05,
+      "loss": 0.75,
+      "step": 4915
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8872567122840024e-05,
+      "loss": 0.5978,
+      "step": 4916
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887202595383731e-05,
+      "loss": 0.7039,
+      "step": 4917
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8871484662747726e-05,
+      "loss": 0.7158,
+      "step": 4918
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8870943249578715e-05,
+      "loss": 0.6143,
+      "step": 4919
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8870401714337736e-05,
+      "loss": 0.7035,
+      "step": 4920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8869860057032234e-05,
+      "loss": 0.6874,
+      "step": 4921
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886931827766966e-05,
+      "loss": 0.649,
+      "step": 4922
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8868776376257476e-05,
+      "loss": 0.7291,
+      "step": 4923
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8868234352803134e-05,
+      "loss": 0.7257,
+      "step": 4924
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867692207314096e-05,
+      "loss": 0.7361,
+      "step": 4925
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867149939797822e-05,
+      "loss": 0.6421,
+      "step": 4926
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8866607550261773e-05,
+      "loss": 0.6819,
+      "step": 4927
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8866065038713414e-05,
+      "loss": 0.704,
+      "step": 4928
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8865522405160207e-05,
+      "loss": 0.6018,
+      "step": 4929
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8864979649609624e-05,
+      "loss": 0.6629,
+      "step": 4930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886443677206913e-05,
+      "loss": 0.7101,
+      "step": 4931
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88638937725462e-05,
+      "loss": 0.6666,
+      "step": 4932
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88633506510483e-05,
+      "loss": 0.7021,
+      "step": 4933
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8862807407582907e-05,
+      "loss": 0.6686,
+      "step": 4934
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88622640421575e-05,
+      "loss": 0.6572,
+      "step": 4935
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886172055477955e-05,
+      "loss": 0.7215,
+      "step": 4936
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861176945456542e-05,
+      "loss": 0.6239,
+      "step": 4937
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8860633214195947e-05,
+      "loss": 0.6523,
+      "step": 4938
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8860089361005255e-05,
+      "loss": 0.7115,
+      "step": 4939
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8859545385891952e-05,
+      "loss": 0.7747,
+      "step": 4940
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885900128886352e-05,
+      "loss": 0.7103,
+      "step": 4941
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8858457069927443e-05,
+      "loss": 0.6867,
+      "step": 4942
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8857912729091213e-05,
+      "loss": 0.719,
+      "step": 4943
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885736826636232e-05,
+      "loss": 0.7057,
+      "step": 4944
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885682368174826e-05,
+      "loss": 0.7125,
+      "step": 4945
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885627897525652e-05,
+      "loss": 0.6455,
+      "step": 4946
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88557341468946e-05,
+      "loss": 0.663,
+      "step": 4947
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8855189196670004e-05,
+      "loss": 0.7341,
+      "step": 4948
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8854644124590216e-05,
+      "loss": 0.6734,
+      "step": 4949
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8854098930662744e-05,
+      "loss": 0.6793,
+      "step": 4950
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8853553614895098e-05,
+      "loss": 0.7525,
+      "step": 4951
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8853008177294766e-05,
+      "loss": 0.6233,
+      "step": 4952
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885246261786927e-05,
+      "loss": 0.6619,
+      "step": 4953
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851916936626104e-05,
+      "loss": 0.6373,
+      "step": 4954
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851371133572787e-05,
+      "loss": 0.7049,
+      "step": 4955
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850825208716824e-05,
+      "loss": 0.6054,
+      "step": 4956
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885027916206573e-05,
+      "loss": 0.6971,
+      "step": 4957
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884973299362702e-05,
+      "loss": 0.7295,
+      "step": 4958
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8849186703408205e-05,
+      "loss": 0.7194,
+      "step": 4959
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8848640291416807e-05,
+      "loss": 0.6753,
+      "step": 4960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884809375766034e-05,
+      "loss": 0.6727,
+      "step": 4961
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884754710214633e-05,
+      "loss": 0.6678,
+      "step": 4962
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88470003248823e-05,
+      "loss": 0.6199,
+      "step": 4963
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884645342587577e-05,
+      "loss": 0.6643,
+      "step": 4964
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884590640513427e-05,
+      "loss": 0.8042,
+      "step": 4965
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8845359262665322e-05,
+      "loss": 0.7449,
+      "step": 4966
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884481199847646e-05,
+      "loss": 0.6026,
+      "step": 4967
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8844264612575213e-05,
+      "loss": 0.6796,
+      "step": 4968
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8843717104969114e-05,
+      "loss": 0.7371,
+      "step": 4969
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8843169475665698e-05,
+      "loss": 0.6805,
+      "step": 4970
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8842621724672498e-05,
+      "loss": 0.6831,
+      "step": 4971
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884207385199705e-05,
+      "loss": 0.7407,
+      "step": 4972
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8841525857646905e-05,
+      "loss": 0.6821,
+      "step": 4973
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884097774162959e-05,
+      "loss": 0.6762,
+      "step": 4974
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884042950395265e-05,
+      "loss": 0.6568,
+      "step": 4975
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883988114462364e-05,
+      "loss": 0.6861,
+      "step": 4976
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839332663650092e-05,
+      "loss": 0.6611,
+      "step": 4977
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883878406103956e-05,
+      "loss": 0.7373,
+      "step": 4978
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8838235336799593e-05,
+      "loss": 0.6875,
+      "step": 4979
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8837686490937744e-05,
+      "loss": 0.6573,
+      "step": 4980
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8837137523461558e-05,
+      "loss": 0.6474,
+      "step": 4981
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836588434378603e-05,
+      "loss": 0.7333,
+      "step": 4982
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883603922369642e-05,
+      "loss": 0.6165,
+      "step": 4983
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883548989142258e-05,
+      "loss": 0.6636,
+      "step": 4984
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8834940437564628e-05,
+      "loss": 0.6569,
+      "step": 4985
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8834390862130134e-05,
+      "loss": 0.7062,
+      "step": 4986
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833841165126662e-05,
+      "loss": 0.6617,
+      "step": 4987
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833291346561772e-05,
+      "loss": 0.6891,
+      "step": 4988
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8832741406443033e-05,
+      "loss": 0.7523,
+      "step": 4989
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883219134477801e-05,
+      "loss": 0.6657,
+      "step": 4990
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831641161574272e-05,
+      "loss": 0.654,
+      "step": 4991
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831090856839393e-05,
+      "loss": 0.6662,
+      "step": 4992
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8830540430580947e-05,
+      "loss": 0.6394,
+      "step": 4993
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88299898828065e-05,
+      "loss": 0.626,
+      "step": 4994
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829439213523636e-05,
+      "loss": 0.6078,
+      "step": 4995
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828888422739933e-05,
+      "loss": 0.6543,
+      "step": 4996
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882833751046296e-05,
+      "loss": 0.5953,
+      "step": 4997
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882778647670031e-05,
+      "loss": 0.5916,
+      "step": 4998
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8827235321459565e-05,
+      "loss": 0.711,
+      "step": 4999
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826684044748303e-05,
+      "loss": 0.7605,
+      "step": 5000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882613264657411e-05,
+      "loss": 0.7463,
+      "step": 5001
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882558112694458e-05,
+      "loss": 0.6883,
+      "step": 5002
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88250294858673e-05,
+      "loss": 0.7807,
+      "step": 5003
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824477723349853e-05,
+      "loss": 0.6822,
+      "step": 5004
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823925839399845e-05,
+      "loss": 0.6862,
+      "step": 5005
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823373834024862e-05,
+      "loss": 0.7235,
+      "step": 5006
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822821707232503e-05,
+      "loss": 0.7303,
+      "step": 5007
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822269459030362e-05,
+      "loss": 0.654,
+      "step": 5008
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821717089426045e-05,
+      "loss": 0.6943,
+      "step": 5009
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821164598427148e-05,
+      "loss": 0.6736,
+      "step": 5010
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820611986041277e-05,
+      "loss": 0.601,
+      "step": 5011
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882005925227603e-05,
+      "loss": 0.7027,
+      "step": 5012
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881950639713902e-05,
+      "loss": 0.6436,
+      "step": 5013
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818953420637855e-05,
+      "loss": 0.6466,
+      "step": 5014
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881840032278014e-05,
+      "loss": 0.7,
+      "step": 5015
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817847103573484e-05,
+      "loss": 0.6219,
+      "step": 5016
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881729376302551e-05,
+      "loss": 0.699,
+      "step": 5017
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816740301143823e-05,
+      "loss": 0.6617,
+      "step": 5018
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816186717936045e-05,
+      "loss": 0.6718,
+      "step": 5019
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881563301340979e-05,
+      "loss": 0.6174,
+      "step": 5020
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881507918757268e-05,
+      "loss": 0.7117,
+      "step": 5021
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814525240432327e-05,
+      "loss": 0.6348,
+      "step": 5022
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881397117199637e-05,
+      "loss": 0.6916,
+      "step": 5023
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813416982272422e-05,
+      "loss": 0.7486,
+      "step": 5024
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812862671268112e-05,
+      "loss": 0.7095,
+      "step": 5025
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881230823899107e-05,
+      "loss": 0.6787,
+      "step": 5026
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881175368544892e-05,
+      "loss": 0.6728,
+      "step": 5027
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88111990106493e-05,
+      "loss": 0.6836,
+      "step": 5028
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881064421459984e-05,
+      "loss": 0.6473,
+      "step": 5029
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881008929730817e-05,
+      "loss": 0.718,
+      "step": 5030
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880953425878193e-05,
+      "loss": 0.6585,
+      "step": 5031
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808979099028765e-05,
+      "loss": 0.7383,
+      "step": 5032
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88084238180563e-05,
+      "loss": 0.7494,
+      "step": 5033
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807868415872187e-05,
+      "loss": 0.6857,
+      "step": 5034
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807312892484062e-05,
+      "loss": 0.5984,
+      "step": 5035
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806757247899577e-05,
+      "loss": 0.6293,
+      "step": 5036
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880620148212637e-05,
+      "loss": 0.6491,
+      "step": 5037
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8805645595172096e-05,
+      "loss": 0.6991,
+      "step": 5038
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8805089587044397e-05,
+      "loss": 0.6879,
+      "step": 5039
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804533457750933e-05,
+      "loss": 0.6899,
+      "step": 5040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803977207299352e-05,
+      "loss": 0.7262,
+      "step": 5041
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803420835697307e-05,
+      "loss": 0.6444,
+      "step": 5042
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802864342952457e-05,
+      "loss": 0.709,
+      "step": 5043
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802307729072456e-05,
+      "loss": 0.7178,
+      "step": 5044
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801750994064968e-05,
+      "loss": 0.7193,
+      "step": 5045
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801194137937654e-05,
+      "loss": 0.7122,
+      "step": 5046
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800637160698175e-05,
+      "loss": 0.6141,
+      "step": 5047
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800080062354197e-05,
+      "loss": 0.5729,
+      "step": 5048
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879952284291338e-05,
+      "loss": 0.6241,
+      "step": 5049
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798965502383398e-05,
+      "loss": 0.7347,
+      "step": 5050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798408040771922e-05,
+      "loss": 0.7252,
+      "step": 5051
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879785045808662e-05,
+      "loss": 0.6102,
+      "step": 5052
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797292754335164e-05,
+      "loss": 0.6438,
+      "step": 5053
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796734929525227e-05,
+      "loss": 0.6804,
+      "step": 5054
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796176983664494e-05,
+      "loss": 0.6422,
+      "step": 5055
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795618916760634e-05,
+      "loss": 0.7228,
+      "step": 5056
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879506072882133e-05,
+      "loss": 0.6315,
+      "step": 5057
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879450241985426e-05,
+      "loss": 0.6773,
+      "step": 5058
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879394398986711e-05,
+      "loss": 0.7415,
+      "step": 5059
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793385438867564e-05,
+      "loss": 0.6806,
+      "step": 5060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879282676686331e-05,
+      "loss": 0.6928,
+      "step": 5061
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879226797386203e-05,
+      "loss": 0.6789,
+      "step": 5062
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879170905987142e-05,
+      "loss": 0.6283,
+      "step": 5063
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791150024899164e-05,
+      "loss": 0.6971,
+      "step": 5064
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790590868952963e-05,
+      "loss": 0.6937,
+      "step": 5065
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879003159204051e-05,
+      "loss": 0.5745,
+      "step": 5066
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789472194169493e-05,
+      "loss": 0.6586,
+      "step": 5067
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788912675347617e-05,
+      "loss": 0.6972,
+      "step": 5068
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788353035582583e-05,
+      "loss": 0.6039,
+      "step": 5069
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787793274882084e-05,
+      "loss": 0.6977,
+      "step": 5070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787233393253832e-05,
+      "loss": 0.6896,
+      "step": 5071
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8786673390705528e-05,
+      "loss": 0.6605,
+      "step": 5072
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8786113267244873e-05,
+      "loss": 0.7004,
+      "step": 5073
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785553022879583e-05,
+      "loss": 0.6971,
+      "step": 5074
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878499265761736e-05,
+      "loss": 0.7415,
+      "step": 5075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8784432171465923e-05,
+      "loss": 0.7371,
+      "step": 5076
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878387156443298e-05,
+      "loss": 0.6978,
+      "step": 5077
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8783310836526246e-05,
+      "loss": 0.683,
+      "step": 5078
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8782749987753437e-05,
+      "loss": 0.724,
+      "step": 5079
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878218901812227e-05,
+      "loss": 0.7334,
+      "step": 5080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781627927640465e-05,
+      "loss": 0.6096,
+      "step": 5081
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781066716315742e-05,
+      "loss": 0.6862,
+      "step": 5082
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878050538415583e-05,
+      "loss": 0.7114,
+      "step": 5083
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8779943931168446e-05,
+      "loss": 0.7273,
+      "step": 5084
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877938235736132e-05,
+      "loss": 0.6317,
+      "step": 5085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8778820662742178e-05,
+      "loss": 0.7415,
+      "step": 5086
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8778258847318748e-05,
+      "loss": 0.7257,
+      "step": 5087
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777696911098762e-05,
+      "loss": 0.6854,
+      "step": 5088
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777134854089958e-05,
+      "loss": 0.6442,
+      "step": 5089
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776572676300066e-05,
+      "loss": 0.7001,
+      "step": 5090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776010377736818e-05,
+      "loss": 0.7706,
+      "step": 5091
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877544795840796e-05,
+      "loss": 0.7037,
+      "step": 5092
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877488541832123e-05,
+      "loss": 0.586,
+      "step": 5093
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877432275748436e-05,
+      "loss": 0.7092,
+      "step": 5094
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8773759975905098e-05,
+      "loss": 0.6978,
+      "step": 5095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8773197073591195e-05,
+      "loss": 0.6734,
+      "step": 5096
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772634050550388e-05,
+      "loss": 0.5957,
+      "step": 5097
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772070906790428e-05,
+      "loss": 0.647,
+      "step": 5098
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771507642319067e-05,
+      "loss": 0.679,
+      "step": 5099
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877094425714405e-05,
+      "loss": 0.7513,
+      "step": 5100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877038075127314e-05,
+      "loss": 0.8349,
+      "step": 5101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876981712471408e-05,
+      "loss": 0.6996,
+      "step": 5102
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876925337747463e-05,
+      "loss": 0.6359,
+      "step": 5103
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768689509562548e-05,
+      "loss": 0.7283,
+      "step": 5104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768125520985594e-05,
+      "loss": 0.7355,
+      "step": 5105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767561411751528e-05,
+      "loss": 0.6138,
+      "step": 5106
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8766997181868115e-05,
+      "loss": 0.7328,
+      "step": 5107
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8766432831343117e-05,
+      "loss": 0.727,
+      "step": 5108
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87658683601843e-05,
+      "loss": 0.7306,
+      "step": 5109
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876530376839943e-05,
+      "loss": 0.6928,
+      "step": 5110
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876473905599628e-05,
+      "loss": 0.5999,
+      "step": 5111
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8764174222982617e-05,
+      "loss": 0.6867,
+      "step": 5112
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8763609269366217e-05,
+      "loss": 0.6272,
+      "step": 5113
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876304419515486e-05,
+      "loss": 0.6814,
+      "step": 5114
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8762479000356307e-05,
+      "loss": 0.6569,
+      "step": 5115
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8761913684978346e-05,
+      "loss": 0.6747,
+      "step": 5116
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876134824902875e-05,
+      "loss": 0.5902,
+      "step": 5117
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876078269251531e-05,
+      "loss": 0.6972,
+      "step": 5118
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87602170154458e-05,
+      "loss": 0.7617,
+      "step": 5119
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759651217828002e-05,
+      "loss": 0.6934,
+      "step": 5120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759085299669713e-05,
+      "loss": 0.7372,
+      "step": 5121
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875851926097871e-05,
+      "loss": 0.6375,
+      "step": 5122
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757953101762786e-05,
+      "loss": 0.5759,
+      "step": 5123
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757386822029733e-05,
+      "loss": 0.6453,
+      "step": 5124
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875682042178734e-05,
+      "loss": 0.7007,
+      "step": 5125
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875625390104341e-05,
+      "loss": 0.637,
+      "step": 5126
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875568725980573e-05,
+      "loss": 0.6983,
+      "step": 5127
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87551204980821e-05,
+      "loss": 0.7068,
+      "step": 5128
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8754553615880312e-05,
+      "loss": 0.6566,
+      "step": 5129
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753986613208183e-05,
+      "loss": 0.7755,
+      "step": 5130
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87534194900735e-05,
+      "loss": 0.6126,
+      "step": 5131
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752852246484077e-05,
+      "loss": 0.7349,
+      "step": 5132
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752284882447713e-05,
+      "loss": 0.6238,
+      "step": 5133
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751717397972222e-05,
+      "loss": 0.6644,
+      "step": 5134
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875114979306541e-05,
+      "loss": 0.617,
+      "step": 5135
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8750582067735084e-05,
+      "loss": 0.6931,
+      "step": 5136
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875001422198906e-05,
+      "loss": 0.6057,
+      "step": 5137
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749446255835154e-05,
+      "loss": 0.6992,
+      "step": 5138
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874887816928118e-05,
+      "loss": 0.6614,
+      "step": 5139
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874830996233495e-05,
+      "loss": 0.5915,
+      "step": 5140
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747741635004296e-05,
+      "loss": 0.5622,
+      "step": 5141
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747173187297025e-05,
+      "loss": 0.7422,
+      "step": 5142
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8746604619220968e-05,
+      "loss": 0.6871,
+      "step": 5143
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8746035930783942e-05,
+      "loss": 0.7508,
+      "step": 5144
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874546712199378e-05,
+      "loss": 0.7622,
+      "step": 5145
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744898192858305e-05,
+      "loss": 0.6273,
+      "step": 5146
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744329143385346e-05,
+      "loss": 0.6963,
+      "step": 5147
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743759973582735e-05,
+      "loss": 0.7287,
+      "step": 5148
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743190683458308e-05,
+      "loss": 0.7256,
+      "step": 5149
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874262127301989e-05,
+      "loss": 0.6827,
+      "step": 5150
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874205174227532e-05,
+      "loss": 0.6497,
+      "step": 5151
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741482091232442e-05,
+      "loss": 0.7881,
+      "step": 5152
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8740912319899084e-05,
+      "loss": 0.6417,
+      "step": 5153
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8740342428283096e-05,
+      "loss": 0.7274,
+      "step": 5154
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739772416392313e-05,
+      "loss": 0.7177,
+      "step": 5155
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739202284234584e-05,
+      "loss": 0.6433,
+      "step": 5156
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738632031817754e-05,
+      "loss": 0.6792,
+      "step": 5157
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873806165914967e-05,
+      "loss": 0.6571,
+      "step": 5158
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737491166238175e-05,
+      "loss": 0.6465,
+      "step": 5159
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8736920553091127e-05,
+      "loss": 0.7196,
+      "step": 5160
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873634981971637e-05,
+      "loss": 0.6839,
+      "step": 5161
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8735778966121772e-05,
+      "loss": 0.5833,
+      "step": 5162
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8735207992315173e-05,
+      "loss": 0.7551,
+      "step": 5163
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734636898304438e-05,
+      "loss": 0.698,
+      "step": 5164
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734065684097427e-05,
+      "loss": 0.7251,
+      "step": 5165
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8733494349701994e-05,
+      "loss": 0.6624,
+      "step": 5166
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732922895126006e-05,
+      "loss": 0.6364,
+      "step": 5167
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732351320377327e-05,
+      "loss": 0.6857,
+      "step": 5168
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873177962546382e-05,
+      "loss": 0.7275,
+      "step": 5169
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873120781039335e-05,
+      "loss": 0.707,
+      "step": 5170
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730635875173796e-05,
+      "loss": 0.6969,
+      "step": 5171
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730063819813016e-05,
+      "loss": 0.599,
+      "step": 5172
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8729491644318887e-05,
+      "loss": 0.6312,
+      "step": 5173
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728919348699285e-05,
+      "loss": 0.7405,
+      "step": 5174
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872834693296208e-05,
+      "loss": 0.7738,
+      "step": 5175
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872777439711515e-05,
+      "loss": 0.7389,
+      "step": 5176
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727201741166377e-05,
+      "loss": 0.7575,
+      "step": 5177
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726628965123643e-05,
+      "loss": 0.5916,
+      "step": 5178
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726056068994825e-05,
+      "loss": 0.687,
+      "step": 5179
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8725483052787807e-05,
+      "loss": 0.7358,
+      "step": 5180
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8724909916510478e-05,
+      "loss": 0.7658,
+      "step": 5181
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8724336660170718e-05,
+      "loss": 0.6873,
+      "step": 5182
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872376328377642e-05,
+      "loss": 0.7689,
+      "step": 5183
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8723189787335475e-05,
+      "loss": 0.7329,
+      "step": 5184
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8722616170855774e-05,
+      "loss": 0.6455,
+      "step": 5185
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872204243434521e-05,
+      "loss": 0.6983,
+      "step": 5186
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872146857781168e-05,
+      "loss": 0.7737,
+      "step": 5187
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720894601263077e-05,
+      "loss": 0.6003,
+      "step": 5188
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87203205047073e-05,
+      "loss": 0.741,
+      "step": 5189
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719746288152252e-05,
+      "loss": 0.6555,
+      "step": 5190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719171951605835e-05,
+      "loss": 0.7091,
+      "step": 5191
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871859749507595e-05,
+      "loss": 0.6276,
+      "step": 5192
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87180229185705e-05,
+      "loss": 0.6827,
+      "step": 5193
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8717448222097395e-05,
+      "loss": 0.7188,
+      "step": 5194
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8716873405664546e-05,
+      "loss": 0.6629,
+      "step": 5195
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8716298469279858e-05,
+      "loss": 0.7358,
+      "step": 5196
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8715723412951243e-05,
+      "loss": 0.6604,
+      "step": 5197
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8715148236686617e-05,
+      "loss": 0.6549,
+      "step": 5198
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8714572940493894e-05,
+      "loss": 0.7258,
+      "step": 5199
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8713997524380987e-05,
+      "loss": 0.6683,
+      "step": 5200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8713421988355817e-05,
+      "loss": 0.7079,
+      "step": 5201
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8712846332426307e-05,
+      "loss": 0.6881,
+      "step": 5202
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8712270556600374e-05,
+      "loss": 0.5928,
+      "step": 5203
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711694660885947e-05,
+      "loss": 0.592,
+      "step": 5204
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711118645290943e-05,
+      "loss": 0.6231,
+      "step": 5205
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871054250982329e-05,
+      "loss": 0.7277,
+      "step": 5206
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709966254490923e-05,
+      "loss": 0.597,
+      "step": 5207
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709389879301766e-05,
+      "loss": 0.7521,
+      "step": 5208
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870881338426375e-05,
+      "loss": 0.653,
+      "step": 5209
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870823676938481e-05,
+      "loss": 0.6628,
+      "step": 5210
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870766003467288e-05,
+      "loss": 0.6737,
+      "step": 5211
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8707083180135898e-05,
+      "loss": 0.6742,
+      "step": 5212
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87065062057818e-05,
+      "loss": 0.6885,
+      "step": 5213
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8705929111618525e-05,
+      "loss": 0.6202,
+      "step": 5214
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8705351897654016e-05,
+      "loss": 0.7509,
+      "step": 5215
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704774563896216e-05,
+      "loss": 0.6323,
+      "step": 5216
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704197110353073e-05,
+      "loss": 0.6692,
+      "step": 5217
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8703619537032526e-05,
+      "loss": 0.7834,
+      "step": 5218
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8703041843942526e-05,
+      "loss": 0.5765,
+      "step": 5219
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702464031091026e-05,
+      "loss": 0.7431,
+      "step": 5220
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870188609848597e-05,
+      "loss": 0.6085,
+      "step": 5221
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870130804613532e-05,
+      "loss": 0.6629,
+      "step": 5222
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870072987404702e-05,
+      "loss": 0.741,
+      "step": 5223
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700151582229035e-05,
+      "loss": 0.6678,
+      "step": 5224
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8699573170689322e-05,
+      "loss": 0.6305,
+      "step": 5225
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8698994639435833e-05,
+      "loss": 0.6809,
+      "step": 5226
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8698415988476538e-05,
+      "loss": 0.7178,
+      "step": 5227
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8697837217819398e-05,
+      "loss": 0.6811,
+      "step": 5228
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869725832747237e-05,
+      "loss": 0.7385,
+      "step": 5229
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8696679317443428e-05,
+      "loss": 0.7152,
+      "step": 5230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8696100187740535e-05,
+      "loss": 0.7759,
+      "step": 5231
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8695520938371667e-05,
+      "loss": 0.635,
+      "step": 5232
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8694941569344785e-05,
+      "loss": 0.6275,
+      "step": 5233
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869436208066787e-05,
+      "loss": 0.6849,
+      "step": 5234
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869378247234889e-05,
+      "loss": 0.7289,
+      "step": 5235
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869320274439583e-05,
+      "loss": 0.7287,
+      "step": 5236
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692622896816653e-05,
+      "loss": 0.7123,
+      "step": 5237
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692042929619352e-05,
+      "loss": 0.6841,
+      "step": 5238
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8691462842811906e-05,
+      "loss": 0.6086,
+      "step": 5239
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869088263640229e-05,
+      "loss": 0.7474,
+      "step": 5240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8690302310398495e-05,
+      "loss": 0.716,
+      "step": 5241
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689721864808502e-05,
+      "loss": 0.6308,
+      "step": 5242
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86891412996403e-05,
+      "loss": 0.7017,
+      "step": 5243
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8688560614901884e-05,
+      "loss": 0.643,
+      "step": 5244
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8687979810601234e-05,
+      "loss": 0.7604,
+      "step": 5245
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868739888674635e-05,
+      "loss": 0.7145,
+      "step": 5246
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8686817843345225e-05,
+      "loss": 0.6499,
+      "step": 5247
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8686236680405848e-05,
+      "loss": 0.5641,
+      "step": 5248
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8685655397936228e-05,
+      "loss": 0.735,
+      "step": 5249
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8685073995944356e-05,
+      "loss": 0.6982,
+      "step": 5250
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8684492474438233e-05,
+      "loss": 0.5908,
+      "step": 5251
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8683910833425868e-05,
+      "loss": 0.6376,
+      "step": 5252
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8683329072915253e-05,
+      "loss": 0.6988,
+      "step": 5253
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86827471929144e-05,
+      "loss": 0.6952,
+      "step": 5254
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868216519343132e-05,
+      "loss": 0.7019,
+      "step": 5255
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8681583074474018e-05,
+      "loss": 0.7669,
+      "step": 5256
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8681000836050504e-05,
+      "loss": 0.6782,
+      "step": 5257
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868041847816879e-05,
+      "loss": 0.6033,
+      "step": 5258
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8679836000836894e-05,
+      "loss": 0.5972,
+      "step": 5259
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867925340406282e-05,
+      "loss": 0.7917,
+      "step": 5260
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86786706878546e-05,
+      "loss": 0.6053,
+      "step": 5261
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8678087852220242e-05,
+      "loss": 0.6941,
+      "step": 5262
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867750489716777e-05,
+      "loss": 0.769,
+      "step": 5263
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867692182270521e-05,
+      "loss": 0.63,
+      "step": 5264
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8676338628840577e-05,
+      "loss": 0.6479,
+      "step": 5265
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675755315581903e-05,
+      "loss": 0.6606,
+      "step": 5266
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867517188293721e-05,
+      "loss": 0.5846,
+      "step": 5267
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8674588330914535e-05,
+      "loss": 0.6747,
+      "step": 5268
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8674004659521895e-05,
+      "loss": 0.7082,
+      "step": 5269
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8673420868767333e-05,
+      "loss": 0.5987,
+      "step": 5270
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867283695865888e-05,
+      "loss": 0.7075,
+      "step": 5271
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867225292920457e-05,
+      "loss": 0.6196,
+      "step": 5272
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8671668780412436e-05,
+      "loss": 0.7472,
+      "step": 5273
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8671084512290522e-05,
+      "loss": 0.6707,
+      "step": 5274
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8670500124846866e-05,
+      "loss": 0.6309,
+      "step": 5275
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866991561808951e-05,
+      "loss": 0.7684,
+      "step": 5276
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86693309920265e-05,
+      "loss": 0.6842,
+      "step": 5277
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8668746246665873e-05,
+      "loss": 0.7091,
+      "step": 5278
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8668161382015683e-05,
+      "loss": 0.6879,
+      "step": 5279
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8667576398083972e-05,
+      "loss": 0.7014,
+      "step": 5280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8666991294878798e-05,
+      "loss": 0.6448,
+      "step": 5281
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8666406072408206e-05,
+      "loss": 0.8334,
+      "step": 5282
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665820730680254e-05,
+      "loss": 0.6435,
+      "step": 5283
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665235269702993e-05,
+      "loss": 0.6073,
+      "step": 5284
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866464968948448e-05,
+      "loss": 0.7096,
+      "step": 5285
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8664063990032775e-05,
+      "loss": 0.6345,
+      "step": 5286
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866347817135594e-05,
+      "loss": 0.6029,
+      "step": 5287
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8662892233462028e-05,
+      "loss": 0.6832,
+      "step": 5288
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8662306176359103e-05,
+      "loss": 0.7419,
+      "step": 5289
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8661720000055243e-05,
+      "loss": 0.6974,
+      "step": 5290
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86611337045585e-05,
+      "loss": 0.7072,
+      "step": 5291
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8660547289876945e-05,
+      "loss": 0.7016,
+      "step": 5292
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8659960756018652e-05,
+      "loss": 0.6138,
+      "step": 5293
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8659374102991693e-05,
+      "loss": 0.6905,
+      "step": 5294
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8658787330804133e-05,
+      "loss": 0.6729,
+      "step": 5295
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865820043946405e-05,
+      "loss": 0.6547,
+      "step": 5296
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8657613428979526e-05,
+      "loss": 0.7561,
+      "step": 5297
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865702629935863e-05,
+      "loss": 0.6291,
+      "step": 5298
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865643905060945e-05,
+      "loss": 0.678,
+      "step": 5299
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865585168274006e-05,
+      "loss": 0.6493,
+      "step": 5300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8655264195758542e-05,
+      "loss": 0.7494,
+      "step": 5301
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8654676589672988e-05,
+      "loss": 0.6204,
+      "step": 5302
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8654088864491474e-05,
+      "loss": 0.6633,
+      "step": 5303
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8653501020222098e-05,
+      "loss": 0.6513,
+      "step": 5304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8652913056872944e-05,
+      "loss": 0.6247,
+      "step": 5305
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86523249744521e-05,
+      "loss": 0.7329,
+      "step": 5306
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8651736772967663e-05,
+      "loss": 0.6904,
+      "step": 5307
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865114845242773e-05,
+      "loss": 0.6274,
+      "step": 5308
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650560012840387e-05,
+      "loss": 0.6903,
+      "step": 5309
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864997145421374e-05,
+      "loss": 0.6039,
+      "step": 5310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649382776555885e-05,
+      "loss": 0.6411,
+      "step": 5311
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8648793979874924e-05,
+      "loss": 0.7297,
+      "step": 5312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8648205064178957e-05,
+      "loss": 0.629,
+      "step": 5313
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864761602947609e-05,
+      "loss": 0.6748,
+      "step": 5314
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8647026875774426e-05,
+      "loss": 0.6147,
+      "step": 5315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8646437603082078e-05,
+      "loss": 0.6673,
+      "step": 5316
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864584821140715e-05,
+      "loss": 0.7524,
+      "step": 5317
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864525870075775e-05,
+      "loss": 0.7239,
+      "step": 5318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8644669071141995e-05,
+      "loss": 0.644,
+      "step": 5319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8644079322568004e-05,
+      "loss": 0.675,
+      "step": 5320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864348945504388e-05,
+      "loss": 0.6028,
+      "step": 5321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642899468577747e-05,
+      "loss": 0.6511,
+      "step": 5322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642309363177723e-05,
+      "loss": 0.7356,
+      "step": 5323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864171913885193e-05,
+      "loss": 0.6215,
+      "step": 5324
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8641128795608486e-05,
+      "loss": 0.684,
+      "step": 5325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864053833345552e-05,
+      "loss": 0.6231,
+      "step": 5326
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8639947752401152e-05,
+      "loss": 0.6174,
+      "step": 5327
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8639357052453514e-05,
+      "loss": 0.6202,
+      "step": 5328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8638766233620725e-05,
+      "loss": 0.7047,
+      "step": 5329
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863817529591093e-05,
+      "loss": 0.7237,
+      "step": 5330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8637584239332248e-05,
+      "loss": 0.6144,
+      "step": 5331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8636993063892822e-05,
+      "loss": 0.6541,
+      "step": 5332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8636401769600776e-05,
+      "loss": 0.7008,
+      "step": 5333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8635810356464254e-05,
+      "loss": 0.6104,
+      "step": 5334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8635218824491398e-05,
+      "loss": 0.6472,
+      "step": 5335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863462717369034e-05,
+      "loss": 0.6151,
+      "step": 5336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8634035404069225e-05,
+      "loss": 0.6532,
+      "step": 5337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633443515636197e-05,
+      "loss": 0.6741,
+      "step": 5338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8632851508399402e-05,
+      "loss": 0.6921,
+      "step": 5339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8632259382366985e-05,
+      "loss": 0.6953,
+      "step": 5340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8631667137547088e-05,
+      "loss": 0.6484,
+      "step": 5341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8631074773947872e-05,
+      "loss": 0.5831,
+      "step": 5342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630482291577483e-05,
+      "loss": 0.7032,
+      "step": 5343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8629889690444073e-05,
+      "loss": 0.6109,
+      "step": 5344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86292969705558e-05,
+      "loss": 0.5526,
+      "step": 5345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8628704131920815e-05,
+      "loss": 0.6986,
+      "step": 5346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8628111174547282e-05,
+      "loss": 0.7188,
+      "step": 5347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8627518098443355e-05,
+      "loss": 0.5678,
+      "step": 5348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86269249036172e-05,
+      "loss": 0.7468,
+      "step": 5349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8626331590076978e-05,
+      "loss": 0.5864,
+      "step": 5350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625738157830855e-05,
+      "loss": 0.6912,
+      "step": 5351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625144606886993e-05,
+      "loss": 0.7605,
+      "step": 5352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862455093725356e-05,
+      "loss": 0.7346,
+      "step": 5353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8623957148938735e-05,
+      "loss": 0.6833,
+      "step": 5354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862336324195068e-05,
+      "loss": 0.5762,
+      "step": 5355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8622769216297567e-05,
+      "loss": 0.7121,
+      "step": 5356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8622175071987576e-05,
+      "loss": 0.6094,
+      "step": 5357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8621580809028876e-05,
+      "loss": 0.7,
+      "step": 5358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8620986427429652e-05,
+      "loss": 0.7551,
+      "step": 5359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8620391927198077e-05,
+      "loss": 0.6503,
+      "step": 5360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8619797308342337e-05,
+      "loss": 0.6473,
+      "step": 5361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8619202570870613e-05,
+      "loss": 0.6365,
+      "step": 5362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8618607714791084e-05,
+      "loss": 0.6749,
+      "step": 5363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861801274011194e-05,
+      "loss": 0.7005,
+      "step": 5364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861741764684137e-05,
+      "loss": 0.6462,
+      "step": 5365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8616822434987564e-05,
+      "loss": 0.6739,
+      "step": 5366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8616227104558703e-05,
+      "loss": 0.7284,
+      "step": 5367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8615631655562993e-05,
+      "loss": 0.6364,
+      "step": 5368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861503608800862e-05,
+      "loss": 0.679,
+      "step": 5369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8614440401903776e-05,
+      "loss": 0.6356,
+      "step": 5370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8613844597256668e-05,
+      "loss": 0.7471,
+      "step": 5371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861324867407549e-05,
+      "loss": 0.6682,
+      "step": 5372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8612652632368436e-05,
+      "loss": 0.7116,
+      "step": 5373
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8612056472143717e-05,
+      "loss": 0.698,
+      "step": 5374
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8611460193409535e-05,
+      "loss": 0.6999,
+      "step": 5375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861086379617409e-05,
+      "loss": 0.5788,
+      "step": 5376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86102672804456e-05,
+      "loss": 0.6353,
+      "step": 5377
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860967064623226e-05,
+      "loss": 0.7517,
+      "step": 5378
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860907389354229e-05,
+      "loss": 0.6635,
+      "step": 5379
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86084770223839e-05,
+      "loss": 0.7474,
+      "step": 5380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8607880032765298e-05,
+      "loss": 0.5916,
+      "step": 5381
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8607282924694706e-05,
+      "loss": 0.6991,
+      "step": 5382
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860668569818034e-05,
+      "loss": 0.6737,
+      "step": 5383
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860608835323041e-05,
+      "loss": 0.6308,
+      "step": 5384
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8605490889853145e-05,
+      "loss": 0.6331,
+      "step": 5385
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604893308056765e-05,
+      "loss": 0.7228,
+      "step": 5386
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604295607849495e-05,
+      "loss": 0.6181,
+      "step": 5387
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8603697789239555e-05,
+      "loss": 0.6847,
+      "step": 5388
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860309985223517e-05,
+      "loss": 0.6952,
+      "step": 5389
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8602501796844578e-05,
+      "loss": 0.6733,
+      "step": 5390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8601903623075996e-05,
+      "loss": 0.5826,
+      "step": 5391
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8601305330937664e-05,
+      "loss": 0.7434,
+      "step": 5392
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8600706920437815e-05,
+      "loss": 0.7117,
+      "step": 5393
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860010839158468e-05,
+      "loss": 0.7525,
+      "step": 5394
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8599509744386498e-05,
+      "loss": 0.6652,
+      "step": 5395
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8598910978851504e-05,
+      "loss": 0.565,
+      "step": 5396
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859831209498794e-05,
+      "loss": 0.6186,
+      "step": 5397
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8597713092804045e-05,
+      "loss": 0.6277,
+      "step": 5398
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8597113972308064e-05,
+      "loss": 0.7218,
+      "step": 5399
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8596514733508244e-05,
+      "loss": 0.6545,
+      "step": 5400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595915376412823e-05,
+      "loss": 0.6647,
+      "step": 5401
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595315901030057e-05,
+      "loss": 0.6161,
+      "step": 5402
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859471630736819e-05,
+      "loss": 0.7752,
+      "step": 5403
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859411659543547e-05,
+      "loss": 0.5711,
+      "step": 5404
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8593516765240164e-05,
+      "loss": 0.7073,
+      "step": 5405
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859291681679051e-05,
+      "loss": 0.751,
+      "step": 5406
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859231675009477e-05,
+      "loss": 0.7225,
+      "step": 5407
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8591716565161203e-05,
+      "loss": 0.6875,
+      "step": 5408
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8591116261998068e-05,
+      "loss": 0.6828,
+      "step": 5409
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8590515840613622e-05,
+      "loss": 0.7668,
+      "step": 5410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858991530101613e-05,
+      "loss": 0.7103,
+      "step": 5411
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589314643213855e-05,
+      "loss": 0.6753,
+      "step": 5412
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858871386721507e-05,
+      "loss": 0.6935,
+      "step": 5413
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8588112973028026e-05,
+      "loss": 0.601,
+      "step": 5414
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8587511960661006e-05,
+      "loss": 0.663,
+      "step": 5415
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8586910830122277e-05,
+      "loss": 0.7029,
+      "step": 5416
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858630958142011e-05,
+      "loss": 0.7503,
+      "step": 5417
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585708214562776e-05,
+      "loss": 0.6975,
+      "step": 5418
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585106729558553e-05,
+      "loss": 0.7288,
+      "step": 5419
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858450512641572e-05,
+      "loss": 0.7046,
+      "step": 5420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8583903405142552e-05,
+      "loss": 0.7489,
+      "step": 5421
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858330156574733e-05,
+      "loss": 0.5977,
+      "step": 5422
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858269960823834e-05,
+      "loss": 0.7505,
+      "step": 5423
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8582097532623856e-05,
+      "loss": 0.6541,
+      "step": 5424
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858149533891217e-05,
+      "loss": 0.6385,
+      "step": 5425
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8580893027111574e-05,
+      "loss": 0.7306,
+      "step": 5426
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8580290597230342e-05,
+      "loss": 0.624,
+      "step": 5427
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8579688049276777e-05,
+      "loss": 0.7554,
+      "step": 5428
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8579085383259165e-05,
+      "loss": 0.6981,
+      "step": 5429
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8578482599185804e-05,
+      "loss": 0.5884,
+      "step": 5430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8577879697064978e-05,
+      "loss": 0.7171,
+      "step": 5431
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857727667690499e-05,
+      "loss": 0.7084,
+      "step": 5432
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8576673538714144e-05,
+      "loss": 0.6703,
+      "step": 5433
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8576070282500732e-05,
+      "loss": 0.71,
+      "step": 5434
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8575466908273056e-05,
+      "loss": 0.6517,
+      "step": 5435
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857486341603942e-05,
+      "loss": 0.6269,
+      "step": 5436
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857425980580813e-05,
+      "loss": 0.6037,
+      "step": 5437
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857365607758749e-05,
+      "loss": 0.6955,
+      "step": 5438
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857305223138581e-05,
+      "loss": 0.7137,
+      "step": 5439
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8572448267211397e-05,
+      "loss": 0.6808,
+      "step": 5440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8571844185072563e-05,
+      "loss": 0.6765,
+      "step": 5441
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8571239984977625e-05,
+      "loss": 0.592,
+      "step": 5442
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8570635666934888e-05,
+      "loss": 0.7185,
+      "step": 5443
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8570031230952676e-05,
+      "loss": 0.7419,
+      "step": 5444
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8569426677039306e-05,
+      "loss": 0.6726,
+      "step": 5445
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8568822005203094e-05,
+      "loss": 0.5908,
+      "step": 5446
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856821721545236e-05,
+      "loss": 0.6183,
+      "step": 5447
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856761230779543e-05,
+      "loss": 0.6788,
+      "step": 5448
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8567007282240627e-05,
+      "loss": 0.6783,
+      "step": 5449
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8566402138796277e-05,
+      "loss": 0.5839,
+      "step": 5450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85657968774707e-05,
+      "loss": 0.6779,
+      "step": 5451
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8565191498272236e-05,
+      "loss": 0.6623,
+      "step": 5452
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8564586001209214e-05,
+      "loss": 0.5812,
+      "step": 5453
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856398038628996e-05,
+      "loss": 0.6893,
+      "step": 5454
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8563374653522814e-05,
+      "loss": 0.7025,
+      "step": 5455
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8562768802916104e-05,
+      "loss": 0.5607,
+      "step": 5456
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8562162834478173e-05,
+      "loss": 0.6928,
+      "step": 5457
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8561556748217357e-05,
+      "loss": 0.6645,
+      "step": 5458
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8560950544142e-05,
+      "loss": 0.6392,
+      "step": 5459
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856034422226044e-05,
+      "loss": 0.6026,
+      "step": 5460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8559737782581017e-05,
+      "loss": 0.7106,
+      "step": 5461
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8559131225112085e-05,
+      "loss": 0.6685,
+      "step": 5462
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558524549861986e-05,
+      "loss": 0.754,
+      "step": 5463
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855791775683907e-05,
+      "loss": 0.6968,
+      "step": 5464
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8557310846051687e-05,
+      "loss": 0.6037,
+      "step": 5465
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8556703817508187e-05,
+      "loss": 0.7259,
+      "step": 5466
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855609667121692e-05,
+      "loss": 0.7289,
+      "step": 5467
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855548940718625e-05,
+      "loss": 0.6397,
+      "step": 5468
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8554882025424526e-05,
+      "loss": 0.6246,
+      "step": 5469
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855427452594011e-05,
+      "loss": 0.6917,
+      "step": 5470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855366690874136e-05,
+      "loss": 0.636,
+      "step": 5471
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8553059173836634e-05,
+      "loss": 0.69,
+      "step": 5472
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8552451321234303e-05,
+      "loss": 0.6149,
+      "step": 5473
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8551843350942723e-05,
+      "loss": 0.6359,
+      "step": 5474
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8551235262970264e-05,
+      "loss": 0.7192,
+      "step": 5475
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8550627057325298e-05,
+      "loss": 0.7295,
+      "step": 5476
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8550018734016187e-05,
+      "loss": 0.693,
+      "step": 5477
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8549410293051305e-05,
+      "loss": 0.7476,
+      "step": 5478
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854880173443903e-05,
+      "loss": 0.727,
+      "step": 5479
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8548193058187725e-05,
+      "loss": 0.6778,
+      "step": 5480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8547584264305777e-05,
+      "loss": 0.7321,
+      "step": 5481
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8546975352801557e-05,
+      "loss": 0.6203,
+      "step": 5482
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8546366323683445e-05,
+      "loss": 0.709,
+      "step": 5483
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8545757176959826e-05,
+      "loss": 0.6988,
+      "step": 5484
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854514791263908e-05,
+      "loss": 0.672,
+      "step": 5485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8544538530729587e-05,
+      "loss": 0.6281,
+      "step": 5486
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8543929031239737e-05,
+      "loss": 0.7302,
+      "step": 5487
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8543319414177916e-05,
+      "loss": 0.6574,
+      "step": 5488
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8542709679552513e-05,
+      "loss": 0.6965,
+      "step": 5489
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854209982737192e-05,
+      "loss": 0.6338,
+      "step": 5490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8541489857644527e-05,
+      "loss": 0.6744,
+      "step": 5491
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854087977037873e-05,
+      "loss": 0.652,
+      "step": 5492
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8540269565582918e-05,
+      "loss": 0.5938,
+      "step": 5493
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8539659243265497e-05,
+      "loss": 0.6041,
+      "step": 5494
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8539048803434858e-05,
+      "loss": 0.6012,
+      "step": 5495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8538438246099405e-05,
+      "loss": 0.6305,
+      "step": 5496
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8537827571267542e-05,
+      "loss": 0.6101,
+      "step": 5497
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8537216778947666e-05,
+      "loss": 0.6877,
+      "step": 5498
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8536605869148186e-05,
+      "loss": 0.6924,
+      "step": 5499
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8535994841877512e-05,
+      "loss": 0.6964,
+      "step": 5500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8535383697144046e-05,
+      "loss": 0.5803,
+      "step": 5501
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85347724349562e-05,
+      "loss": 0.6316,
+      "step": 5502
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8534161055322388e-05,
+      "loss": 0.6488,
+      "step": 5503
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533549558251018e-05,
+      "loss": 0.603,
+      "step": 5504
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853293794375051e-05,
+      "loss": 0.7582,
+      "step": 5505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853232621182928e-05,
+      "loss": 0.669,
+      "step": 5506
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853171436249574e-05,
+      "loss": 0.6565,
+      "step": 5507
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8531102395758318e-05,
+      "loss": 0.704,
+      "step": 5508
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8530490311625428e-05,
+      "loss": 0.7064,
+      "step": 5509
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8529878110105498e-05,
+      "loss": 0.6559,
+      "step": 5510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8529265791206947e-05,
+      "loss": 0.6355,
+      "step": 5511
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8528653354938205e-05,
+      "loss": 0.775,
+      "step": 5512
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8528040801307702e-05,
+      "loss": 0.6209,
+      "step": 5513
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852742813032386e-05,
+      "loss": 0.6061,
+      "step": 5514
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8526815341995117e-05,
+      "loss": 0.6026,
+      "step": 5515
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85262024363299e-05,
+      "loss": 0.6585,
+      "step": 5516
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852558941333665e-05,
+      "loss": 0.6834,
+      "step": 5517
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8524976273023795e-05,
+      "loss": 0.6598,
+      "step": 5518
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8524363015399774e-05,
+      "loss": 0.6728,
+      "step": 5519
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852374964047303e-05,
+      "loss": 0.7359,
+      "step": 5520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8523136148252003e-05,
+      "loss": 0.6555,
+      "step": 5521
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8522522538745133e-05,
+      "loss": 0.7449,
+      "step": 5522
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521908811960862e-05,
+      "loss": 0.6753,
+      "step": 5523
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852129496790764e-05,
+      "loss": 0.7224,
+      "step": 5524
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8520681006593913e-05,
+      "loss": 0.7106,
+      "step": 5525
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8520066928028124e-05,
+      "loss": 0.6864,
+      "step": 5526
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8519452732218732e-05,
+      "loss": 0.6964,
+      "step": 5527
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851883841917418e-05,
+      "loss": 0.7214,
+      "step": 5528
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8518223988902934e-05,
+      "loss": 0.7196,
+      "step": 5529
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8517609441413434e-05,
+      "loss": 0.7104,
+      "step": 5530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8516994776714145e-05,
+      "loss": 0.6974,
+      "step": 5531
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8516379994813526e-05,
+      "loss": 0.7164,
+      "step": 5532
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8515765095720033e-05,
+      "loss": 0.6002,
+      "step": 5533
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851515007944213e-05,
+      "loss": 0.6902,
+      "step": 5534
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851453494598828e-05,
+      "loss": 0.7212,
+      "step": 5535
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8513919695366952e-05,
+      "loss": 0.6305,
+      "step": 5536
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85133043275866e-05,
+      "loss": 0.713,
+      "step": 5537
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8512688842655705e-05,
+      "loss": 0.6197,
+      "step": 5538
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8512073240582732e-05,
+      "loss": 0.7104,
+      "step": 5539
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851145752137615e-05,
+      "loss": 0.6452,
+      "step": 5540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8510841685044436e-05,
+      "loss": 0.6477,
+      "step": 5541
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8510225731596058e-05,
+      "loss": 0.5921,
+      "step": 5542
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85096096610395e-05,
+      "loss": 0.6942,
+      "step": 5543
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8508993473383234e-05,
+      "loss": 0.7159,
+      "step": 5544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8508377168635743e-05,
+      "loss": 0.7651,
+      "step": 5545
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507760746805507e-05,
+      "loss": 0.675,
+      "step": 5546
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507144207901003e-05,
+      "loss": 0.6653,
+      "step": 5547
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850652755193072e-05,
+      "loss": 0.6758,
+      "step": 5548
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8505910778903146e-05,
+      "loss": 0.6066,
+      "step": 5549
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8505293888826764e-05,
+      "loss": 0.7372,
+      "step": 5550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8504676881710066e-05,
+      "loss": 0.6258,
+      "step": 5551
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850405975756154e-05,
+      "loss": 0.6973,
+      "step": 5552
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8503442516389684e-05,
+      "loss": 0.7553,
+      "step": 5553
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8502825158202983e-05,
+      "loss": 0.5721,
+      "step": 5554
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850220768300994e-05,
+      "loss": 0.721,
+      "step": 5555
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8501590090819044e-05,
+      "loss": 0.6775,
+      "step": 5556
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8500972381638797e-05,
+      "loss": 0.5937,
+      "step": 5557
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8500354555477706e-05,
+      "loss": 0.5917,
+      "step": 5558
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8499736612344265e-05,
+      "loss": 0.71,
+      "step": 5559
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849911855224698e-05,
+      "loss": 0.6333,
+      "step": 5560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849850037519435e-05,
+      "loss": 0.5757,
+      "step": 5561
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8497882081194896e-05,
+      "loss": 0.6861,
+      "step": 5562
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8497263670257112e-05,
+      "loss": 0.7536,
+      "step": 5563
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8496645142389516e-05,
+      "loss": 0.6185,
+      "step": 5564
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849602649760062e-05,
+      "loss": 0.7166,
+      "step": 5565
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849540773589893e-05,
+      "loss": 0.6871,
+      "step": 5566
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8494788857292968e-05,
+      "loss": 0.7096,
+      "step": 5567
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849416986179124e-05,
+      "loss": 0.7455,
+      "step": 5568
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8493550749402278e-05,
+      "loss": 0.73,
+      "step": 5569
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8492931520134593e-05,
+      "loss": 0.6314,
+      "step": 5570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8492312173996707e-05,
+      "loss": 0.6764,
+      "step": 5571
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849169271099714e-05,
+      "loss": 0.6921,
+      "step": 5572
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8491073131144427e-05,
+      "loss": 0.6289,
+      "step": 5573
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8490453434447085e-05,
+      "loss": 0.706,
+      "step": 5574
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8489833620913644e-05,
+      "loss": 0.6241,
+      "step": 5575
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8489213690552627e-05,
+      "loss": 0.6699,
+      "step": 5576
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8488593643372574e-05,
+      "loss": 0.6924,
+      "step": 5577
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8487973479382013e-05,
+      "loss": 0.6682,
+      "step": 5578
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848735319858948e-05,
+      "loss": 0.6722,
+      "step": 5579
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8486732801003508e-05,
+      "loss": 0.7253,
+      "step": 5580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8486112286632634e-05,
+      "loss": 0.6962,
+      "step": 5581
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84854916554854e-05,
+      "loss": 0.6837,
+      "step": 5582
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848487090757035e-05,
+      "loss": 0.6036,
+      "step": 5583
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8484250042896014e-05,
+      "loss": 0.6901,
+      "step": 5584
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8483629061470946e-05,
+      "loss": 0.7004,
+      "step": 5585
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8483007963303684e-05,
+      "loss": 0.706,
+      "step": 5586
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8482386748402784e-05,
+      "loss": 0.7159,
+      "step": 5587
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8481765416776786e-05,
+      "loss": 0.7148,
+      "step": 5588
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8481143968434242e-05,
+      "loss": 0.6805,
+      "step": 5589
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8480522403383704e-05,
+      "loss": 0.6447,
+      "step": 5590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847990072163373e-05,
+      "loss": 0.6881,
+      "step": 5591
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847927892319287e-05,
+      "loss": 0.6982,
+      "step": 5592
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847865700806968e-05,
+      "loss": 0.7842,
+      "step": 5593
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8478034976272717e-05,
+      "loss": 0.7052,
+      "step": 5594
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8477412827810547e-05,
+      "loss": 0.7108,
+      "step": 5595
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847679056269173e-05,
+      "loss": 0.64,
+      "step": 5596
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847616818092482e-05,
+      "loss": 0.715,
+      "step": 5597
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8475545682518393e-05,
+      "loss": 0.6788,
+      "step": 5598
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847492306748101e-05,
+      "loss": 0.7517,
+      "step": 5599
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8474300335821235e-05,
+      "loss": 0.6386,
+      "step": 5600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8473677487547644e-05,
+      "loss": 0.6805,
+      "step": 5601
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8473054522668804e-05,
+      "loss": 0.6642,
+      "step": 5602
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8472431441193293e-05,
+      "loss": 0.623,
+      "step": 5603
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8471808243129678e-05,
+      "loss": 0.6819,
+      "step": 5604
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8471184928486536e-05,
+      "loss": 0.6618,
+      "step": 5605
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8470561497272446e-05,
+      "loss": 0.6382,
+      "step": 5606
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8469937949495987e-05,
+      "loss": 0.6596,
+      "step": 5607
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8469314285165742e-05,
+      "loss": 0.6686,
+      "step": 5608
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846869050429029e-05,
+      "loss": 0.6872,
+      "step": 5609
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846806660687821e-05,
+      "loss": 0.6384,
+      "step": 5610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84674425929381e-05,
+      "loss": 0.6381,
+      "step": 5611
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8466818462478536e-05,
+      "loss": 0.7193,
+      "step": 5612
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8466194215508112e-05,
+      "loss": 0.652,
+      "step": 5613
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8465569852035415e-05,
+      "loss": 0.7577,
+      "step": 5614
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8464945372069037e-05,
+      "loss": 0.7534,
+      "step": 5615
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8464320775617577e-05,
+      "loss": 0.6471,
+      "step": 5616
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8463696062689623e-05,
+      "loss": 0.7724,
+      "step": 5617
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8463071233293775e-05,
+      "loss": 0.6684,
+      "step": 5618
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8462446287438632e-05,
+      "loss": 0.6839,
+      "step": 5619
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846182122513279e-05,
+      "loss": 0.6983,
+      "step": 5620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8461196046384853e-05,
+      "loss": 0.5807,
+      "step": 5621
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8460570751203423e-05,
+      "loss": 0.6601,
+      "step": 5622
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845994533959711e-05,
+      "loss": 0.6387,
+      "step": 5623
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8459319811574508e-05,
+      "loss": 0.7545,
+      "step": 5624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845869416714424e-05,
+      "loss": 0.7525,
+      "step": 5625
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8458068406314903e-05,
+      "loss": 0.6453,
+      "step": 5626
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8457442529095114e-05,
+      "loss": 0.6459,
+      "step": 5627
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8456816535493484e-05,
+      "loss": 0.5703,
+      "step": 5628
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845619042551863e-05,
+      "loss": 0.7253,
+      "step": 5629
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8455564199179162e-05,
+      "loss": 0.612,
+      "step": 5630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8454937856483703e-05,
+      "loss": 0.6071,
+      "step": 5631
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8454311397440873e-05,
+      "loss": 0.6378,
+      "step": 5632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8453684822059284e-05,
+      "loss": 0.7208,
+      "step": 5633
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8453058130347565e-05,
+      "loss": 0.6942,
+      "step": 5634
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845243132231434e-05,
+      "loss": 0.675,
+      "step": 5635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845180439796823e-05,
+      "loss": 0.676,
+      "step": 5636
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8451177357317868e-05,
+      "loss": 0.7406,
+      "step": 5637
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845055020037188e-05,
+      "loss": 0.6422,
+      "step": 5638
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8449922927138897e-05,
+      "loss": 0.6305,
+      "step": 5639
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844929553762755e-05,
+      "loss": 0.6804,
+      "step": 5640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8448668031846468e-05,
+      "loss": 0.6718,
+      "step": 5641
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8448040409804294e-05,
+      "loss": 0.7366,
+      "step": 5642
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844741267150966e-05,
+      "loss": 0.5746,
+      "step": 5643
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446784816971203e-05,
+      "loss": 0.6938,
+      "step": 5644
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446156846197565e-05,
+      "loss": 0.6819,
+      "step": 5645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8445528759197387e-05,
+      "loss": 0.761,
+      "step": 5646
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8444900555979318e-05,
+      "loss": 0.7161,
+      "step": 5647
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844427223655199e-05,
+      "loss": 0.6536,
+      "step": 5648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8443643800924063e-05,
+      "loss": 0.7242,
+      "step": 5649
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8443015249104174e-05,
+      "loss": 0.6372,
+      "step": 5650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8442386581100977e-05,
+      "loss": 0.6038,
+      "step": 5651
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844175779692312e-05,
+      "loss": 0.6617,
+      "step": 5652
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8441128896579264e-05,
+      "loss": 0.5905,
+      "step": 5653
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8440499880078052e-05,
+      "loss": 0.7226,
+      "step": 5654
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8439870747428148e-05,
+      "loss": 0.5694,
+      "step": 5655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8439241498638206e-05,
+      "loss": 0.7015,
+      "step": 5656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8438612133716883e-05,
+      "loss": 0.7165,
+      "step": 5657
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8437982652672848e-05,
+      "loss": 0.5771,
+      "step": 5658
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843735305551475e-05,
+      "loss": 0.6322,
+      "step": 5659
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8436723342251267e-05,
+      "loss": 0.6134,
+      "step": 5660
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8436093512891052e-05,
+      "loss": 0.655,
+      "step": 5661
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8435463567442783e-05,
+      "loss": 0.7143,
+      "step": 5662
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8434833505915123e-05,
+      "loss": 0.6757,
+      "step": 5663
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843420332831674e-05,
+      "loss": 0.6854,
+      "step": 5664
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8433573034656308e-05,
+      "loss": 0.6915,
+      "step": 5665
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84329426249425e-05,
+      "loss": 0.7148,
+      "step": 5666
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8432312099183993e-05,
+      "loss": 0.6173,
+      "step": 5667
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8431681457389463e-05,
+      "loss": 0.6904,
+      "step": 5668
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8431050699567585e-05,
+      "loss": 0.7095,
+      "step": 5669
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8430419825727044e-05,
+      "loss": 0.5908,
+      "step": 5670
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8429788835876515e-05,
+      "loss": 0.6749,
+      "step": 5671
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842915773002469e-05,
+      "loss": 0.6537,
+      "step": 5672
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8428526508180245e-05,
+      "loss": 0.6266,
+      "step": 5673
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8427895170351868e-05,
+      "loss": 0.6823,
+      "step": 5674
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8427263716548248e-05,
+      "loss": 0.7063,
+      "step": 5675
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8426632146778075e-05,
+      "loss": 0.7018,
+      "step": 5676
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8426000461050045e-05,
+      "loss": 0.7214,
+      "step": 5677
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842536865937284e-05,
+      "loss": 0.6848,
+      "step": 5678
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8424736741755156e-05,
+      "loss": 0.691,
+      "step": 5679
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8424104708205694e-05,
+      "loss": 0.666,
+      "step": 5680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842347255873315e-05,
+      "loss": 0.614,
+      "step": 5681
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842284029334622e-05,
+      "loss": 0.729,
+      "step": 5682
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8422207912053612e-05,
+      "loss": 0.5621,
+      "step": 5683
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842157541486402e-05,
+      "loss": 0.5983,
+      "step": 5684
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842094280178615e-05,
+      "loss": 0.7006,
+      "step": 5685
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8420310072828707e-05,
+      "loss": 0.6046,
+      "step": 5686
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84196772280004e-05,
+      "loss": 0.6887,
+      "step": 5687
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841904426730994e-05,
+      "loss": 0.6491,
+      "step": 5688
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8418411190766026e-05,
+      "loss": 0.6344,
+      "step": 5689
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841777799837738e-05,
+      "loss": 0.6017,
+      "step": 5690
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8417144690152712e-05,
+      "loss": 0.6199,
+      "step": 5691
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8416511266100736e-05,
+      "loss": 0.6787,
+      "step": 5692
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841587772623017e-05,
+      "loss": 0.6862,
+      "step": 5693
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8415244070549733e-05,
+      "loss": 0.5818,
+      "step": 5694
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8414610299068142e-05,
+      "loss": 0.5825,
+      "step": 5695
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841397641179412e-05,
+      "loss": 0.6422,
+      "step": 5696
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8413342408736393e-05,
+      "loss": 0.7183,
+      "step": 5697
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8412708289903676e-05,
+      "loss": 0.7076,
+      "step": 5698
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8412074055304704e-05,
+      "loss": 0.6803,
+      "step": 5699
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84114397049482e-05,
+      "loss": 0.6513,
+      "step": 5700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8410805238842898e-05,
+      "loss": 0.6569,
+      "step": 5701
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841017065699752e-05,
+      "loss": 0.5957,
+      "step": 5702
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840953595942081e-05,
+      "loss": 0.7563,
+      "step": 5703
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408901146121493e-05,
+      "loss": 0.6088,
+      "step": 5704
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408266217108305e-05,
+      "loss": 0.6802,
+      "step": 5705
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840763117238999e-05,
+      "loss": 0.6121,
+      "step": 5706
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840699601197528e-05,
+      "loss": 0.6709,
+      "step": 5707
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840636073587292e-05,
+      "loss": 0.6938,
+      "step": 5708
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840572534409165e-05,
+      "loss": 0.7361,
+      "step": 5709
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840508983664021e-05,
+      "loss": 0.6992,
+      "step": 5710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840445421352735e-05,
+      "loss": 0.6133,
+      "step": 5711
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8403818474761815e-05,
+      "loss": 0.6788,
+      "step": 5712
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8403182620352352e-05,
+      "loss": 0.6623,
+      "step": 5713
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8402546650307717e-05,
+      "loss": 0.5664,
+      "step": 5714
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8401910564636653e-05,
+      "loss": 0.6493,
+      "step": 5715
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8401274363347917e-05,
+      "loss": 0.6331,
+      "step": 5716
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8400638046450266e-05,
+      "loss": 0.714,
+      "step": 5717
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840000161395245e-05,
+      "loss": 0.6726,
+      "step": 5718
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839936506586324e-05,
+      "loss": 0.6589,
+      "step": 5719
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8398728402191376e-05,
+      "loss": 0.6452,
+      "step": 5720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8398091622945636e-05,
+      "loss": 0.7041,
+      "step": 5721
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8397454728134775e-05,
+      "loss": 0.6743,
+      "step": 5722
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8396817717767557e-05,
+      "loss": 0.5799,
+      "step": 5723
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839618059185275e-05,
+      "loss": 0.7514,
+      "step": 5724
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8395543350399117e-05,
+      "loss": 0.8337,
+      "step": 5725
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8394905993415436e-05,
+      "loss": 0.7426,
+      "step": 5726
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8394268520910467e-05,
+      "loss": 0.6391,
+      "step": 5727
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8393630932892992e-05,
+      "loss": 0.6856,
+      "step": 5728
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8392993229371774e-05,
+      "loss": 0.5885,
+      "step": 5729
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83923554103556e-05,
+      "loss": 0.6855,
+      "step": 5730
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8391717475853242e-05,
+      "loss": 0.7524,
+      "step": 5731
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8391079425873475e-05,
+      "loss": 0.6592,
+      "step": 5732
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8390441260425082e-05,
+      "loss": 0.706,
+      "step": 5733
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8389802979516846e-05,
+      "loss": 0.6595,
+      "step": 5734
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838916458315755e-05,
+      "loss": 0.6772,
+      "step": 5735
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8388526071355976e-05,
+      "loss": 0.6467,
+      "step": 5736
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8387887444120914e-05,
+      "loss": 0.7192,
+      "step": 5737
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838724870146115e-05,
+      "loss": 0.6146,
+      "step": 5738
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8386609843385473e-05,
+      "loss": 0.659,
+      "step": 5739
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8385970869902675e-05,
+      "loss": 0.6313,
+      "step": 5740
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838533178102155e-05,
+      "loss": 0.6294,
+      "step": 5741
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8384692576750892e-05,
+      "loss": 0.6979,
+      "step": 5742
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8384053257099498e-05,
+      "loss": 0.7489,
+      "step": 5743
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838341382207616e-05,
+      "loss": 0.7169,
+      "step": 5744
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838277427168969e-05,
+      "loss": 0.6271,
+      "step": 5745
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8382134605948872e-05,
+      "loss": 0.7123,
+      "step": 5746
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838149482486252e-05,
+      "loss": 0.7016,
+      "step": 5747
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8380854928439434e-05,
+      "loss": 0.6844,
+      "step": 5748
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838021491668842e-05,
+      "loss": 0.7253,
+      "step": 5749
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8379574789618284e-05,
+      "loss": 0.6521,
+      "step": 5750
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837893454723784e-05,
+      "loss": 0.6312,
+      "step": 5751
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837829418955589e-05,
+      "loss": 0.6679,
+      "step": 5752
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837765371658125e-05,
+      "loss": 0.6242,
+      "step": 5753
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8377013128322736e-05,
+      "loss": 0.6723,
+      "step": 5754
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837637242478916e-05,
+      "loss": 0.5695,
+      "step": 5755
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837573160598934e-05,
+      "loss": 0.7023,
+      "step": 5756
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8375090671932093e-05,
+      "loss": 0.6991,
+      "step": 5757
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837444962262624e-05,
+      "loss": 0.7393,
+      "step": 5758
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83738084580806e-05,
+      "loss": 0.721,
+      "step": 5759
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8373167178304e-05,
+      "loss": 0.6822,
+      "step": 5760
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837252578330526e-05,
+      "loss": 0.7006,
+      "step": 5761
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8371884273093213e-05,
+      "loss": 0.6366,
+      "step": 5762
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8371242647676678e-05,
+      "loss": 0.6511,
+      "step": 5763
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837060090706449e-05,
+      "loss": 0.6231,
+      "step": 5764
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8369959051265476e-05,
+      "loss": 0.5789,
+      "step": 5765
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836931708028847e-05,
+      "loss": 0.7278,
+      "step": 5766
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8368674994142314e-05,
+      "loss": 0.6745,
+      "step": 5767
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836803279283583e-05,
+      "loss": 0.748,
+      "step": 5768
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8367390476377864e-05,
+      "loss": 0.7147,
+      "step": 5769
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8366748044777254e-05,
+      "loss": 0.6556,
+      "step": 5770
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8366105498042837e-05,
+      "loss": 0.7106,
+      "step": 5771
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836546283618346e-05,
+      "loss": 0.7797,
+      "step": 5772
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8364820059207964e-05,
+      "loss": 0.6113,
+      "step": 5773
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836417716712519e-05,
+      "loss": 0.6643,
+      "step": 5774
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8363534159943996e-05,
+      "loss": 0.7178,
+      "step": 5775
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8362891037673215e-05,
+      "loss": 0.7431,
+      "step": 5776
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836224780032171e-05,
+      "loss": 0.7713,
+      "step": 5777
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8361604447898325e-05,
+      "loss": 0.6679,
+      "step": 5778
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836096098041192e-05,
+      "loss": 0.628,
+      "step": 5779
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8360317397871342e-05,
+      "loss": 0.6748,
+      "step": 5780
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835967370028545e-05,
+      "loss": 0.7093,
+      "step": 5781
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8359029887663103e-05,
+      "loss": 0.6864,
+      "step": 5782
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835838596001316e-05,
+      "loss": 0.7406,
+      "step": 5783
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835774191734448e-05,
+      "loss": 0.6181,
+      "step": 5784
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835709775966593e-05,
+      "loss": 0.6149,
+      "step": 5785
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835645348698637e-05,
+      "loss": 0.6983,
+      "step": 5786
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835580909931467e-05,
+      "loss": 0.7403,
+      "step": 5787
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8355164596659693e-05,
+      "loss": 0.6992,
+      "step": 5788
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835451997903031e-05,
+      "loss": 0.6855,
+      "step": 5789
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835387524643539e-05,
+      "loss": 0.7069,
+      "step": 5790
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8353230398883805e-05,
+      "loss": 0.6659,
+      "step": 5791
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8352585436384432e-05,
+      "loss": 0.6869,
+      "step": 5792
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8351940358946145e-05,
+      "loss": 0.6426,
+      "step": 5793
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835129516657782e-05,
+      "loss": 0.718,
+      "step": 5794
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8350649859288336e-05,
+      "loss": 0.6902,
+      "step": 5795
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835000443708657e-05,
+      "loss": 0.6947,
+      "step": 5796
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834935889998141e-05,
+      "loss": 0.6711,
+      "step": 5797
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834871324798173e-05,
+      "loss": 0.7219,
+      "step": 5798
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8348067481096423e-05,
+      "loss": 0.6451,
+      "step": 5799
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8347421599334368e-05,
+      "loss": 0.7299,
+      "step": 5800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8346775602704464e-05,
+      "loss": 0.6715,
+      "step": 5801
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834612949121559e-05,
+      "loss": 0.7056,
+      "step": 5802
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834548326487664e-05,
+      "loss": 0.7239,
+      "step": 5803
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8344836923696506e-05,
+      "loss": 0.763,
+      "step": 5804
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8344190467684087e-05,
+      "loss": 0.5984,
+      "step": 5805
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8343543896848275e-05,
+      "loss": 0.669,
+      "step": 5806
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8342897211197964e-05,
+      "loss": 0.6313,
+      "step": 5807
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834225041074206e-05,
+      "loss": 0.6034,
+      "step": 5808
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834160349548946e-05,
+      "loss": 0.7207,
+      "step": 5809
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8340956465449066e-05,
+      "loss": 0.6707,
+      "step": 5810
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834030932062978e-05,
+      "loss": 0.6473,
+      "step": 5811
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833966206104051e-05,
+      "loss": 0.75,
+      "step": 5812
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8339014686690166e-05,
+      "loss": 0.6844,
+      "step": 5813
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833836719758765e-05,
+      "loss": 0.636,
+      "step": 5814
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8337719593741874e-05,
+      "loss": 0.6166,
+      "step": 5815
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833707187516175e-05,
+      "loss": 0.6793,
+      "step": 5816
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8336424041856193e-05,
+      "loss": 0.7289,
+      "step": 5817
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8335776093834116e-05,
+      "loss": 0.6879,
+      "step": 5818
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8335128031104436e-05,
+      "loss": 0.6938,
+      "step": 5819
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8334479853676065e-05,
+      "loss": 0.7066,
+      "step": 5820
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8333831561557933e-05,
+      "loss": 0.7015,
+      "step": 5821
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8333183154758957e-05,
+      "loss": 0.6115,
+      "step": 5822
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8332534633288056e-05,
+      "loss": 0.6603,
+      "step": 5823
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8331885997154154e-05,
+      "loss": 0.6677,
+      "step": 5824
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8331237246366183e-05,
+      "loss": 0.6105,
+      "step": 5825
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833058838093307e-05,
+      "loss": 0.6158,
+      "step": 5826
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832993940086374e-05,
+      "loss": 0.6756,
+      "step": 5827
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832929030616712e-05,
+      "loss": 0.6516,
+      "step": 5828
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8328641096852147e-05,
+      "loss": 0.7419,
+      "step": 5829
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832799177292776e-05,
+      "loss": 0.6938,
+      "step": 5830
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8327342334402883e-05,
+      "loss": 0.7065,
+      "step": 5831
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832669278128646e-05,
+      "loss": 0.6853,
+      "step": 5832
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832604311358743e-05,
+      "loss": 0.6801,
+      "step": 5833
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8325393331314728e-05,
+      "loss": 0.7111,
+      "step": 5834
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8324743434477302e-05,
+      "loss": 0.7133,
+      "step": 5835
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8324093423084087e-05,
+      "loss": 0.6597,
+      "step": 5836
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8323443297144036e-05,
+      "loss": 0.5985,
+      "step": 5837
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832279305666609e-05,
+      "loss": 0.6685,
+      "step": 5838
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83221427016592e-05,
+      "loss": 0.616,
+      "step": 5839
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8321492232132312e-05,
+      "loss": 0.6574,
+      "step": 5840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832084164809438e-05,
+      "loss": 0.6908,
+      "step": 5841
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8320190949554356e-05,
+      "loss": 0.6779,
+      "step": 5842
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831954013652119e-05,
+      "loss": 0.6871,
+      "step": 5843
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8318889209003847e-05,
+      "loss": 0.7133,
+      "step": 5844
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8318238167011276e-05,
+      "loss": 0.6936,
+      "step": 5845
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831758701055244e-05,
+      "loss": 0.6919,
+      "step": 5846
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8316935739636298e-05,
+      "loss": 0.6846,
+      "step": 5847
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831628435427181e-05,
+      "loss": 0.5749,
+      "step": 5848
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8315632854467944e-05,
+      "loss": 0.7028,
+      "step": 5849
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8314981240233663e-05,
+      "loss": 0.6916,
+      "step": 5850
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8314329511577933e-05,
+      "loss": 0.7051,
+      "step": 5851
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8313677668509723e-05,
+      "loss": 0.7462,
+      "step": 5852
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8313025711038007e-05,
+      "loss": 0.6318,
+      "step": 5853
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831237363917175e-05,
+      "loss": 0.6115,
+      "step": 5854
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831172145291993e-05,
+      "loss": 0.5925,
+      "step": 5855
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8311069152291513e-05,
+      "loss": 0.7297,
+      "step": 5856
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8310416737295486e-05,
+      "loss": 0.6788,
+      "step": 5857
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8309764207940826e-05,
+      "loss": 0.7346,
+      "step": 5858
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8309111564236504e-05,
+      "loss": 0.7217,
+      "step": 5859
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8308458806191507e-05,
+      "loss": 0.7446,
+      "step": 5860
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8307805933814818e-05,
+      "loss": 0.6457,
+      "step": 5861
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8307152947115418e-05,
+      "loss": 0.6565,
+      "step": 5862
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8306499846102294e-05,
+      "loss": 0.6559,
+      "step": 5863
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8305846630784436e-05,
+      "loss": 0.7154,
+      "step": 5864
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8305193301170828e-05,
+      "loss": 0.6305,
+      "step": 5865
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8304539857270463e-05,
+      "loss": 0.6884,
+      "step": 5866
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830388629909233e-05,
+      "loss": 0.7095,
+      "step": 5867
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8303232626645425e-05,
+      "loss": 0.6886,
+      "step": 5868
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8302578839938748e-05,
+      "loss": 0.5817,
+      "step": 5869
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830192493898129e-05,
+      "loss": 0.6985,
+      "step": 5870
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830127092378205e-05,
+      "loss": 0.7291,
+      "step": 5871
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8300616794350025e-05,
+      "loss": 0.6383,
+      "step": 5872
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829996255069422e-05,
+      "loss": 0.5864,
+      "step": 5873
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829930819282364e-05,
+      "loss": 0.572,
+      "step": 5874
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8298653720747283e-05,
+      "loss": 0.575,
+      "step": 5875
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8297999134474162e-05,
+      "loss": 0.7182,
+      "step": 5876
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829734443401328e-05,
+      "loss": 0.5666,
+      "step": 5877
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829668961937365e-05,
+      "loss": 0.5796,
+      "step": 5878
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8296034690564276e-05,
+      "loss": 0.7081,
+      "step": 5879
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829537964759418e-05,
+      "loss": 0.6718,
+      "step": 5880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829472449047237e-05,
+      "loss": 0.6184,
+      "step": 5881
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829406921920786e-05,
+      "loss": 0.6352,
+      "step": 5882
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829341383380967e-05,
+      "loss": 0.6326,
+      "step": 5883
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8292758334286822e-05,
+      "loss": 0.6647,
+      "step": 5884
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8292102720648333e-05,
+      "loss": 0.7492,
+      "step": 5885
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829144699290322e-05,
+      "loss": 0.7077,
+      "step": 5886
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8290791151060513e-05,
+      "loss": 0.6213,
+      "step": 5887
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8290135195129237e-05,
+      "loss": 0.7448,
+      "step": 5888
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8289479125118412e-05,
+      "loss": 0.6672,
+      "step": 5889
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8288822941037074e-05,
+      "loss": 0.6753,
+      "step": 5890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828816664289425e-05,
+      "loss": 0.6254,
+      "step": 5891
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828751023069897e-05,
+      "loss": 0.7364,
+      "step": 5892
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8286853704460264e-05,
+      "loss": 0.6589,
+      "step": 5893
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828619706418717e-05,
+      "loss": 0.7357,
+      "step": 5894
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8285540309888724e-05,
+      "loss": 0.6956,
+      "step": 5895
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8284883441573968e-05,
+      "loss": 0.7239,
+      "step": 5896
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8284226459251928e-05,
+      "loss": 0.7199,
+      "step": 5897
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8283569362931657e-05,
+      "loss": 0.6733,
+      "step": 5898
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828291215262219e-05,
+      "loss": 0.6285,
+      "step": 5899
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8282254828332578e-05,
+      "loss": 0.6633,
+      "step": 5900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8281597390071862e-05,
+      "loss": 0.7808,
+      "step": 5901
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8280939837849087e-05,
+      "loss": 0.7378,
+      "step": 5902
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8280282171673303e-05,
+      "loss": 0.7406,
+      "step": 5903
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827962439155356e-05,
+      "loss": 0.6659,
+      "step": 5904
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8278966497498912e-05,
+      "loss": 0.582,
+      "step": 5905
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8278308489518412e-05,
+      "loss": 0.7061,
+      "step": 5906
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827765036762111e-05,
+      "loss": 0.6038,
+      "step": 5907
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827699213181607e-05,
+      "loss": 0.6103,
+      "step": 5908
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827633378211234e-05,
+      "loss": 0.672,
+      "step": 5909
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8275675318518988e-05,
+      "loss": 0.6761,
+      "step": 5910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8275016741045074e-05,
+      "loss": 0.754,
+      "step": 5911
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8274358049699656e-05,
+      "loss": 0.6683,
+      "step": 5912
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82736992444918e-05,
+      "loss": 0.631,
+      "step": 5913
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8273040325430575e-05,
+      "loss": 0.7518,
+      "step": 5914
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8272381292525044e-05,
+      "loss": 0.7494,
+      "step": 5915
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8271722145784277e-05,
+      "loss": 0.6346,
+      "step": 5916
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8271062885217346e-05,
+      "loss": 0.7524,
+      "step": 5917
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827040351083332e-05,
+      "loss": 0.5999,
+      "step": 5918
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8269744022641278e-05,
+      "loss": 0.6948,
+      "step": 5919
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826908442065029e-05,
+      "loss": 0.6888,
+      "step": 5920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8268424704869436e-05,
+      "loss": 0.6759,
+      "step": 5921
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826776487530779e-05,
+      "loss": 0.702,
+      "step": 5922
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826710493197444e-05,
+      "loss": 0.6708,
+      "step": 5923
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8266444874878456e-05,
+      "loss": 0.6609,
+      "step": 5924
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8265784704028934e-05,
+      "loss": 0.6464,
+      "step": 5925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8265124419434945e-05,
+      "loss": 0.6622,
+      "step": 5926
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826446402110558e-05,
+      "loss": 0.6915,
+      "step": 5927
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8263803509049936e-05,
+      "loss": 0.6836,
+      "step": 5928
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826314288327709e-05,
+      "loss": 0.7253,
+      "step": 5929
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8262482143796136e-05,
+      "loss": 0.6582,
+      "step": 5930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826182129061617e-05,
+      "loss": 0.6698,
+      "step": 5931
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8261160323746283e-05,
+      "loss": 0.6597,
+      "step": 5932
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826049924319557e-05,
+      "loss": 0.7028,
+      "step": 5933
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8259838048973132e-05,
+      "loss": 0.6476,
+      "step": 5934
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8259176741088062e-05,
+      "loss": 0.7434,
+      "step": 5935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8258515319549463e-05,
+      "loss": 0.668,
+      "step": 5936
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8257853784366436e-05,
+      "loss": 0.7025,
+      "step": 5937
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8257192135548087e-05,
+      "loss": 0.622,
+      "step": 5938
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8256530373103518e-05,
+      "loss": 0.7058,
+      "step": 5939
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8255868497041835e-05,
+      "loss": 0.6227,
+      "step": 5940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825520650737215e-05,
+      "loss": 0.5949,
+      "step": 5941
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825454440410357e-05,
+      "loss": 0.6903,
+      "step": 5942
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82538821872452e-05,
+      "loss": 0.6522,
+      "step": 5943
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8253219856806164e-05,
+      "loss": 0.7293,
+      "step": 5944
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8252557412795572e-05,
+      "loss": 0.5972,
+      "step": 5945
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8251894855222537e-05,
+      "loss": 0.7142,
+      "step": 5946
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8251232184096177e-05,
+      "loss": 0.6783,
+      "step": 5947
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8250569399425613e-05,
+      "loss": 0.6989,
+      "step": 5948
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8249906501219962e-05,
+      "loss": 0.6454,
+      "step": 5949
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8249243489488352e-05,
+      "loss": 0.554,
+      "step": 5950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82485803642399e-05,
+      "loss": 0.6762,
+      "step": 5951
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8247917125483735e-05,
+      "loss": 0.6908,
+      "step": 5952
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8247253773228987e-05,
+      "loss": 0.6041,
+      "step": 5953
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8246590307484776e-05,
+      "loss": 0.6893,
+      "step": 5954
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8245926728260238e-05,
+      "loss": 0.687,
+      "step": 5955
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8245263035564503e-05,
+      "loss": 0.666,
+      "step": 5956
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8244599229406702e-05,
+      "loss": 0.629,
+      "step": 5957
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8243935309795974e-05,
+      "loss": 0.6808,
+      "step": 5958
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824327127674145e-05,
+      "loss": 0.7684,
+      "step": 5959
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824260713025227e-05,
+      "loss": 0.6321,
+      "step": 5960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8241942870337575e-05,
+      "loss": 0.7055,
+      "step": 5961
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8241278497006503e-05,
+      "loss": 0.6779,
+      "step": 5962
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82406140102682e-05,
+      "loss": 0.6445,
+      "step": 5963
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8239949410131803e-05,
+      "loss": 0.6958,
+      "step": 5964
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8239284696606465e-05,
+      "loss": 0.692,
+      "step": 5965
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8238619869701328e-05,
+      "loss": 0.7032,
+      "step": 5966
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8237954929425544e-05,
+      "loss": 0.6089,
+      "step": 5967
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823728987578826e-05,
+      "loss": 0.655,
+      "step": 5968
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823662470879863e-05,
+      "loss": 0.629,
+      "step": 5969
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8235959428465806e-05,
+      "loss": 0.6829,
+      "step": 5970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8235294034798948e-05,
+      "loss": 0.6895,
+      "step": 5971
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8234628527807202e-05,
+      "loss": 0.7225,
+      "step": 5972
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8233962907499737e-05,
+      "loss": 0.6846,
+      "step": 5973
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8233297173885702e-05,
+      "loss": 0.6443,
+      "step": 5974
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8232631326974268e-05,
+      "loss": 0.6839,
+      "step": 5975
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8231965366774592e-05,
+      "loss": 0.6063,
+      "step": 5976
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8231299293295837e-05,
+      "loss": 0.5959,
+      "step": 5977
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823063310654717e-05,
+      "loss": 0.6515,
+      "step": 5978
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8229966806537766e-05,
+      "loss": 0.7153,
+      "step": 5979
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8229300393276783e-05,
+      "loss": 0.7041,
+      "step": 5980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8228633866773396e-05,
+      "loss": 0.6326,
+      "step": 5981
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822796722703678e-05,
+      "loss": 0.7128,
+      "step": 5982
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8227300474076098e-05,
+      "loss": 0.6701,
+      "step": 5983
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822663360790054e-05,
+      "loss": 0.7042,
+      "step": 5984
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822596662851927e-05,
+      "loss": 0.712,
+      "step": 5985
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8225299535941473e-05,
+      "loss": 0.6451,
+      "step": 5986
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8224632330176325e-05,
+      "loss": 0.6885,
+      "step": 5987
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822396501123301e-05,
+      "loss": 0.6909,
+      "step": 5988
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8223297579120713e-05,
+      "loss": 0.6566,
+      "step": 5989
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8222630033848615e-05,
+      "loss": 0.732,
+      "step": 5990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8221962375425902e-05,
+      "loss": 0.7621,
+      "step": 5991
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822129460386176e-05,
+      "loss": 0.7114,
+      "step": 5992
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8220626719165383e-05,
+      "loss": 0.7535,
+      "step": 5993
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8219958721345957e-05,
+      "loss": 0.7472,
+      "step": 5994
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821929061041268e-05,
+      "loss": 0.7055,
+      "step": 5995
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821862238637474e-05,
+      "loss": 0.7005,
+      "step": 5996
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8217954049241334e-05,
+      "loss": 0.5797,
+      "step": 5997
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8217285599021658e-05,
+      "loss": 0.7244,
+      "step": 5998
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8216617035724913e-05,
+      "loss": 0.6286,
+      "step": 5999
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82159483593603e-05,
+      "loss": 0.623,
+      "step": 6000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8215279569937016e-05,
+      "loss": 0.6763,
+      "step": 6001
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8214610667464264e-05,
+      "loss": 0.5913,
+      "step": 6002
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8213941651951255e-05,
+      "loss": 0.7128,
+      "step": 6003
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8213272523407192e-05,
+      "loss": 0.7085,
+      "step": 6004
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8212603281841283e-05,
+      "loss": 0.7684,
+      "step": 6005
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8211933927262734e-05,
+      "loss": 0.6188,
+      "step": 6006
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8211264459680757e-05,
+      "loss": 0.6942,
+      "step": 6007
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821059487910457e-05,
+      "loss": 0.6718,
+      "step": 6008
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820992518554338e-05,
+      "loss": 0.6893,
+      "step": 6009
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820925537900641e-05,
+      "loss": 0.6745,
+      "step": 6010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8208585459502868e-05,
+      "loss": 0.7522,
+      "step": 6011
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8207915427041977e-05,
+      "loss": 0.7205,
+      "step": 6012
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820724528163296e-05,
+      "loss": 0.7348,
+      "step": 6013
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8206575023285035e-05,
+      "loss": 0.5786,
+      "step": 6014
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820590465200743e-05,
+      "loss": 0.7654,
+      "step": 6015
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8205234167809363e-05,
+      "loss": 0.6931,
+      "step": 6016
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8204563570700063e-05,
+      "loss": 0.6839,
+      "step": 6017
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8203892860688763e-05,
+      "loss": 0.7088,
+      "step": 6018
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820322203778469e-05,
+      "loss": 0.7265,
+      "step": 6019
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8202551101997068e-05,
+      "loss": 0.6946,
+      "step": 6020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820188005333514e-05,
+      "loss": 0.8001,
+      "step": 6021
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8201208891808132e-05,
+      "loss": 0.6281,
+      "step": 6022
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8200537617425288e-05,
+      "loss": 0.6964,
+      "step": 6023
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819986623019584e-05,
+      "loss": 0.7394,
+      "step": 6024
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8199194730129025e-05,
+      "loss": 0.7328,
+      "step": 6025
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8198523117234085e-05,
+      "loss": 0.6203,
+      "step": 6026
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8197851391520265e-05,
+      "loss": 0.6898,
+      "step": 6027
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8197179552996805e-05,
+      "loss": 0.6281,
+      "step": 6028
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8196507601672954e-05,
+      "loss": 0.6693,
+      "step": 6029
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819583553755796e-05,
+      "loss": 0.7109,
+      "step": 6030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819516336066106e-05,
+      "loss": 0.6901,
+      "step": 6031
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8194491070991517e-05,
+      "loss": 0.8044,
+      "step": 6032
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8193818668558573e-05,
+      "loss": 0.6183,
+      "step": 6033
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8193146153371483e-05,
+      "loss": 0.638,
+      "step": 6034
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8192473525439506e-05,
+      "loss": 0.6718,
+      "step": 6035
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8191800784771893e-05,
+      "loss": 0.6574,
+      "step": 6036
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8191127931377903e-05,
+      "loss": 0.6048,
+      "step": 6037
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81904549652668e-05,
+      "loss": 0.6293,
+      "step": 6038
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8189781886447834e-05,
+      "loss": 0.6446,
+      "step": 6039
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818910869493028e-05,
+      "loss": 0.5771,
+      "step": 6040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8188435390723386e-05,
+      "loss": 0.6883,
+      "step": 6041
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818776197383643e-05,
+      "loss": 0.7528,
+      "step": 6042
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8187088444278675e-05,
+      "loss": 0.7216,
+      "step": 6043
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818641480205939e-05,
+      "loss": 0.6459,
+      "step": 6044
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8185741047187843e-05,
+      "loss": 0.6133,
+      "step": 6045
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818506717967331e-05,
+      "loss": 0.6771,
+      "step": 6046
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8184393199525054e-05,
+      "loss": 0.6745,
+      "step": 6047
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818371910675236e-05,
+      "loss": 0.7127,
+      "step": 6048
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8183044901364502e-05,
+      "loss": 0.5605,
+      "step": 6049
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8182370583370753e-05,
+      "loss": 0.7302,
+      "step": 6050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8181696152780395e-05,
+      "loss": 0.6436,
+      "step": 6051
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8181021609602712e-05,
+      "loss": 0.7075,
+      "step": 6052
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818034695384698e-05,
+      "loss": 0.7516,
+      "step": 6053
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8179672185522487e-05,
+      "loss": 0.7848,
+      "step": 6054
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8178997304638515e-05,
+      "loss": 0.6839,
+      "step": 6055
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817832231120436e-05,
+      "loss": 0.6255,
+      "step": 6056
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8177647205229295e-05,
+      "loss": 0.6782,
+      "step": 6057
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8176971986722625e-05,
+      "loss": 0.6489,
+      "step": 6058
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8176296655693635e-05,
+      "loss": 0.688,
+      "step": 6059
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8175621212151616e-05,
+      "loss": 0.6871,
+      "step": 6060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817494565610587e-05,
+      "loss": 0.6675,
+      "step": 6061
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8174269987565686e-05,
+      "loss": 0.7338,
+      "step": 6062
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8173594206540363e-05,
+      "loss": 0.6988,
+      "step": 6063
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8172918313039206e-05,
+      "loss": 0.6821,
+      "step": 6064
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8172242307071507e-05,
+      "loss": 0.6341,
+      "step": 6065
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8171566188646575e-05,
+      "loss": 0.6943,
+      "step": 6066
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817088995777371e-05,
+      "loss": 0.7088,
+      "step": 6067
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8170213614462223e-05,
+      "loss": 0.6063,
+      "step": 6068
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8169537158721415e-05,
+      "loss": 0.6591,
+      "step": 6069
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81688605905606e-05,
+      "loss": 0.6488,
+      "step": 6070
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8168183909989085e-05,
+      "loss": 0.6573,
+      "step": 6071
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816750711701618e-05,
+      "loss": 0.6326,
+      "step": 6072
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81668302116512e-05,
+      "loss": 0.7055,
+      "step": 6073
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8166153193903464e-05,
+      "loss": 0.5725,
+      "step": 6074
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816547606378228e-05,
+      "loss": 0.5992,
+      "step": 6075
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8164798821296977e-05,
+      "loss": 0.7139,
+      "step": 6076
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816412146645686e-05,
+      "loss": 0.6714,
+      "step": 6077
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8163443999271267e-05,
+      "loss": 0.6687,
+      "step": 6078
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8162766419749503e-05,
+      "loss": 0.616,
+      "step": 6079
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8162088727900905e-05,
+      "loss": 0.7068,
+      "step": 6080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8161410923734795e-05,
+      "loss": 0.6974,
+      "step": 6081
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81607330072605e-05,
+      "loss": 0.5684,
+      "step": 6082
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8160054978487345e-05,
+      "loss": 0.7272,
+      "step": 6083
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8159376837424665e-05,
+      "loss": 0.701,
+      "step": 6084
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8158698584081788e-05,
+      "loss": 0.7,
+      "step": 6085
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815802021846805e-05,
+      "loss": 0.6486,
+      "step": 6086
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815734174059279e-05,
+      "loss": 0.6515,
+      "step": 6087
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8156663150465338e-05,
+      "loss": 0.7257,
+      "step": 6088
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8155984448095032e-05,
+      "loss": 0.6693,
+      "step": 6089
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8155305633491217e-05,
+      "loss": 0.6837,
+      "step": 6090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8154626706663225e-05,
+      "loss": 0.6242,
+      "step": 6091
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815394766762041e-05,
+      "loss": 0.7011,
+      "step": 6092
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8153268516372105e-05,
+      "loss": 0.6092,
+      "step": 6093
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8152589252927666e-05,
+      "loss": 0.6858,
+      "step": 6094
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815190987729643e-05,
+      "loss": 0.6844,
+      "step": 6095
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8151230389487757e-05,
+      "loss": 0.7129,
+      "step": 6096
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815055078951099e-05,
+      "loss": 0.734,
+      "step": 6097
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8149871077375482e-05,
+      "loss": 0.6579,
+      "step": 6098
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8149191253090586e-05,
+      "loss": 0.7015,
+      "step": 6099
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8148511316665657e-05,
+      "loss": 0.6061,
+      "step": 6100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8147831268110052e-05,
+      "loss": 0.7321,
+      "step": 6101
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8147151107433133e-05,
+      "loss": 0.717,
+      "step": 6102
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8146470834644255e-05,
+      "loss": 0.7072,
+      "step": 6103
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8145790449752778e-05,
+      "loss": 0.6866,
+      "step": 6104
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8145109952768068e-05,
+      "loss": 0.7253,
+      "step": 6105
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8144429343699486e-05,
+      "loss": 0.6478,
+      "step": 6106
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8143748622556404e-05,
+      "loss": 0.6407,
+      "step": 6107
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8143067789348183e-05,
+      "loss": 0.6387,
+      "step": 6108
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8142386844084195e-05,
+      "loss": 0.6053,
+      "step": 6109
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814170578677381e-05,
+      "loss": 0.603,
+      "step": 6110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81410246174264e-05,
+      "loss": 0.719,
+      "step": 6111
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8140343336051334e-05,
+      "loss": 0.6844,
+      "step": 6112
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8139661942657995e-05,
+      "loss": 0.6093,
+      "step": 6113
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8138980437255754e-05,
+      "loss": 0.7063,
+      "step": 6114
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813829881985399e-05,
+      "loss": 0.6612,
+      "step": 6115
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8137617090462082e-05,
+      "loss": 0.5362,
+      "step": 6116
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8136935249089416e-05,
+      "loss": 0.6096,
+      "step": 6117
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813625329574537e-05,
+      "loss": 0.6633,
+      "step": 6118
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813557123043933e-05,
+      "loss": 0.6289,
+      "step": 6119
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8134889053180677e-05,
+      "loss": 0.6735,
+      "step": 6120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813420676397881e-05,
+      "loss": 0.7126,
+      "step": 6121
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8133524362843105e-05,
+      "loss": 0.6322,
+      "step": 6122
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8132841849782957e-05,
+      "loss": 0.7103,
+      "step": 6123
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8132159224807765e-05,
+      "loss": 0.5984,
+      "step": 6124
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8131476487926914e-05,
+      "loss": 0.623,
+      "step": 6125
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8130793639149798e-05,
+      "loss": 0.6979,
+      "step": 6126
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813011067848582e-05,
+      "loss": 0.7445,
+      "step": 6127
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812942760594438e-05,
+      "loss": 0.6573,
+      "step": 6128
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812874442153487e-05,
+      "loss": 0.5668,
+      "step": 6129
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8128061125266686e-05,
+      "loss": 0.6705,
+      "step": 6130
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812737771714925e-05,
+      "loss": 0.6805,
+      "step": 6131
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8126694197191947e-05,
+      "loss": 0.6966,
+      "step": 6132
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81260105654042e-05,
+      "loss": 0.6325,
+      "step": 6133
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81253268217954e-05,
+      "loss": 0.5915,
+      "step": 6134
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8124642966374964e-05,
+      "loss": 0.5975,
+      "step": 6135
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8123958999152305e-05,
+      "loss": 0.7352,
+      "step": 6136
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812327492013683e-05,
+      "loss": 0.6645,
+      "step": 6137
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8122590729337957e-05,
+      "loss": 0.619,
+      "step": 6138
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8121906426765096e-05,
+      "loss": 0.7416,
+      "step": 6139
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8121222012427666e-05,
+      "loss": 0.7163,
+      "step": 6140
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8120537486335085e-05,
+      "loss": 0.6644,
+      "step": 6141
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8119852848496774e-05,
+      "loss": 0.5929,
+      "step": 6142
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8119168098922152e-05,
+      "loss": 0.7232,
+      "step": 6143
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8118483237620643e-05,
+      "loss": 0.5997,
+      "step": 6144
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811779826460167e-05,
+      "loss": 0.5974,
+      "step": 6145
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811711317987466e-05,
+      "loss": 0.7108,
+      "step": 6146
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8116427983449045e-05,
+      "loss": 0.68,
+      "step": 6147
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8115742675334244e-05,
+      "loss": 0.7355,
+      "step": 6148
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8115057255539694e-05,
+      "loss": 0.7098,
+      "step": 6149
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8114371724074827e-05,
+      "loss": 0.5853,
+      "step": 6150
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8113686080949076e-05,
+      "loss": 0.6564,
+      "step": 6151
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8113000326171874e-05,
+      "loss": 0.7095,
+      "step": 6152
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811231445975266e-05,
+      "loss": 0.6464,
+      "step": 6153
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8111628481700865e-05,
+      "loss": 0.5798,
+      "step": 6154
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811094239202594e-05,
+      "loss": 0.6532,
+      "step": 6155
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811025619073732e-05,
+      "loss": 0.6637,
+      "step": 6156
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8109569877844447e-05,
+      "loss": 0.5689,
+      "step": 6157
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810888345335677e-05,
+      "loss": 0.5726,
+      "step": 6158
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810819691728373e-05,
+      "loss": 0.7019,
+      "step": 6159
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8107510269634775e-05,
+      "loss": 0.7667,
+      "step": 6160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8106823510419354e-05,
+      "loss": 0.5851,
+      "step": 6161
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810613663964692e-05,
+      "loss": 0.6878,
+      "step": 6162
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8105449657326924e-05,
+      "loss": 0.6932,
+      "step": 6163
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8104762563468816e-05,
+      "loss": 0.6619,
+      "step": 6164
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8104075358082057e-05,
+      "loss": 0.6801,
+      "step": 6165
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81033880411761e-05,
+      "loss": 0.7194,
+      "step": 6166
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81027006127604e-05,
+      "loss": 0.7105,
+      "step": 6167
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810201307284442e-05,
+      "loss": 0.7298,
+      "step": 6168
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8101325421437622e-05,
+      "loss": 0.6736,
+      "step": 6169
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810063765854947e-05,
+      "loss": 0.6096,
+      "step": 6170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8099949784189425e-05,
+      "loss": 0.5465,
+      "step": 6171
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8099261798366952e-05,
+      "loss": 0.6403,
+      "step": 6172
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809857370109152e-05,
+      "loss": 0.6595,
+      "step": 6173
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.80978854923726e-05,
+      "loss": 0.7257,
+      "step": 6174
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809719717221966e-05,
+      "loss": 0.6644,
+      "step": 6175
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809650874064217e-05,
+      "loss": 0.723,
+      "step": 6176
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8095820197649607e-05,
+      "loss": 0.6556,
+      "step": 6177
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8095131543251442e-05,
+      "loss": 0.6076,
+      "step": 6178
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809444277745716e-05,
+      "loss": 0.609,
+      "step": 6179
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8093753900276226e-05,
+      "loss": 0.578,
+      "step": 6180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809306491171813e-05,
+      "loss": 0.5953,
+      "step": 6181
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809237581179235e-05,
+      "loss": 0.7109,
+      "step": 6182
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809168660050837e-05,
+      "loss": 0.7528,
+      "step": 6183
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8090997277875665e-05,
+      "loss": 0.6815,
+      "step": 6184
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8090307843903738e-05,
+      "loss": 0.6819,
+      "step": 6185
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8089618298602058e-05,
+      "loss": 0.6735,
+      "step": 6186
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8088928641980127e-05,
+      "loss": 0.6952,
+      "step": 6187
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8088238874047428e-05,
+      "loss": 0.7837,
+      "step": 6188
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8087548994813455e-05,
+      "loss": 0.6937,
+      "step": 6189
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8086859004287704e-05,
+      "loss": 0.6729,
+      "step": 6190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8086168902479665e-05,
+      "loss": 0.6618,
+      "step": 6191
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8085478689398836e-05,
+      "loss": 0.6741,
+      "step": 6192
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8084788365054718e-05,
+      "loss": 0.6929,
+      "step": 6193
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808409792945681e-05,
+      "loss": 0.6644,
+      "step": 6194
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8083407382614604e-05,
+      "loss": 0.7387,
+      "step": 6195
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8082716724537617e-05,
+      "loss": 0.7155,
+      "step": 6196
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8082025955235344e-05,
+      "loss": 0.6451,
+      "step": 6197
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808133507471729e-05,
+      "loss": 0.6655,
+      "step": 6198
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8080644082992967e-05,
+      "loss": 0.5152,
+      "step": 6199
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807995298007188e-05,
+      "loss": 0.5926,
+      "step": 6200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8079261765963537e-05,
+      "loss": 0.6166,
+      "step": 6201
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8078570440677457e-05,
+      "loss": 0.6923,
+      "step": 6202
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8077879004223148e-05,
+      "loss": 0.7484,
+      "step": 6203
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8077187456610127e-05,
+      "loss": 0.7185,
+      "step": 6204
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807649579784791e-05,
+      "loss": 0.5917,
+      "step": 6205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8075804027946014e-05,
+      "loss": 0.6366,
+      "step": 6206
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8075112146913957e-05,
+      "loss": 0.6529,
+      "step": 6207
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8074420154761263e-05,
+      "loss": 0.635,
+      "step": 6208
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807372805149745e-05,
+      "loss": 0.6891,
+      "step": 6209
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8073035837132047e-05,
+      "loss": 0.759,
+      "step": 6210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8072343511674576e-05,
+      "loss": 0.5858,
+      "step": 6211
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8071651075134568e-05,
+      "loss": 0.7247,
+      "step": 6212
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8070958527521547e-05,
+      "loss": 0.7711,
+      "step": 6213
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8070265868845048e-05,
+      "loss": 0.6773,
+      "step": 6214
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8069573099114595e-05,
+      "loss": 0.7056,
+      "step": 6215
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8068880218339728e-05,
+      "loss": 0.7359,
+      "step": 6216
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806818722652998e-05,
+      "loss": 0.5986,
+      "step": 6217
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8067494123694884e-05,
+      "loss": 0.6898,
+      "step": 6218
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806680090984398e-05,
+      "loss": 0.6036,
+      "step": 6219
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806610758498681e-05,
+      "loss": 0.6179,
+      "step": 6220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806541414913291e-05,
+      "loss": 0.6798,
+      "step": 6221
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8064720602291825e-05,
+      "loss": 0.5716,
+      "step": 6222
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80640269444731e-05,
+      "loss": 0.7024,
+      "step": 6223
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8063333175686275e-05,
+      "loss": 0.7392,
+      "step": 6224
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8062639295940903e-05,
+      "loss": 0.7012,
+      "step": 6225
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8061945305246527e-05,
+      "loss": 0.6555,
+      "step": 6226
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8061251203612702e-05,
+      "loss": 0.631,
+      "step": 6227
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806055699104898e-05,
+      "loss": 0.6609,
+      "step": 6228
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8059862667564906e-05,
+      "loss": 0.6171,
+      "step": 6229
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8059168233170045e-05,
+      "loss": 0.6482,
+      "step": 6230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805847368787394e-05,
+      "loss": 0.6318,
+      "step": 6231
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8057779031686162e-05,
+      "loss": 0.7064,
+      "step": 6232
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8057084264616264e-05,
+      "loss": 0.599,
+      "step": 6233
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8056389386673805e-05,
+      "loss": 0.7033,
+      "step": 6234
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8055694397868354e-05,
+      "loss": 0.6961,
+      "step": 6235
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8054999298209465e-05,
+      "loss": 0.5716,
+      "step": 6236
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8054304087706706e-05,
+      "loss": 0.7041,
+      "step": 6237
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805360876636965e-05,
+      "loss": 0.7087,
+      "step": 6238
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805291333420786e-05,
+      "loss": 0.7131,
+      "step": 6239
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8052217791230908e-05,
+      "loss": 0.6081,
+      "step": 6240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805152213744836e-05,
+      "loss": 0.7211,
+      "step": 6241
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8050826372869793e-05,
+      "loss": 0.7261,
+      "step": 6242
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8050130497504782e-05,
+      "loss": 0.685,
+      "step": 6243
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8049434511362904e-05,
+      "loss": 0.5718,
+      "step": 6244
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8048738414453732e-05,
+      "loss": 0.6645,
+      "step": 6245
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804804220678685e-05,
+      "loss": 0.638,
+      "step": 6246
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8047345888371833e-05,
+      "loss": 0.7053,
+      "step": 6247
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8046649459218267e-05,
+      "loss": 0.6968,
+      "step": 6248
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8045952919335732e-05,
+      "loss": 0.6786,
+      "step": 6249
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8045256268733818e-05,
+      "loss": 0.7292,
+      "step": 6250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8044559507422105e-05,
+      "loss": 0.7372,
+      "step": 6251
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8043862635410186e-05,
+      "loss": 0.7556,
+      "step": 6252
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804316565270765e-05,
+      "loss": 0.61,
+      "step": 6253
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8042468559324086e-05,
+      "loss": 0.6931,
+      "step": 6254
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804177135526909e-05,
+      "loss": 0.5854,
+      "step": 6255
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8041074040552253e-05,
+      "loss": 0.736,
+      "step": 6256
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804037661518317e-05,
+      "loss": 0.589,
+      "step": 6257
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8039679079171443e-05,
+      "loss": 0.7373,
+      "step": 6258
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8038981432526664e-05,
+      "loss": 0.681,
+      "step": 6259
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803828367525844e-05,
+      "loss": 0.6397,
+      "step": 6260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8037585807376366e-05,
+      "loss": 0.7261,
+      "step": 6261
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803688782889005e-05,
+      "loss": 0.6433,
+      "step": 6262
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80361897398091e-05,
+      "loss": 0.7152,
+      "step": 6263
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8035491540143114e-05,
+      "loss": 0.6162,
+      "step": 6264
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8034793229901706e-05,
+      "loss": 0.702,
+      "step": 6265
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803409480909448e-05,
+      "loss": 0.6235,
+      "step": 6266
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8033396277731053e-05,
+      "loss": 0.608,
+      "step": 6267
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8032697635821033e-05,
+      "loss": 0.7317,
+      "step": 6268
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8031998883374036e-05,
+      "loss": 0.7115,
+      "step": 6269
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8031300020399675e-05,
+      "loss": 0.7133,
+      "step": 6270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803060104690757e-05,
+      "loss": 0.7106,
+      "step": 6271
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802990196290734e-05,
+      "loss": 0.7431,
+      "step": 6272
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80292027684086e-05,
+      "loss": 0.7278,
+      "step": 6273
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8028503463420975e-05,
+      "loss": 0.6238,
+      "step": 6274
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802780404795409e-05,
+      "loss": 0.686,
+      "step": 6275
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802710452201757e-05,
+      "loss": 0.6457,
+      "step": 6276
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8026404885621036e-05,
+      "loss": 0.6352,
+      "step": 6277
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8025705138774117e-05,
+      "loss": 0.6258,
+      "step": 6278
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8025005281486444e-05,
+      "loss": 0.5678,
+      "step": 6279
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8024305313767648e-05,
+      "loss": 0.638,
+      "step": 6280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8023605235627356e-05,
+      "loss": 0.699,
+      "step": 6281
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802290504707521e-05,
+      "loss": 0.6517,
+      "step": 6282
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802220474812084e-05,
+      "loss": 0.6007,
+      "step": 6283
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8021504338773884e-05,
+      "loss": 0.6991,
+      "step": 6284
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802080381904398e-05,
+      "loss": 0.6387,
+      "step": 6285
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8020103188940767e-05,
+      "loss": 0.7333,
+      "step": 6286
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8019402448473887e-05,
+      "loss": 0.5626,
+      "step": 6287
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8018701597652987e-05,
+      "loss": 0.6982,
+      "step": 6288
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80180006364877e-05,
+      "loss": 0.6835,
+      "step": 6289
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801729956498768e-05,
+      "loss": 0.6935,
+      "step": 6290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8016598383162578e-05,
+      "loss": 0.6988,
+      "step": 6291
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8015897091022035e-05,
+      "loss": 0.707,
+      "step": 6292
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801519568857571e-05,
+      "loss": 0.689,
+      "step": 6293
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8014494175833244e-05,
+      "loss": 0.6603,
+      "step": 6294
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8013792552804298e-05,
+      "loss": 0.6123,
+      "step": 6295
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8013090819498523e-05,
+      "loss": 0.6787,
+      "step": 6296
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801238897592558e-05,
+      "loss": 0.7426,
+      "step": 6297
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8011687022095122e-05,
+      "loss": 0.6794,
+      "step": 6298
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8010984958016812e-05,
+      "loss": 0.6506,
+      "step": 6299
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8010282783700307e-05,
+      "loss": 0.7663,
+      "step": 6300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8009580499155276e-05,
+      "loss": 0.7181,
+      "step": 6301
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8008878104391382e-05,
+      "loss": 0.619,
+      "step": 6302
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800817559941828e-05,
+      "loss": 0.7432,
+      "step": 6303
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800747298424565e-05,
+      "loss": 0.6804,
+      "step": 6304
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800677025888316e-05,
+      "loss": 0.7025,
+      "step": 6305
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800606742334047e-05,
+      "loss": 0.6881,
+      "step": 6306
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800536447762726e-05,
+      "loss": 0.5786,
+      "step": 6307
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80046614217532e-05,
+      "loss": 0.6166,
+      "step": 6308
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8003958255727965e-05,
+      "loss": 0.6799,
+      "step": 6309
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800325497956123e-05,
+      "loss": 0.5843,
+      "step": 6310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8002551593262675e-05,
+      "loss": 0.5877,
+      "step": 6311
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800184809684198e-05,
+      "loss": 0.5955,
+      "step": 6312
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8001144490308823e-05,
+      "loss": 0.6471,
+      "step": 6313
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8000440773672888e-05,
+      "loss": 0.5989,
+      "step": 6314
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7999736946943856e-05,
+      "loss": 0.7122,
+      "step": 6315
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7999033010131417e-05,
+      "loss": 0.6949,
+      "step": 6316
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7998328963245254e-05,
+      "loss": 0.6499,
+      "step": 6317
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799762480629506e-05,
+      "loss": 0.6182,
+      "step": 6318
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7996920539290518e-05,
+      "loss": 0.6778,
+      "step": 6319
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7996216162241324e-05,
+      "loss": 0.6437,
+      "step": 6320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7995511675157166e-05,
+      "loss": 0.6485,
+      "step": 6321
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7994807078047745e-05,
+      "loss": 0.694,
+      "step": 6322
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799410237092275e-05,
+      "loss": 0.6601,
+      "step": 6323
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799339755379189e-05,
+      "loss": 0.7075,
+      "step": 6324
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799269262666485e-05,
+      "loss": 0.6901,
+      "step": 6325
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7991987589551338e-05,
+      "loss": 0.6653,
+      "step": 6326
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7991282442461052e-05,
+      "loss": 0.5996,
+      "step": 6327
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7990577185403702e-05,
+      "loss": 0.5559,
+      "step": 6328
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7989871818388986e-05,
+      "loss": 0.7065,
+      "step": 6329
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7989166341426615e-05,
+      "loss": 0.7108,
+      "step": 6330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7988460754526293e-05,
+      "loss": 0.6974,
+      "step": 6331
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7987755057697734e-05,
+      "loss": 0.7555,
+      "step": 6332
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7987049250950644e-05,
+      "loss": 0.6743,
+      "step": 6333
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798634333429474e-05,
+      "loss": 0.6678,
+      "step": 6334
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7985637307739735e-05,
+      "loss": 0.634,
+      "step": 6335
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7984931171295344e-05,
+      "loss": 0.6743,
+      "step": 6336
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7984224924971287e-05,
+      "loss": 0.6358,
+      "step": 6337
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7983518568777275e-05,
+      "loss": 0.572,
+      "step": 6338
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798281210272303e-05,
+      "loss": 0.6875,
+      "step": 6339
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798210552681828e-05,
+      "loss": 0.6341,
+      "step": 6340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7981398841072747e-05,
+      "loss": 0.6914,
+      "step": 6341
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7980692045496146e-05,
+      "loss": 0.7652,
+      "step": 6342
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7979985140098212e-05,
+      "loss": 0.6984,
+      "step": 6343
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7979278124888675e-05,
+      "loss": 0.6712,
+      "step": 6344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7978570999877256e-05,
+      "loss": 0.7061,
+      "step": 6345
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7977863765073692e-05,
+      "loss": 0.7367,
+      "step": 6346
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797715642048771e-05,
+      "loss": 0.5866,
+      "step": 6347
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7976448966129047e-05,
+      "loss": 0.6928,
+      "step": 6348
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7975741402007438e-05,
+      "loss": 0.6607,
+      "step": 6349
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7975033728132622e-05,
+      "loss": 0.7359,
+      "step": 6350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7974325944514327e-05,
+      "loss": 0.6574,
+      "step": 6351
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7973618051162307e-05,
+      "loss": 0.7047,
+      "step": 6352
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7972910048086294e-05,
+      "loss": 0.7057,
+      "step": 6353
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7972201935296033e-05,
+      "loss": 0.671,
+      "step": 6354
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7971493712801268e-05,
+      "loss": 0.6503,
+      "step": 6355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7970785380611747e-05,
+      "loss": 0.6849,
+      "step": 6356
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7970076938737214e-05,
+      "loss": 0.6917,
+      "step": 6357
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7969368387187417e-05,
+      "loss": 0.7052,
+      "step": 6358
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7968659725972113e-05,
+      "loss": 0.6593,
+      "step": 6359
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7967950955101047e-05,
+      "loss": 0.6715,
+      "step": 6360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7967242074583976e-05,
+      "loss": 0.6911,
+      "step": 6361
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796653308443065e-05,
+      "loss": 0.6166,
+      "step": 6362
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7965823984650834e-05,
+      "loss": 0.6629,
+      "step": 6363
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7965114775254274e-05,
+      "loss": 0.6966,
+      "step": 6364
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796440545625074e-05,
+      "loss": 0.6981,
+      "step": 6365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796369602764999e-05,
+      "loss": 0.747,
+      "step": 6366
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796298648946178e-05,
+      "loss": 0.6287,
+      "step": 6367
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7962276841695884e-05,
+      "loss": 0.6826,
+      "step": 6368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7961567084362063e-05,
+      "loss": 0.6618,
+      "step": 6369
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7960857217470077e-05,
+      "loss": 0.7417,
+      "step": 6370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7960147241029706e-05,
+      "loss": 0.6261,
+      "step": 6371
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7959437155050713e-05,
+      "loss": 0.6359,
+      "step": 6372
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795872695954287e-05,
+      "loss": 0.6583,
+      "step": 6373
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7958016654515952e-05,
+      "loss": 0.6476,
+      "step": 6374
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7957306239979733e-05,
+      "loss": 0.667,
+      "step": 6375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7956595715943987e-05,
+      "loss": 0.6571,
+      "step": 6376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7955885082418495e-05,
+      "loss": 0.7237,
+      "step": 6377
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795517433941303e-05,
+      "loss": 0.6433,
+      "step": 6378
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795446348693738e-05,
+      "loss": 0.7286,
+      "step": 6379
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795375252500132e-05,
+      "loss": 0.7666,
+      "step": 6380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7953041453614635e-05,
+      "loss": 0.7085,
+      "step": 6381
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7952330272787116e-05,
+      "loss": 0.6626,
+      "step": 6382
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7951618982528544e-05,
+      "loss": 0.6781,
+      "step": 6383
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7950907582848708e-05,
+      "loss": 0.5573,
+      "step": 6384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795019607375739e-05,
+      "loss": 0.714,
+      "step": 6385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7949484455264398e-05,
+      "loss": 0.7124,
+      "step": 6386
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794877272737951e-05,
+      "loss": 0.5642,
+      "step": 6387
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7948060890112527e-05,
+      "loss": 0.6839,
+      "step": 6388
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794734894347324e-05,
+      "loss": 0.6843,
+      "step": 6389
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7946636887471453e-05,
+      "loss": 0.6407,
+      "step": 6390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7945924722116956e-05,
+      "loss": 0.6337,
+      "step": 6391
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7945212447419552e-05,
+      "loss": 0.7197,
+      "step": 6392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7944500063389046e-05,
+      "loss": 0.6854,
+      "step": 6393
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7943787570035237e-05,
+      "loss": 0.6828,
+      "step": 6394
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7943074967367936e-05,
+      "loss": 0.6831,
+      "step": 6395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7942362255396936e-05,
+      "loss": 0.6321,
+      "step": 6396
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7941649434132058e-05,
+      "loss": 0.7002,
+      "step": 6397
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7940936503583105e-05,
+      "loss": 0.7226,
+      "step": 6398
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7940223463759884e-05,
+      "loss": 0.6986,
+      "step": 6399
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793951031467222e-05,
+      "loss": 0.6764,
+      "step": 6400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7938797056329907e-05,
+      "loss": 0.6323,
+      "step": 6401
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7938083688742778e-05,
+      "loss": 0.6166,
+      "step": 6402
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793737021192064e-05,
+      "loss": 0.7267,
+      "step": 6403
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7936656625873316e-05,
+      "loss": 0.6817,
+      "step": 6404
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793594293061062e-05,
+      "loss": 0.5953,
+      "step": 6405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7935229126142376e-05,
+      "loss": 0.6793,
+      "step": 6406
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7934515212478408e-05,
+      "loss": 0.6468,
+      "step": 6407
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793380118962854e-05,
+      "loss": 0.6656,
+      "step": 6408
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7933087057602595e-05,
+      "loss": 0.7055,
+      "step": 6409
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79323728164104e-05,
+      "loss": 0.6146,
+      "step": 6410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7931658466061788e-05,
+      "loss": 0.6411,
+      "step": 6411
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7930944006566584e-05,
+      "loss": 0.6168,
+      "step": 6412
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7930229437934624e-05,
+      "loss": 0.7031,
+      "step": 6413
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7929514760175736e-05,
+      "loss": 0.6654,
+      "step": 6414
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7928799973299756e-05,
+      "loss": 0.6534,
+      "step": 6415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7928085077316528e-05,
+      "loss": 0.6371,
+      "step": 6416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792737007223588e-05,
+      "loss": 0.7499,
+      "step": 6417
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792665495806765e-05,
+      "loss": 0.7261,
+      "step": 6418
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792593973482169e-05,
+      "loss": 0.6143,
+      "step": 6419
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7925224402507832e-05,
+      "loss": 0.6922,
+      "step": 6420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792450896113592e-05,
+      "loss": 0.6752,
+      "step": 6421
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7923793410715804e-05,
+      "loss": 0.6874,
+      "step": 6422
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7923077751257328e-05,
+      "loss": 0.6158,
+      "step": 6423
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792236198277034e-05,
+      "loss": 0.7201,
+      "step": 6424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7921646105264688e-05,
+      "loss": 0.7271,
+      "step": 6425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792093011875023e-05,
+      "loss": 0.7268,
+      "step": 6426
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7920214023236807e-05,
+      "loss": 0.6677,
+      "step": 6427
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7919497818734287e-05,
+      "loss": 0.6427,
+      "step": 6428
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791878150525251e-05,
+      "loss": 0.7152,
+      "step": 6429
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7918065082801348e-05,
+      "loss": 0.6635,
+      "step": 6430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7917348551390652e-05,
+      "loss": 0.6944,
+      "step": 6431
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7916631911030283e-05,
+      "loss": 0.6788,
+      "step": 6432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79159151617301e-05,
+      "loss": 0.6177,
+      "step": 6433
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791519830349997e-05,
+      "loss": 0.598,
+      "step": 6434
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7914481336349757e-05,
+      "loss": 0.6984,
+      "step": 6435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7913764260289328e-05,
+      "loss": 0.6429,
+      "step": 6436
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791304707532855e-05,
+      "loss": 0.7479,
+      "step": 6437
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7912329781477287e-05,
+      "loss": 0.7129,
+      "step": 6438
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7911612378745418e-05,
+      "loss": 0.7096,
+      "step": 6439
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791089486714281e-05,
+      "loss": 0.6766,
+      "step": 6440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7910177246679336e-05,
+      "loss": 0.6869,
+      "step": 6441
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790945951736487e-05,
+      "loss": 0.6902,
+      "step": 6442
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7908741679209295e-05,
+      "loss": 0.6221,
+      "step": 6443
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7908023732222485e-05,
+      "loss": 0.7379,
+      "step": 6444
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790730567641432e-05,
+      "loss": 0.7299,
+      "step": 6445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790658751179468e-05,
+      "loss": 0.714,
+      "step": 6446
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7905869238373448e-05,
+      "loss": 0.6699,
+      "step": 6447
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790515085616051e-05,
+      "loss": 0.6526,
+      "step": 6448
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7904432365165748e-05,
+      "loss": 0.6222,
+      "step": 6449
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7903713765399054e-05,
+      "loss": 0.642,
+      "step": 6450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790299505687031e-05,
+      "loss": 0.6151,
+      "step": 6451
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7902276239589413e-05,
+      "loss": 0.6962,
+      "step": 6452
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790155731356625e-05,
+      "loss": 0.737,
+      "step": 6453
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790083827881072e-05,
+      "loss": 0.6359,
+      "step": 6454
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7900119135332707e-05,
+      "loss": 0.7071,
+      "step": 6455
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7899399883142115e-05,
+      "loss": 0.6303,
+      "step": 6456
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789868052224884e-05,
+      "loss": 0.6665,
+      "step": 6457
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789796105266278e-05,
+      "loss": 0.6278,
+      "step": 6458
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7897241474393837e-05,
+      "loss": 0.7346,
+      "step": 6459
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7896521787451912e-05,
+      "loss": 0.6789,
+      "step": 6460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789580199184691e-05,
+      "loss": 0.5763,
+      "step": 6461
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789508208758873e-05,
+      "loss": 0.6721,
+      "step": 6462
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7894362074687288e-05,
+      "loss": 0.71,
+      "step": 6463
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7893641953152484e-05,
+      "loss": 0.7062,
+      "step": 6464
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7892921722994235e-05,
+      "loss": 0.6828,
+      "step": 6465
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7892201384222443e-05,
+      "loss": 0.7117,
+      "step": 6466
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7891480936847027e-05,
+      "loss": 0.6802,
+      "step": 6467
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.78907603808779e-05,
+      "loss": 0.6674,
+      "step": 6468
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789003971632498e-05,
+      "loss": 0.6732,
+      "step": 6469
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7889318943198174e-05,
+      "loss": 0.5937,
+      "step": 6470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788859806150741e-05,
+      "loss": 0.6553,
+      "step": 6471
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7887877071262605e-05,
+      "loss": 0.5597,
+      "step": 6472
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7887155972473682e-05,
+      "loss": 0.6574,
+      "step": 6473
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7886434765150562e-05,
+      "loss": 0.6034,
+      "step": 6474
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788571344930317e-05,
+      "loss": 0.6709,
+      "step": 6475
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788499202494143e-05,
+      "loss": 0.6424,
+      "step": 6476
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7884270492075275e-05,
+      "loss": 0.7282,
+      "step": 6477
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788354885071463e-05,
+      "loss": 0.7052,
+      "step": 6478
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788282710086942e-05,
+      "loss": 0.761,
+      "step": 6479
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788210524254959e-05,
+      "loss": 0.6643,
+      "step": 6480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7881383275765062e-05,
+      "loss": 0.7404,
+      "step": 6481
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7880661200525777e-05,
+      "loss": 0.6736,
+      "step": 6482
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7879939016841666e-05,
+      "loss": 0.7086,
+      "step": 6483
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787921672472267e-05,
+      "loss": 0.6095,
+      "step": 6484
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7878494324178732e-05,
+      "loss": 0.7266,
+      "step": 6485
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787777181521979e-05,
+      "loss": 0.7098,
+      "step": 6486
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7877049197855783e-05,
+      "loss": 0.7199,
+      "step": 6487
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787632647209666e-05,
+      "loss": 0.6962,
+      "step": 6488
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787560363795236e-05,
+      "loss": 0.7041,
+      "step": 6489
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7874880695432836e-05,
+      "loss": 0.6531,
+      "step": 6490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7874157644548035e-05,
+      "loss": 0.6158,
+      "step": 6491
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7873434485307907e-05,
+      "loss": 0.7367,
+      "step": 6492
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7872711217722402e-05,
+      "loss": 0.6772,
+      "step": 6493
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7871987841801467e-05,
+      "loss": 0.7673,
+      "step": 6494
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787126435755507e-05,
+      "loss": 0.612,
+      "step": 6495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7870540764993155e-05,
+      "loss": 0.7333,
+      "step": 6496
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7869817064125686e-05,
+      "loss": 0.6424,
+      "step": 6497
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7869093254962618e-05,
+      "loss": 0.6423,
+      "step": 6498
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7868369337513913e-05,
+      "loss": 0.7582,
+      "step": 6499
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7867645311789533e-05,
+      "loss": 0.7089,
+      "step": 6500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7866921177799437e-05,
+      "loss": 0.6779,
+      "step": 6501
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7866196935553593e-05,
+      "loss": 0.5909,
+      "step": 6502
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786547258506197e-05,
+      "loss": 0.6681,
+      "step": 6503
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7864748126334534e-05,
+      "loss": 0.7348,
+      "step": 6504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7864023559381253e-05,
+      "loss": 0.6972,
+      "step": 6505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7863298884212093e-05,
+      "loss": 0.7149,
+      "step": 6506
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786257410083704e-05,
+      "loss": 0.611,
+      "step": 6507
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7861849209266052e-05,
+      "loss": 0.6711,
+      "step": 6508
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786112420950911e-05,
+      "loss": 0.601,
+      "step": 6509
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7860399101576194e-05,
+      "loss": 0.6602,
+      "step": 6510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7859673885477276e-05,
+      "loss": 0.6367,
+      "step": 6511
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7858948561222344e-05,
+      "loss": 0.689,
+      "step": 6512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7858223128821376e-05,
+      "loss": 0.6529,
+      "step": 6513
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785749758828435e-05,
+      "loss": 0.6063,
+      "step": 6514
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785677193962125e-05,
+      "loss": 0.6894,
+      "step": 6515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785604618284207e-05,
+      "loss": 0.7294,
+      "step": 6516
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7855320317956785e-05,
+      "loss": 0.7144,
+      "step": 6517
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7854594344975398e-05,
+      "loss": 0.655,
+      "step": 6518
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7853868263907885e-05,
+      "loss": 0.7076,
+      "step": 6519
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7853142074764246e-05,
+      "loss": 0.6625,
+      "step": 6520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7852415777554473e-05,
+      "loss": 0.7418,
+      "step": 6521
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7851689372288555e-05,
+      "loss": 0.686,
+      "step": 6522
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7850962858976496e-05,
+      "loss": 0.6419,
+      "step": 6523
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785023623762829e-05,
+      "loss": 0.6365,
+      "step": 6524
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784950950825393e-05,
+      "loss": 0.6643,
+      "step": 6525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7848782670863428e-05,
+      "loss": 0.6071,
+      "step": 6526
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784805572546678e-05,
+      "loss": 0.7082,
+      "step": 6527
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7847328672073983e-05,
+      "loss": 0.7242,
+      "step": 6528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7846601510695055e-05,
+      "loss": 0.6565,
+      "step": 6529
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7845874241339992e-05,
+      "loss": 0.6235,
+      "step": 6530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7845146864018806e-05,
+      "loss": 0.6837,
+      "step": 6531
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7844419378741506e-05,
+      "loss": 0.6325,
+      "step": 6532
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7843691785518103e-05,
+      "loss": 0.6485,
+      "step": 6533
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784296408435861e-05,
+      "loss": 0.6169,
+      "step": 6534
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7842236275273038e-05,
+      "loss": 0.6723,
+      "step": 6535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7841508358271403e-05,
+      "loss": 0.7139,
+      "step": 6536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7840780333363724e-05,
+      "loss": 0.7288,
+      "step": 6537
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784005220056002e-05,
+      "loss": 0.6823,
+      "step": 6538
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7839323959870306e-05,
+      "loss": 0.6856,
+      "step": 6539
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783859561130461e-05,
+      "loss": 0.7455,
+      "step": 6540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7837867154872948e-05,
+      "loss": 0.6553,
+      "step": 6541
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7837138590585347e-05,
+      "loss": 0.6176,
+      "step": 6542
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783640991845183e-05,
+      "loss": 0.6318,
+      "step": 6543
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7835681138482434e-05,
+      "loss": 0.6775,
+      "step": 6544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7834952250687173e-05,
+      "loss": 0.7628,
+      "step": 6545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7834223255076087e-05,
+      "loss": 0.5273,
+      "step": 6546
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7833494151659208e-05,
+      "loss": 0.6578,
+      "step": 6547
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7832764940446564e-05,
+      "loss": 0.6472,
+      "step": 6548
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7832035621448192e-05,
+      "loss": 0.6143,
+      "step": 6549
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783130619467413e-05,
+      "loss": 0.6899,
+      "step": 6550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7830576660134412e-05,
+      "loss": 0.6073,
+      "step": 6551
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7829847017839078e-05,
+      "loss": 0.6995,
+      "step": 6552
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7829117267798172e-05,
+      "loss": 0.6937,
+      "step": 6553
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782838741002173e-05,
+      "loss": 0.6462,
+      "step": 6554
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78276574445198e-05,
+      "loss": 0.5817,
+      "step": 6555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7826927371302426e-05,
+      "loss": 0.6569,
+      "step": 6556
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782619719037965e-05,
+      "loss": 0.7208,
+      "step": 6557
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782546690176153e-05,
+      "loss": 0.6557,
+      "step": 6558
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7824736505458106e-05,
+      "loss": 0.6359,
+      "step": 6559
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782400600147943e-05,
+      "loss": 0.707,
+      "step": 6560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782327538983556e-05,
+      "loss": 0.5483,
+      "step": 6561
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7822544670536544e-05,
+      "loss": 0.6595,
+      "step": 6562
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7821813843592447e-05,
+      "loss": 0.6034,
+      "step": 6563
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782108290901331e-05,
+      "loss": 0.6405,
+      "step": 6564
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7820351866809204e-05,
+      "loss": 0.7104,
+      "step": 6565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7819620716990184e-05,
+      "loss": 0.6858,
+      "step": 6566
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7818889459566314e-05,
+      "loss": 0.7485,
+      "step": 6567
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781815809454765e-05,
+      "loss": 0.6337,
+      "step": 6568
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7817426621944266e-05,
+      "loss": 0.6125,
+      "step": 6569
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781669504176622e-05,
+      "loss": 0.6456,
+      "step": 6570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7815963354023586e-05,
+      "loss": 0.6736,
+      "step": 6571
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7815231558726422e-05,
+      "loss": 0.664,
+      "step": 6572
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781449965588481e-05,
+      "loss": 0.7314,
+      "step": 6573
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7813767645508812e-05,
+      "loss": 0.6784,
+      "step": 6574
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7813035527608508e-05,
+      "loss": 0.6723,
+      "step": 6575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7812303302193966e-05,
+      "loss": 0.6709,
+      "step": 6576
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7811570969275274e-05,
+      "loss": 0.6457,
+      "step": 6577
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7810838528862496e-05,
+      "loss": 0.698,
+      "step": 6578
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7810105980965717e-05,
+      "loss": 0.6417,
+      "step": 6579
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780937332559502e-05,
+      "loss": 0.7635,
+      "step": 6580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780864056276048e-05,
+      "loss": 0.5918,
+      "step": 6581
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7807907692472188e-05,
+      "loss": 0.6933,
+      "step": 6582
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7807174714740225e-05,
+      "loss": 0.6646,
+      "step": 6583
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7806441629574675e-05,
+      "loss": 0.6926,
+      "step": 6584
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7805708436985632e-05,
+      "loss": 0.7018,
+      "step": 6585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780497513698318e-05,
+      "loss": 0.7118,
+      "step": 6586
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7804241729577413e-05,
+      "loss": 0.6941,
+      "step": 6587
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7803508214778423e-05,
+      "loss": 0.572,
+      "step": 6588
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7802774592596302e-05,
+      "loss": 0.7365,
+      "step": 6589
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7802040863041148e-05,
+      "loss": 0.6757,
+      "step": 6590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7801307026123056e-05,
+      "loss": 0.6741,
+      "step": 6591
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7800573081852124e-05,
+      "loss": 0.7062,
+      "step": 6592
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779983903023845e-05,
+      "loss": 0.6423,
+      "step": 6593
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779910487129214e-05,
+      "loss": 0.5726,
+      "step": 6594
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7798370605023295e-05,
+      "loss": 0.7248,
+      "step": 6595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7797636231442018e-05,
+      "loss": 0.6827,
+      "step": 6596
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779690175055841e-05,
+      "loss": 0.6178,
+      "step": 6597
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779616716238259e-05,
+      "loss": 0.6956,
+      "step": 6598
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7795432466924656e-05,
+      "loss": 0.5833,
+      "step": 6599
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7794697664194722e-05,
+      "loss": 0.7075,
+      "step": 6600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77939627542029e-05,
+      "loss": 0.6175,
+      "step": 6601
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77932277369593e-05,
+      "loss": 0.6301,
+      "step": 6602
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7792492612474038e-05,
+      "loss": 0.6799,
+      "step": 6603
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7791757380757234e-05,
+      "loss": 0.646,
+      "step": 6604
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7791022041819e-05,
+      "loss": 0.6515,
+      "step": 6605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7790286595669455e-05,
+      "loss": 0.6877,
+      "step": 6606
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7789551042318724e-05,
+      "loss": 0.6893,
+      "step": 6607
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778881538177693e-05,
+      "loss": 0.6738,
+      "step": 6608
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7788079614054186e-05,
+      "loss": 0.6791,
+      "step": 6609
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7787343739160624e-05,
+      "loss": 0.659,
+      "step": 6610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7786607757106372e-05,
+      "loss": 0.6184,
+      "step": 6611
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7785871667901555e-05,
+      "loss": 0.6895,
+      "step": 6612
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7785135471556302e-05,
+      "loss": 0.6185,
+      "step": 6613
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778439916808074e-05,
+      "loss": 0.7021,
+      "step": 6614
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778366275748501e-05,
+      "loss": 0.6126,
+      "step": 6615
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7782926239779242e-05,
+      "loss": 0.6936,
+      "step": 6616
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778218961497357e-05,
+      "loss": 0.7068,
+      "step": 6617
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778145288307813e-05,
+      "loss": 0.5218,
+      "step": 6618
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7780716044103062e-05,
+      "loss": 0.7005,
+      "step": 6619
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7779979098058506e-05,
+      "loss": 0.7238,
+      "step": 6620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7779242044954596e-05,
+      "loss": 0.7105,
+      "step": 6621
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7778504884801483e-05,
+      "loss": 0.6427,
+      "step": 6622
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7777767617609307e-05,
+      "loss": 0.6955,
+      "step": 6623
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7777030243388217e-05,
+      "loss": 0.7245,
+      "step": 6624
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7776292762148356e-05,
+      "loss": 0.5864,
+      "step": 6625
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7775555173899876e-05,
+      "loss": 0.6156,
+      "step": 6626
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777481747865292e-05,
+      "loss": 0.7289,
+      "step": 6627
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7774079676417644e-05,
+      "loss": 0.7041,
+      "step": 6628
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7773341767204205e-05,
+      "loss": 0.7418,
+      "step": 6629
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777260375102275e-05,
+      "loss": 0.6762,
+      "step": 6630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777186562788344e-05,
+      "loss": 0.6986,
+      "step": 6631
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7771127397796427e-05,
+      "loss": 0.6793,
+      "step": 6632
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7770389060771875e-05,
+      "loss": 0.6974,
+      "step": 6633
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776965061681994e-05,
+      "loss": 0.6786,
+      "step": 6634
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7768912065950786e-05,
+      "loss": 0.6881,
+      "step": 6635
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776817340817458e-05,
+      "loss": 0.6804,
+      "step": 6636
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7767434643501476e-05,
+      "loss": 0.5931,
+      "step": 6637
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776669577194165e-05,
+      "loss": 0.6921,
+      "step": 6638
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7765956793505265e-05,
+      "loss": 0.6618,
+      "step": 6639
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776521770820249e-05,
+      "loss": 0.6975,
+      "step": 6640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7764478516043495e-05,
+      "loss": 0.7009,
+      "step": 6641
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7763739217038455e-05,
+      "loss": 0.6263,
+      "step": 6642
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7762999811197538e-05,
+      "loss": 0.5649,
+      "step": 6643
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776226029853093e-05,
+      "loss": 0.6616,
+      "step": 6644
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7761520679048794e-05,
+      "loss": 0.73,
+      "step": 6645
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7760780952761314e-05,
+      "loss": 0.6291,
+      "step": 6646
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776004111967867e-05,
+      "loss": 0.6604,
+      "step": 6647
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775930117981104e-05,
+      "loss": 0.7024,
+      "step": 6648
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775856113316861e-05,
+      "loss": 0.6452,
+      "step": 6649
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7757820979761563e-05,
+      "loss": 0.696,
+      "step": 6650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7757080719600078e-05,
+      "loss": 0.5834,
+      "step": 6651
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775634035269435e-05,
+      "loss": 0.745,
+      "step": 6652
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775559987905456e-05,
+      "loss": 0.7129,
+      "step": 6653
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7754859298690905e-05,
+      "loss": 0.7173,
+      "step": 6654
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7754118611613573e-05,
+      "loss": 0.72,
+      "step": 6655
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7753377817832752e-05,
+      "loss": 0.7,
+      "step": 6656
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7752636917358642e-05,
+      "loss": 0.7047,
+      "step": 6657
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7751895910201434e-05,
+      "loss": 0.7043,
+      "step": 6658
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7751154796371325e-05,
+      "loss": 0.6622,
+      "step": 6659
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7750413575878518e-05,
+      "loss": 0.5712,
+      "step": 6660
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7749672248733208e-05,
+      "loss": 0.6391,
+      "step": 6661
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7748930814945602e-05,
+      "loss": 0.6789,
+      "step": 6662
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7748189274525896e-05,
+      "loss": 0.6159,
+      "step": 6663
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77474476274843e-05,
+      "loss": 0.6118,
+      "step": 6664
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7746705873831014e-05,
+      "loss": 0.6686,
+      "step": 6665
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7745964013576248e-05,
+      "loss": 0.6516,
+      "step": 6666
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7745222046730216e-05,
+      "loss": 0.6631,
+      "step": 6667
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774447997330312e-05,
+      "loss": 0.6533,
+      "step": 6668
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7743737793305175e-05,
+      "loss": 0.6921,
+      "step": 6669
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774299550674659e-05,
+      "loss": 0.6204,
+      "step": 6670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7742253113637587e-05,
+      "loss": 0.7001,
+      "step": 6671
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774151061398838e-05,
+      "loss": 0.7419,
+      "step": 6672
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774076800780918e-05,
+      "loss": 0.6896,
+      "step": 6673
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7740025295110214e-05,
+      "loss": 0.6972,
+      "step": 6674
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77392824759017e-05,
+      "loss": 0.7226,
+      "step": 6675
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7738539550193856e-05,
+      "loss": 0.6064,
+      "step": 6676
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773779651799691e-05,
+      "loss": 0.6449,
+      "step": 6677
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7737053379321084e-05,
+      "loss": 0.6444,
+      "step": 6678
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7736310134176604e-05,
+      "loss": 0.6131,
+      "step": 6679
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77355667825737e-05,
+      "loss": 0.6202,
+      "step": 6680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77348233245226e-05,
+      "loss": 0.6736,
+      "step": 6681
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7734079760033532e-05,
+      "loss": 0.5801,
+      "step": 6682
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7733336089116736e-05,
+      "loss": 0.693,
+      "step": 6683
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7732592311782438e-05,
+      "loss": 0.5848,
+      "step": 6684
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7731848428040873e-05,
+      "loss": 0.6263,
+      "step": 6685
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7731104437902283e-05,
+      "loss": 0.643,
+      "step": 6686
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77303603413769e-05,
+      "loss": 0.5669,
+      "step": 6687
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772961613847497e-05,
+      "loss": 0.7396,
+      "step": 6688
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7728871829206725e-05,
+      "loss": 0.7569,
+      "step": 6689
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7728127413582418e-05,
+      "loss": 0.6986,
+      "step": 6690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7727382891612284e-05,
+      "loss": 0.6328,
+      "step": 6691
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772663826330657e-05,
+      "loss": 0.698,
+      "step": 6692
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7725893528675526e-05,
+      "loss": 0.7468,
+      "step": 6693
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77251486877294e-05,
+      "loss": 0.646,
+      "step": 6694
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772440374047844e-05,
+      "loss": 0.6623,
+      "step": 6695
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7723658686932898e-05,
+      "loss": 0.6396,
+      "step": 6696
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772291352710302e-05,
+      "loss": 0.7434,
+      "step": 6697
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772216826099907e-05,
+      "loss": 0.6558,
+      "step": 6698
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77214228886313e-05,
+      "loss": 0.6535,
+      "step": 6699
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772067741000997e-05,
+      "loss": 0.6293,
+      "step": 6700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7719931825145326e-05,
+      "loss": 0.6712,
+      "step": 6701
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7719186134047643e-05,
+      "loss": 0.686,
+      "step": 6702
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7718440336727173e-05,
+      "loss": 0.6836,
+      "step": 6703
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7717694433194186e-05,
+      "loss": 0.6259,
+      "step": 6704
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771694842345894e-05,
+      "loss": 0.6794,
+      "step": 6705
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7716202307531704e-05,
+      "loss": 0.7128,
+      "step": 6706
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771545608542274e-05,
+      "loss": 0.6636,
+      "step": 6707
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7714709757142324e-05,
+      "loss": 0.6645,
+      "step": 6708
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7713963322700724e-05,
+      "loss": 0.6849,
+      "step": 6709
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771321678210821e-05,
+      "loss": 0.6416,
+      "step": 6710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7712470135375053e-05,
+      "loss": 0.7242,
+      "step": 6711
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7711723382511536e-05,
+      "loss": 0.7054,
+      "step": 6712
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7710976523527925e-05,
+      "loss": 0.7647,
+      "step": 6713
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7710229558434502e-05,
+      "loss": 0.6435,
+      "step": 6714
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7709482487241545e-05,
+      "loss": 0.6516,
+      "step": 6715
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7708735309959335e-05,
+      "loss": 0.6719,
+      "step": 6716
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7707988026598154e-05,
+      "loss": 0.6551,
+      "step": 6717
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7707240637168285e-05,
+      "loss": 0.6626,
+      "step": 6718
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770649314168001e-05,
+      "loss": 0.6731,
+      "step": 6719
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7705745540143622e-05,
+      "loss": 0.5865,
+      "step": 6720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7704997832569405e-05,
+      "loss": 0.6963,
+      "step": 6721
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7704250018967645e-05,
+      "loss": 0.6951,
+      "step": 6722
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7703502099348636e-05,
+      "loss": 0.6971,
+      "step": 6723
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7702754073722668e-05,
+      "loss": 0.7128,
+      "step": 6724
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7702005942100037e-05,
+      "loss": 0.6999,
+      "step": 6725
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7701257704491036e-05,
+      "loss": 0.6613,
+      "step": 6726
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7700509360905962e-05,
+      "loss": 0.6022,
+      "step": 6727
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7699760911355115e-05,
+      "loss": 0.6548,
+      "step": 6728
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7699012355848787e-05,
+      "loss": 0.6793,
+      "step": 6729
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769826369439729e-05,
+      "loss": 0.6872,
+      "step": 6730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7697514927010916e-05,
+      "loss": 0.7051,
+      "step": 6731
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7696766053699978e-05,
+      "loss": 0.6138,
+      "step": 6732
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7696017074474773e-05,
+      "loss": 0.6171,
+      "step": 6733
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769526798934561e-05,
+      "loss": 0.7131,
+      "step": 6734
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7694518798322797e-05,
+      "loss": 0.6378,
+      "step": 6735
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7693769501416645e-05,
+      "loss": 0.666,
+      "step": 6736
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7693020098637464e-05,
+      "loss": 0.723,
+      "step": 6737
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7692270589995573e-05,
+      "loss": 0.6412,
+      "step": 6738
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7691520975501272e-05,
+      "loss": 0.7011,
+      "step": 6739
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7690771255164888e-05,
+      "loss": 0.717,
+      "step": 6740
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7690021428996732e-05,
+      "loss": 0.679,
+      "step": 6741
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7689271497007127e-05,
+      "loss": 0.6662,
+      "step": 6742
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7688521459206386e-05,
+      "loss": 0.6737,
+      "step": 6743
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7687771315604833e-05,
+      "loss": 0.6209,
+      "step": 6744
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7687021066212795e-05,
+      "loss": 0.6134,
+      "step": 6745
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7686270711040593e-05,
+      "loss": 0.6349,
+      "step": 6746
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768552025009855e-05,
+      "loss": 0.6683,
+      "step": 6747
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7684769683396996e-05,
+      "loss": 0.6276,
+      "step": 6748
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7684019010946262e-05,
+      "loss": 0.6885,
+      "step": 6749
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768326823275667e-05,
+      "loss": 0.672,
+      "step": 6750
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7682517348838555e-05,
+      "loss": 0.6493,
+      "step": 6751
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7681766359202256e-05,
+      "loss": 0.5874,
+      "step": 6752
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.76810152638581e-05,
+      "loss": 0.6321,
+      "step": 6753
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768026406281642e-05,
+      "loss": 0.6469,
+      "step": 6754
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7679512756087564e-05,
+      "loss": 0.5892,
+      "step": 6755
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7678761343681864e-05,
+      "loss": 0.701,
+      "step": 6756
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7678009825609658e-05,
+      "loss": 0.6918,
+      "step": 6757
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7677258201881297e-05,
+      "loss": 0.5128,
+      "step": 6758
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767650647250711e-05,
+      "loss": 0.7337,
+      "step": 6759
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7675754637497452e-05,
+      "loss": 0.5868,
+      "step": 6760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7675002696862664e-05,
+      "loss": 0.7345,
+      "step": 6761
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7674250650613096e-05,
+      "loss": 0.7109,
+      "step": 6762
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7673498498759095e-05,
+      "loss": 0.6947,
+      "step": 6763
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7672746241311013e-05,
+      "loss": 0.6109,
+      "step": 6764
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7671993878279204e-05,
+      "loss": 0.6114,
+      "step": 6765
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7671241409674014e-05,
+      "loss": 0.5828,
+      "step": 6766
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7670488835505805e-05,
+      "loss": 0.7526,
+      "step": 6767
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7669736155784924e-05,
+      "loss": 0.7037,
+      "step": 6768
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766898337052174e-05,
+      "loss": 0.5884,
+      "step": 6769
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7668230479726603e-05,
+      "loss": 0.6235,
+      "step": 6770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7667477483409875e-05,
+      "loss": 0.6418,
+      "step": 6771
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7666724381581923e-05,
+      "loss": 0.6603,
+      "step": 6772
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7665971174253106e-05,
+      "loss": 0.6941,
+      "step": 6773
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766521786143379e-05,
+      "loss": 0.6853,
+      "step": 6774
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7664464443134338e-05,
+      "loss": 0.6442,
+      "step": 6775
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7663710919365128e-05,
+      "loss": 0.6664,
+      "step": 6776
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7662957290136515e-05,
+      "loss": 0.6768,
+      "step": 6777
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766220355545888e-05,
+      "loss": 0.6652,
+      "step": 6778
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7661449715342588e-05,
+      "loss": 0.6284,
+      "step": 6779
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7660695769798018e-05,
+      "loss": 0.6778,
+      "step": 6780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7659941718835543e-05,
+      "loss": 0.5894,
+      "step": 6781
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765918756246554e-05,
+      "loss": 0.6137,
+      "step": 6782
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7658433300698384e-05,
+      "loss": 0.6774,
+      "step": 6783
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765767893354446e-05,
+      "loss": 0.7305,
+      "step": 6784
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7656924461014144e-05,
+      "loss": 0.6228,
+      "step": 6785
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765616988311782e-05,
+      "loss": 0.7238,
+      "step": 6786
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765541519986587e-05,
+      "loss": 0.5922,
+      "step": 6787
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7654660411268678e-05,
+      "loss": 0.6624,
+      "step": 6788
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7653905517336632e-05,
+      "loss": 0.6168,
+      "step": 6789
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7653150518080125e-05,
+      "loss": 0.6168,
+      "step": 6790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7652395413509542e-05,
+      "loss": 0.7082,
+      "step": 6791
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7651640203635274e-05,
+      "loss": 0.7447,
+      "step": 6792
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7650884888467712e-05,
+      "loss": 0.7517,
+      "step": 6793
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7650129468017252e-05,
+      "loss": 0.6627,
+      "step": 6794
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7649373942294288e-05,
+      "loss": 0.6901,
+      "step": 6795
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7648618311309218e-05,
+      "loss": 0.731,
+      "step": 6796
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764786257507244e-05,
+      "loss": 0.6748,
+      "step": 6797
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764710673359435e-05,
+      "loss": 0.634,
+      "step": 6798
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7646350786885352e-05,
+      "loss": 0.7072,
+      "step": 6799
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7645594734955854e-05,
+      "loss": 0.7217,
+      "step": 6800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764483857781625e-05,
+      "loss": 0.7174,
+      "step": 6801
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764408231547695e-05,
+      "loss": 0.6673,
+      "step": 6802
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764332594794836e-05,
+      "loss": 0.6512,
+      "step": 6803
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764256947524089e-05,
+      "loss": 0.717,
+      "step": 6804
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764181289736495e-05,
+      "loss": 0.6367,
+      "step": 6805
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7641056214330946e-05,
+      "loss": 0.698,
+      "step": 6806
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76402994261493e-05,
+      "loss": 0.6033,
+      "step": 6807
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7639542532830417e-05,
+      "loss": 0.6622,
+      "step": 6808
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7638785534384718e-05,
+      "loss": 0.5937,
+      "step": 6809
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7638028430822617e-05,
+      "loss": 0.7076,
+      "step": 6810
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763727122215453e-05,
+      "loss": 0.6593,
+      "step": 6811
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7636513908390888e-05,
+      "loss": 0.7228,
+      "step": 6812
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76357564895421e-05,
+      "loss": 0.6695,
+      "step": 6813
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763499896561859e-05,
+      "loss": 0.695,
+      "step": 6814
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7634241336630792e-05,
+      "loss": 0.6327,
+      "step": 6815
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7633483602589122e-05,
+      "loss": 0.639,
+      "step": 6816
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763272576350401e-05,
+      "loss": 0.6849,
+      "step": 6817
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7631967819385883e-05,
+      "loss": 0.6717,
+      "step": 6818
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7631209770245178e-05,
+      "loss": 0.7497,
+      "step": 6819
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7630451616092317e-05,
+      "loss": 0.569,
+      "step": 6820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762969335693774e-05,
+      "loss": 0.6987,
+      "step": 6821
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7628934992791874e-05,
+      "loss": 0.5839,
+      "step": 6822
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7628176523665163e-05,
+      "loss": 0.6153,
+      "step": 6823
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7627417949568037e-05,
+      "loss": 0.6495,
+      "step": 6824
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7626659270510942e-05,
+      "loss": 0.6582,
+      "step": 6825
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7625900486504314e-05,
+      "loss": 0.6764,
+      "step": 6826
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762514159755859e-05,
+      "loss": 0.6904,
+      "step": 6827
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762438260368422e-05,
+      "loss": 0.6263,
+      "step": 6828
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762362350489165e-05,
+      "loss": 0.5695,
+      "step": 6829
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7622864301191315e-05,
+      "loss": 0.6127,
+      "step": 6830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7622104992593675e-05,
+      "loss": 0.6306,
+      "step": 6831
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762134557910917e-05,
+      "loss": 0.7527,
+      "step": 6832
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762058606074825e-05,
+      "loss": 0.5609,
+      "step": 6833
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7619826437521375e-05,
+      "loss": 0.6999,
+      "step": 6834
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761906670943899e-05,
+      "loss": 0.67,
+      "step": 6835
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7618306876511554e-05,
+      "loss": 0.6467,
+      "step": 6836
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761754693874952e-05,
+      "loss": 0.633,
+      "step": 6837
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7616786896163347e-05,
+      "loss": 0.6511,
+      "step": 6838
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7616026748763495e-05,
+      "loss": 0.5927,
+      "step": 6839
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7615266496560425e-05,
+      "loss": 0.6528,
+      "step": 6840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761450613956459e-05,
+      "loss": 0.6535,
+      "step": 6841
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7613745677786464e-05,
+      "loss": 0.6691,
+      "step": 6842
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7612985111236508e-05,
+      "loss": 0.6749,
+      "step": 6843
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7612224439925185e-05,
+      "loss": 0.6923,
+      "step": 6844
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7611463663862968e-05,
+      "loss": 0.5923,
+      "step": 6845
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7610702783060323e-05,
+      "loss": 0.6209,
+      "step": 6846
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7609941797527718e-05,
+      "loss": 0.6103,
+      "step": 6847
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7609180707275627e-05,
+      "loss": 0.7007,
+      "step": 6848
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7608419512314527e-05,
+      "loss": 0.7597,
+      "step": 6849
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7607658212654885e-05,
+      "loss": 0.5763,
+      "step": 6850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7606896808307182e-05,
+      "loss": 0.6599,
+      "step": 6851
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7606135299281897e-05,
+      "loss": 0.7331,
+      "step": 6852
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7605373685589503e-05,
+      "loss": 0.5875,
+      "step": 6853
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760461196724049e-05,
+      "loss": 0.6795,
+      "step": 6854
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7603850144245333e-05,
+      "loss": 0.6386,
+      "step": 6855
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7603088216614512e-05,
+      "loss": 0.6844,
+      "step": 6856
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7602326184358524e-05,
+      "loss": 0.6968,
+      "step": 6857
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760156404748784e-05,
+      "loss": 0.6886,
+      "step": 6858
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760080180601296e-05,
+      "loss": 0.7233,
+      "step": 6859
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7600039459944363e-05,
+      "loss": 0.6812,
+      "step": 6860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759927700929255e-05,
+      "loss": 0.6657,
+      "step": 6861
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7598514454068013e-05,
+      "loss": 0.7098,
+      "step": 6862
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759775179428123e-05,
+      "loss": 0.5672,
+      "step": 6863
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7596989029942714e-05,
+      "loss": 0.6263,
+      "step": 6864
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759622616106295e-05,
+      "loss": 0.737,
+      "step": 6865
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759546318765244e-05,
+      "loss": 0.6471,
+      "step": 6866
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7594700109721686e-05,
+      "loss": 0.5684,
+      "step": 6867
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759393692728118e-05,
+      "loss": 0.6962,
+      "step": 6868
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7593173640341434e-05,
+      "loss": 0.6548,
+      "step": 6869
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7592410248912943e-05,
+      "loss": 0.6903,
+      "step": 6870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7591646753006214e-05,
+      "loss": 0.6875,
+      "step": 6871
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759088315263176e-05,
+      "loss": 0.6376,
+      "step": 6872
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7590119447800083e-05,
+      "loss": 0.7138,
+      "step": 6873
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758935563852169e-05,
+      "loss": 0.5969,
+      "step": 6874
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7588591724807098e-05,
+      "loss": 0.6357,
+      "step": 6875
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7587827706666813e-05,
+      "loss": 0.6503,
+      "step": 6876
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7587063584111354e-05,
+      "loss": 0.5889,
+      "step": 6877
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7586299357151234e-05,
+      "loss": 0.6357,
+      "step": 6878
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7585535025796966e-05,
+      "loss": 0.7274,
+      "step": 6879
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758477059005907e-05,
+      "loss": 0.6865,
+      "step": 6880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7584006049948072e-05,
+      "loss": 0.6727,
+      "step": 6881
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7583241405474482e-05,
+      "loss": 0.6996,
+      "step": 6882
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758247665664883e-05,
+      "loss": 0.6439,
+      "step": 6883
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7581711803481635e-05,
+      "loss": 0.716,
+      "step": 6884
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7580946845983425e-05,
+      "loss": 0.6072,
+      "step": 6885
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758018178416472e-05,
+      "loss": 0.6993,
+      "step": 6886
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757941661803606e-05,
+      "loss": 0.6867,
+      "step": 6887
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7578651347607964e-05,
+      "loss": 0.6558,
+      "step": 6888
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7577885972890966e-05,
+      "loss": 0.6961,
+      "step": 6889
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7577120493895597e-05,
+      "loss": 0.742,
+      "step": 6890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7576354910632397e-05,
+      "loss": 0.6872,
+      "step": 6891
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757558922311189e-05,
+      "loss": 0.6239,
+      "step": 6892
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7574823431344626e-05,
+      "loss": 0.6506,
+      "step": 6893
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7574057535341134e-05,
+      "loss": 0.6019,
+      "step": 6894
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7573291535111955e-05,
+      "loss": 0.6901,
+      "step": 6895
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7572525430667627e-05,
+      "loss": 0.7029,
+      "step": 6896
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7571759222018697e-05,
+      "loss": 0.7132,
+      "step": 6897
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7570992909175706e-05,
+      "loss": 0.6856,
+      "step": 6898
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75702264921492e-05,
+      "loss": 0.6609,
+      "step": 6899
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7569459970949727e-05,
+      "loss": 0.7072,
+      "step": 6900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7568693345587834e-05,
+      "loss": 0.5749,
+      "step": 6901
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756792661607407e-05,
+      "loss": 0.5945,
+      "step": 6902
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7567159782418985e-05,
+      "loss": 0.6841,
+      "step": 6903
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7566392844633133e-05,
+      "loss": 0.6965,
+      "step": 6904
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7565625802727068e-05,
+      "loss": 0.7212,
+      "step": 6905
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7564858656711343e-05,
+      "loss": 0.6158,
+      "step": 6906
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756409140659651e-05,
+      "loss": 0.6386,
+      "step": 6907
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7563324052393142e-05,
+      "loss": 0.6535,
+      "step": 6908
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7562556594111786e-05,
+      "loss": 0.718,
+      "step": 6909
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756178903176301e-05,
+      "loss": 0.6064,
+      "step": 6910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7561021365357366e-05,
+      "loss": 0.6569,
+      "step": 6911
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7560253594905425e-05,
+      "loss": 0.7356,
+      "step": 6912
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7559485720417753e-05,
+      "loss": 0.6304,
+      "step": 6913
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755871774190492e-05,
+      "loss": 0.572,
+      "step": 6914
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755794965937748e-05,
+      "loss": 0.7308,
+      "step": 6915
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755718147284602e-05,
+      "loss": 0.6013,
+      "step": 6916
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75564131823211e-05,
+      "loss": 0.6163,
+      "step": 6917
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755564478781329e-05,
+      "loss": 0.6406,
+      "step": 6918
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7554876289333175e-05,
+      "loss": 0.7133,
+      "step": 6919
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7554107686891324e-05,
+      "loss": 0.6402,
+      "step": 6920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755333898049831e-05,
+      "loss": 0.6928,
+      "step": 6921
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7552570170164718e-05,
+      "loss": 0.5823,
+      "step": 6922
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755180125590112e-05,
+      "loss": 0.6173,
+      "step": 6923
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7551032237718105e-05,
+      "loss": 0.6991,
+      "step": 6924
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755026311562625e-05,
+      "loss": 0.7191,
+      "step": 6925
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754949388963614e-05,
+      "loss": 0.6984,
+      "step": 6926
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754872455975836e-05,
+      "loss": 0.6121,
+      "step": 6927
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75479551260035e-05,
+      "loss": 0.6903,
+      "step": 6928
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754718558838214e-05,
+      "loss": 0.7027,
+      "step": 6929
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754641594690488e-05,
+      "loss": 0.6555,
+      "step": 6930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7545646201582304e-05,
+      "loss": 0.6273,
+      "step": 6931
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7544876352425005e-05,
+      "loss": 0.6929,
+      "step": 6932
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754410639944358e-05,
+      "loss": 0.7198,
+      "step": 6933
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754333634264862e-05,
+      "loss": 0.6057,
+      "step": 6934
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7542566182050725e-05,
+      "loss": 0.716,
+      "step": 6935
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754179591766049e-05,
+      "loss": 0.6683,
+      "step": 6936
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7541025549488517e-05,
+      "loss": 0.67,
+      "step": 6937
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7540255077545408e-05,
+      "loss": 0.6154,
+      "step": 6938
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7539484501841763e-05,
+      "loss": 0.6483,
+      "step": 6939
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7538713822388186e-05,
+      "loss": 0.6513,
+      "step": 6940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7537943039195282e-05,
+      "loss": 0.5215,
+      "step": 6941
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753717215227366e-05,
+      "loss": 0.6735,
+      "step": 6942
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7536401161633927e-05,
+      "loss": 0.6787,
+      "step": 6943
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7535630067286692e-05,
+      "loss": 0.6625,
+      "step": 6944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7534858869242563e-05,
+      "loss": 0.6814,
+      "step": 6945
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7534087567512156e-05,
+      "loss": 0.7039,
+      "step": 6946
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7533316162106088e-05,
+      "loss": 0.753,
+      "step": 6947
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7532544653034964e-05,
+      "loss": 0.6788,
+      "step": 6948
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753177304030941e-05,
+      "loss": 0.7167,
+      "step": 6949
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753100132394004e-05,
+      "loss": 0.6462,
+      "step": 6950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7530229503937474e-05,
+      "loss": 0.7575,
+      "step": 6951
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7529457580312336e-05,
+      "loss": 0.6314,
+      "step": 6952
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7528685553075244e-05,
+      "loss": 0.6735,
+      "step": 6953
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7527913422236828e-05,
+      "loss": 0.6872,
+      "step": 6954
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7527141187807705e-05,
+      "loss": 0.6507,
+      "step": 6955
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7526368849798504e-05,
+      "loss": 0.71,
+      "step": 6956
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7525596408219858e-05,
+      "loss": 0.6579,
+      "step": 6957
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7524823863082387e-05,
+      "loss": 0.6239,
+      "step": 6958
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752405121439673e-05,
+      "loss": 0.6117,
+      "step": 6959
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7523278462173518e-05,
+      "loss": 0.7244,
+      "step": 6960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7522505606423387e-05,
+      "loss": 0.5873,
+      "step": 6961
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7521732647156963e-05,
+      "loss": 0.7156,
+      "step": 6962
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7520959584384895e-05,
+      "loss": 0.6677,
+      "step": 6963
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7520186418117808e-05,
+      "loss": 0.6161,
+      "step": 6964
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751941314836635e-05,
+      "loss": 0.6782,
+      "step": 6965
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7518639775141163e-05,
+      "loss": 0.7381,
+      "step": 6966
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7517866298452886e-05,
+      "loss": 0.7127,
+      "step": 6967
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751709271831216e-05,
+      "loss": 0.6012,
+      "step": 6968
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7516319034729634e-05,
+      "loss": 0.5881,
+      "step": 6969
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7515545247715956e-05,
+      "loss": 0.6646,
+      "step": 6970
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751477135728177e-05,
+      "loss": 0.6949,
+      "step": 6971
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7513997363437728e-05,
+      "loss": 0.6735,
+      "step": 6972
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7513223266194476e-05,
+      "loss": 0.5974,
+      "step": 6973
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7512449065562674e-05,
+      "loss": 0.6796,
+      "step": 6974
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7511674761552973e-05,
+      "loss": 0.6508,
+      "step": 6975
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7510900354176024e-05,
+      "loss": 0.6696,
+      "step": 6976
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7510125843442485e-05,
+      "loss": 0.7297,
+      "step": 6977
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750935122936302e-05,
+      "loss": 0.6245,
+      "step": 6978
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750857651194828e-05,
+      "loss": 0.6271,
+      "step": 6979
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7507801691208932e-05,
+      "loss": 0.562,
+      "step": 6980
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7507026767155634e-05,
+      "loss": 0.5654,
+      "step": 6981
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7506251739799052e-05,
+      "loss": 0.6458,
+      "step": 6982
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750547660914985e-05,
+      "loss": 0.7425,
+      "step": 6983
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7504701375218697e-05,
+      "loss": 0.6748,
+      "step": 6984
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750392603801626e-05,
+      "loss": 0.7068,
+      "step": 6985
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7503150597553205e-05,
+      "loss": 0.639,
+      "step": 6986
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7502375053840205e-05,
+      "loss": 0.674,
+      "step": 6987
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750159940688793e-05,
+      "loss": 0.6605,
+      "step": 6988
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7500823656707056e-05,
+      "loss": 0.6448,
+      "step": 6989
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750004780330826e-05,
+      "loss": 0.7429,
+      "step": 6990
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7499271846702216e-05,
+      "loss": 0.7079,
+      "step": 6991
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74984957868996e-05,
+      "loss": 0.6747,
+      "step": 6992
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7497719623911095e-05,
+      "loss": 0.6707,
+      "step": 6993
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749694335774738e-05,
+      "loss": 0.6652,
+      "step": 6994
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7496166988419135e-05,
+      "loss": 0.6288,
+      "step": 6995
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749539051593705e-05,
+      "loss": 0.6736,
+      "step": 6996
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74946139403118e-05,
+      "loss": 0.6035,
+      "step": 6997
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7493837261554077e-05,
+      "loss": 0.6719,
+      "step": 6998
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7493060479674573e-05,
+      "loss": 0.6235,
+      "step": 6999
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749228359468397e-05,
+      "loss": 0.6829,
+      "step": 7000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7491506606592962e-05,
+      "loss": 0.6387,
+      "step": 7001
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749072951541224e-05,
+      "loss": 0.581,
+      "step": 7002
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74899523211525e-05,
+      "loss": 0.6452,
+      "step": 7003
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7489175023824435e-05,
+      "loss": 0.627,
+      "step": 7004
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7488397623438738e-05,
+      "loss": 0.6878,
+      "step": 7005
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748762012000611e-05,
+      "loss": 0.6623,
+      "step": 7006
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7486842513537253e-05,
+      "loss": 0.6009,
+      "step": 7007
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748606480404286e-05,
+      "loss": 0.6186,
+      "step": 7008
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748528699153364e-05,
+      "loss": 0.7108,
+      "step": 7009
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74845090760203e-05,
+      "loss": 0.6248,
+      "step": 7010
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748373105751353e-05,
+      "loss": 0.7152,
+      "step": 7011
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7482952936024045e-05,
+      "loss": 0.6495,
+      "step": 7012
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748217471156256e-05,
+      "loss": 0.7574,
+      "step": 7013
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748139638413977e-05,
+      "loss": 0.6476,
+      "step": 7014
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7480617953766393e-05,
+      "loss": 0.7076,
+      "step": 7015
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747983942045314e-05,
+      "loss": 0.6355,
+      "step": 7016
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7479060784210723e-05,
+      "loss": 0.6057,
+      "step": 7017
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747828204504986e-05,
+      "loss": 0.6269,
+      "step": 7018
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7477503202981266e-05,
+      "loss": 0.6478,
+      "step": 7019
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7476724258015655e-05,
+      "loss": 0.6901,
+      "step": 7020
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747594521016375e-05,
+      "loss": 0.6603,
+      "step": 7021
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747516605943627e-05,
+      "loss": 0.6901,
+      "step": 7022
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7474386805843936e-05,
+      "loss": 0.6791,
+      "step": 7023
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7473607449397475e-05,
+      "loss": 0.6217,
+      "step": 7024
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7472827990107607e-05,
+      "loss": 0.6872,
+      "step": 7025
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7472048427985056e-05,
+      "loss": 0.6666,
+      "step": 7026
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747126876304056e-05,
+      "loss": 0.6424,
+      "step": 7027
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7470488995284833e-05,
+      "loss": 0.6967,
+      "step": 7028
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746970912472862e-05,
+      "loss": 0.6525,
+      "step": 7029
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7468929151382644e-05,
+      "loss": 0.75,
+      "step": 7030
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7468149075257638e-05,
+      "loss": 0.6692,
+      "step": 7031
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746736889636434e-05,
+      "loss": 0.6172,
+      "step": 7032
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7466588614713484e-05,
+      "loss": 0.6721,
+      "step": 7033
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746580823031581e-05,
+      "loss": 0.6088,
+      "step": 7034
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7465027743182053e-05,
+      "loss": 0.7718,
+      "step": 7035
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7464247153322955e-05,
+      "loss": 0.7044,
+      "step": 7036
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7463466460749257e-05,
+      "loss": 0.6404,
+      "step": 7037
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7462685665471703e-05,
+      "loss": 0.688,
+      "step": 7038
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746190476750104e-05,
+      "loss": 0.6943,
+      "step": 7039
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7461123766848005e-05,
+      "loss": 0.7129,
+      "step": 7040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7460342663523352e-05,
+      "loss": 0.6667,
+      "step": 7041
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7459561457537835e-05,
+      "loss": 0.6389,
+      "step": 7042
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7458780148902193e-05,
+      "loss": 0.6617,
+      "step": 7043
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7457998737627183e-05,
+      "loss": 0.6769,
+      "step": 7044
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7457217223723556e-05,
+      "loss": 0.7037,
+      "step": 7045
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7456435607202067e-05,
+      "loss": 0.5999,
+      "step": 7046
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7455653888073473e-05,
+      "loss": 0.6792,
+      "step": 7047
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7454872066348533e-05,
+      "loss": 0.6771,
+      "step": 7048
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7454090142038e-05,
+      "loss": 0.6901,
+      "step": 7049
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7453308115152638e-05,
+      "loss": 0.7302,
+      "step": 7050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745252598570321e-05,
+      "loss": 0.7099,
+      "step": 7051
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7451743753700472e-05,
+      "loss": 0.7355,
+      "step": 7052
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7450961419155192e-05,
+      "loss": 0.6731,
+      "step": 7053
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745017898207814e-05,
+      "loss": 0.6555,
+      "step": 7054
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7449396442480076e-05,
+      "loss": 0.6551,
+      "step": 7055
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7448613800371772e-05,
+      "loss": 0.6219,
+      "step": 7056
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7447831055763995e-05,
+      "loss": 0.5901,
+      "step": 7057
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744704820866752e-05,
+      "loss": 0.7078,
+      "step": 7058
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744626525909312e-05,
+      "loss": 0.6814,
+      "step": 7059
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7445482207051564e-05,
+      "loss": 0.7151,
+      "step": 7060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7444699052553634e-05,
+      "loss": 0.6718,
+      "step": 7061
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7443915795610104e-05,
+      "loss": 0.6315,
+      "step": 7062
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7443132436231746e-05,
+      "loss": 0.5919,
+      "step": 7063
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744234897442935e-05,
+      "loss": 0.6865,
+      "step": 7064
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7441565410213693e-05,
+      "loss": 0.6635,
+      "step": 7065
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7440781743595556e-05,
+      "loss": 0.6606,
+      "step": 7066
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7439997974585722e-05,
+      "loss": 0.6157,
+      "step": 7067
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7439214103194985e-05,
+      "loss": 0.6661,
+      "step": 7068
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743843012943412e-05,
+      "loss": 0.7342,
+      "step": 7069
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743764605331392e-05,
+      "loss": 0.6556,
+      "step": 7070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7436861874845178e-05,
+      "loss": 0.7201,
+      "step": 7071
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7436077594038678e-05,
+      "loss": 0.6766,
+      "step": 7072
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743529321090522e-05,
+      "loss": 0.5914,
+      "step": 7073
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7434508725455594e-05,
+      "loss": 0.6193,
+      "step": 7074
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7433724137700596e-05,
+      "loss": 0.6822,
+      "step": 7075
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7432939447651018e-05,
+      "loss": 0.6826,
+      "step": 7076
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7432154655317668e-05,
+      "loss": 0.6741,
+      "step": 7077
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7431369760711338e-05,
+      "loss": 0.6258,
+      "step": 7078
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7430584763842824e-05,
+      "loss": 0.6533,
+      "step": 7079
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742979966472294e-05,
+      "loss": 0.7476,
+      "step": 7080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7429014463362483e-05,
+      "loss": 0.7084,
+      "step": 7081
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742822915977226e-05,
+      "loss": 0.6793,
+      "step": 7082
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7427443753963077e-05,
+      "loss": 0.7062,
+      "step": 7083
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7426658245945742e-05,
+      "loss": 0.7156,
+      "step": 7084
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742587263573106e-05,
+      "loss": 0.6715,
+      "step": 7085
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742508692332985e-05,
+      "loss": 0.6946,
+      "step": 7086
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742430110875292e-05,
+      "loss": 0.7171,
+      "step": 7087
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7423515192011078e-05,
+      "loss": 0.5907,
+      "step": 7088
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7422729173115148e-05,
+      "loss": 0.6613,
+      "step": 7089
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7421943052075942e-05,
+      "loss": 0.7312,
+      "step": 7090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742115682890428e-05,
+      "loss": 0.6761,
+      "step": 7091
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7420370503610975e-05,
+      "loss": 0.5967,
+      "step": 7092
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7419584076206854e-05,
+      "loss": 0.7106,
+      "step": 7093
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7418797546702737e-05,
+      "loss": 0.7509,
+      "step": 7094
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741801091510945e-05,
+      "loss": 0.7084,
+      "step": 7095
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7417224181437807e-05,
+      "loss": 0.6654,
+      "step": 7096
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7416437345698648e-05,
+      "loss": 0.7343,
+      "step": 7097
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741565040790279e-05,
+      "loss": 0.6911,
+      "step": 7098
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741486336806107e-05,
+      "loss": 0.69,
+      "step": 7099
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7414076226184313e-05,
+      "loss": 0.7505,
+      "step": 7100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7413288982283355e-05,
+      "loss": 0.5522,
+      "step": 7101
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7412501636369025e-05,
+      "loss": 0.6679,
+      "step": 7102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741171418845216e-05,
+      "loss": 0.5896,
+      "step": 7103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7410926638543594e-05,
+      "loss": 0.6708,
+      "step": 7104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7410138986654167e-05,
+      "loss": 0.7048,
+      "step": 7105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7409351232794716e-05,
+      "loss": 0.6836,
+      "step": 7106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7408563376976082e-05,
+      "loss": 0.624,
+      "step": 7107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7407775419209106e-05,
+      "loss": 0.7188,
+      "step": 7108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740698735950463e-05,
+      "loss": 0.6093,
+      "step": 7109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74061991978735e-05,
+      "loss": 0.7599,
+      "step": 7110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7405410934326562e-05,
+      "loss": 0.6384,
+      "step": 7111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7404622568874662e-05,
+      "loss": 0.6792,
+      "step": 7112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7403834101528653e-05,
+      "loss": 0.6949,
+      "step": 7113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7403045532299376e-05,
+      "loss": 0.6751,
+      "step": 7114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7402256861197687e-05,
+      "loss": 0.7052,
+      "step": 7115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7401468088234445e-05,
+      "loss": 0.7053,
+      "step": 7116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7400679213420492e-05,
+      "loss": 0.6955,
+      "step": 7117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7399890236766693e-05,
+      "loss": 0.6463,
+      "step": 7118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7399101158283905e-05,
+      "loss": 0.6163,
+      "step": 7119
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739831197798298e-05,
+      "loss": 0.7001,
+      "step": 7120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7397522695874786e-05,
+      "loss": 0.6319,
+      "step": 7121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7396733311970175e-05,
+      "loss": 0.6473,
+      "step": 7122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7395943826280018e-05,
+      "loss": 0.6948,
+      "step": 7123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739515423881517e-05,
+      "loss": 0.6576,
+      "step": 7124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7394364549586505e-05,
+      "loss": 0.7044,
+      "step": 7125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7393574758604887e-05,
+      "loss": 0.7257,
+      "step": 7126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7392784865881182e-05,
+      "loss": 0.5774,
+      "step": 7127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7391994871426262e-05,
+      "loss": 0.6821,
+      "step": 7128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7391204775251e-05,
+      "loss": 0.6837,
+      "step": 7129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7390414577366263e-05,
+      "loss": 0.6664,
+      "step": 7130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738962427778293e-05,
+      "loss": 0.5973,
+      "step": 7131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738883387651187e-05,
+      "loss": 0.6837,
+      "step": 7132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7388043373563967e-05,
+      "loss": 0.6814,
+      "step": 7133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7387252768950093e-05,
+      "loss": 0.6289,
+      "step": 7134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7386462062681132e-05,
+      "loss": 0.5664,
+      "step": 7135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7385671254767964e-05,
+      "loss": 0.7249,
+      "step": 7136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7384880345221466e-05,
+      "loss": 0.7334,
+      "step": 7137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7384089334052527e-05,
+      "loss": 0.6557,
+      "step": 7138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738329822127203e-05,
+      "loss": 0.6807,
+      "step": 7139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7382507006890865e-05,
+      "loss": 0.6573,
+      "step": 7140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7381715690919914e-05,
+      "loss": 0.7064,
+      "step": 7141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738092427337007e-05,
+      "loss": 0.6428,
+      "step": 7142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7380132754252222e-05,
+      "loss": 0.7208,
+      "step": 7143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7379341133577262e-05,
+      "loss": 0.7211,
+      "step": 7144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7378549411356083e-05,
+      "loss": 0.6814,
+      "step": 7145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737775758759958e-05,
+      "loss": 0.7202,
+      "step": 7146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7376965662318654e-05,
+      "loss": 0.675,
+      "step": 7147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7376173635524196e-05,
+      "loss": 0.6339,
+      "step": 7148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7375381507227108e-05,
+      "loss": 0.6023,
+      "step": 7149
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737458927743829e-05,
+      "loss": 0.6495,
+      "step": 7150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7373796946168645e-05,
+      "loss": 0.6849,
+      "step": 7151
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7373004513429074e-05,
+      "loss": 0.6151,
+      "step": 7152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7372211979230483e-05,
+      "loss": 0.7044,
+      "step": 7153
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7371419343583777e-05,
+      "loss": 0.5999,
+      "step": 7154
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7370626606499863e-05,
+      "loss": 0.5718,
+      "step": 7155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7369833767989647e-05,
+      "loss": 0.6585,
+      "step": 7156
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7369040828064046e-05,
+      "loss": 0.6529,
+      "step": 7157
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736824778673397e-05,
+      "loss": 0.6413,
+      "step": 7158
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7367454644010327e-05,
+      "loss": 0.6829,
+      "step": 7159
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736666139990404e-05,
+      "loss": 0.7035,
+      "step": 7160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7365868054426012e-05,
+      "loss": 0.6984,
+      "step": 7161
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736507460758717e-05,
+      "loss": 0.6792,
+      "step": 7162
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7364281059398433e-05,
+      "loss": 0.6747,
+      "step": 7163
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7363487409870714e-05,
+      "loss": 0.6059,
+      "step": 7164
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736269365901494e-05,
+      "loss": 0.5855,
+      "step": 7165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7361899806842028e-05,
+      "loss": 0.6543,
+      "step": 7166
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7361105853362912e-05,
+      "loss": 0.6735,
+      "step": 7167
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7360311798588508e-05,
+      "loss": 0.5992,
+      "step": 7168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735951764252975e-05,
+      "loss": 0.756,
+      "step": 7169
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7358723385197557e-05,
+      "loss": 0.6645,
+      "step": 7170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7357929026602867e-05,
+      "loss": 0.6439,
+      "step": 7171
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7357134566756606e-05,
+      "loss": 0.6384,
+      "step": 7172
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735634000566971e-05,
+      "loss": 0.6177,
+      "step": 7173
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7355545343353116e-05,
+      "loss": 0.6946,
+      "step": 7174
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735475057981775e-05,
+      "loss": 0.6434,
+      "step": 7175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7353955715074556e-05,
+      "loss": 0.6907,
+      "step": 7176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7353160749134467e-05,
+      "loss": 0.5798,
+      "step": 7177
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7352365682008426e-05,
+      "loss": 0.6614,
+      "step": 7178
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7351570513707376e-05,
+      "loss": 0.5771,
+      "step": 7179
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7350775244242253e-05,
+      "loss": 0.7194,
+      "step": 7180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7349979873624002e-05,
+      "loss": 0.6829,
+      "step": 7181
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7349184401863573e-05,
+      "loss": 0.6514,
+      "step": 7182
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734838882897191e-05,
+      "loss": 0.5959,
+      "step": 7183
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7347593154959955e-05,
+      "loss": 0.5803,
+      "step": 7184
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7346797379838667e-05,
+      "loss": 0.729,
+      "step": 7185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734600150361899e-05,
+      "loss": 0.6206,
+      "step": 7186
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7345205526311875e-05,
+      "loss": 0.6951,
+      "step": 7187
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7344409447928282e-05,
+      "loss": 0.6712,
+      "step": 7188
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734361326847916e-05,
+      "loss": 0.7563,
+      "step": 7189
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7342816987975465e-05,
+      "loss": 0.6128,
+      "step": 7190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734202060642816e-05,
+      "loss": 0.6877,
+      "step": 7191
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7341224123848196e-05,
+      "loss": 0.7228,
+      "step": 7192
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734042754024654e-05,
+      "loss": 0.6498,
+      "step": 7193
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733963085563415e-05,
+      "loss": 0.6089,
+      "step": 7194
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733883407002199e-05,
+      "loss": 0.6399,
+      "step": 7195
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7338037183421028e-05,
+      "loss": 0.5741,
+      "step": 7196
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7337240195842224e-05,
+      "loss": 0.6261,
+      "step": 7197
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7336443107296547e-05,
+      "loss": 0.6286,
+      "step": 7198
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7335645917794968e-05,
+      "loss": 0.6976,
+      "step": 7199
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7334848627348454e-05,
+      "loss": 0.6283,
+      "step": 7200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733405123596798e-05,
+      "loss": 0.6384,
+      "step": 7201
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7333253743664516e-05,
+      "loss": 0.7254,
+      "step": 7202
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7332456150449037e-05,
+      "loss": 0.6571,
+      "step": 7203
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7331658456332516e-05,
+      "loss": 0.7554,
+      "step": 7204
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7330860661325935e-05,
+      "loss": 0.6,
+      "step": 7205
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733006276544027e-05,
+      "loss": 0.62,
+      "step": 7206
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73292647686865e-05,
+      "loss": 0.7098,
+      "step": 7207
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7328466671075607e-05,
+      "loss": 0.6791,
+      "step": 7208
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732766847261857e-05,
+      "loss": 0.7304,
+      "step": 7209
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732687017332638e-05,
+      "loss": 0.5802,
+      "step": 7210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732607177321002e-05,
+      "loss": 0.6309,
+      "step": 7211
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7325273272280473e-05,
+      "loss": 0.7075,
+      "step": 7212
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732447467054873e-05,
+      "loss": 0.626,
+      "step": 7213
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7323675968025776e-05,
+      "loss": 0.7011,
+      "step": 7214
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7322877164722606e-05,
+      "loss": 0.6405,
+      "step": 7215
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7322078260650216e-05,
+      "loss": 0.6283,
+      "step": 7216
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7321279255819595e-05,
+      "loss": 0.6914,
+      "step": 7217
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7320480150241734e-05,
+      "loss": 0.53,
+      "step": 7218
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7319680943927636e-05,
+      "loss": 0.5566,
+      "step": 7219
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7318881636888297e-05,
+      "loss": 0.5588,
+      "step": 7220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7318082229134718e-05,
+      "loss": 0.7426,
+      "step": 7221
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7317282720677893e-05,
+      "loss": 0.7214,
+      "step": 7222
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731648311152883e-05,
+      "loss": 0.6464,
+      "step": 7223
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7315683401698532e-05,
+      "loss": 0.638,
+      "step": 7224
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7314883591198005e-05,
+      "loss": 0.6704,
+      "step": 7225
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7314083680038244e-05,
+      "loss": 0.7144,
+      "step": 7226
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731328366823027e-05,
+      "loss": 0.6563,
+      "step": 7227
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7312483555785087e-05,
+      "loss": 0.6918,
+      "step": 7228
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7311683342713707e-05,
+      "loss": 0.6005,
+      "step": 7229
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7310883029027135e-05,
+      "loss": 0.6812,
+      "step": 7230
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731008261473639e-05,
+      "loss": 0.7671,
+      "step": 7231
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7309282099852484e-05,
+      "loss": 0.6805,
+      "step": 7232
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7308481484386436e-05,
+      "loss": 0.6107,
+      "step": 7233
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730768076834926e-05,
+      "loss": 0.6024,
+      "step": 7234
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7306879951751977e-05,
+      "loss": 0.7393,
+      "step": 7235
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73060790346056e-05,
+      "loss": 0.6413,
+      "step": 7236
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7305278016921162e-05,
+      "loss": 0.6747,
+      "step": 7237
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730447689870968e-05,
+      "loss": 0.7058,
+      "step": 7238
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7303675679982173e-05,
+      "loss": 0.5627,
+      "step": 7239
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7302874360749674e-05,
+      "loss": 0.6221,
+      "step": 7240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7302072941023204e-05,
+      "loss": 0.7264,
+      "step": 7241
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7301271420813795e-05,
+      "loss": 0.5707,
+      "step": 7242
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730046980013248e-05,
+      "loss": 0.6693,
+      "step": 7243
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7299668078990283e-05,
+      "loss": 0.7072,
+      "step": 7244
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7298866257398237e-05,
+      "loss": 0.5841,
+      "step": 7245
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7298064335367382e-05,
+      "loss": 0.6991,
+      "step": 7246
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7297262312908747e-05,
+      "loss": 0.6848,
+      "step": 7247
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729646019003337e-05,
+      "loss": 0.6581,
+      "step": 7248
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729565796675229e-05,
+      "loss": 0.601,
+      "step": 7249
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7294855643076548e-05,
+      "loss": 0.5528,
+      "step": 7250
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729405321901718e-05,
+      "loss": 0.6518,
+      "step": 7251
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729325069458523e-05,
+      "loss": 0.6178,
+      "step": 7252
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7292448069791742e-05,
+      "loss": 0.6918,
+      "step": 7253
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7291645344647764e-05,
+      "loss": 0.6284,
+      "step": 7254
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7290842519164334e-05,
+      "loss": 0.5981,
+      "step": 7255
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729003959335251e-05,
+      "loss": 0.6128,
+      "step": 7256
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7289236567223337e-05,
+      "loss": 0.707,
+      "step": 7257
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7288433440787857e-05,
+      "loss": 0.6606,
+      "step": 7258
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728763021405713e-05,
+      "loss": 0.685,
+      "step": 7259
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7286826887042213e-05,
+      "loss": 0.7781,
+      "step": 7260
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7286023459754152e-05,
+      "loss": 0.6364,
+      "step": 7261
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7285219932204004e-05,
+      "loss": 0.6306,
+      "step": 7262
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7284416304402832e-05,
+      "loss": 0.6539,
+      "step": 7263
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728361257636169e-05,
+      "loss": 0.6188,
+      "step": 7264
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7282808748091637e-05,
+      "loss": 0.7688,
+      "step": 7265
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7282004819603736e-05,
+      "loss": 0.6446,
+      "step": 7266
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7281200790909056e-05,
+      "loss": 0.7427,
+      "step": 7267
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728039666201865e-05,
+      "loss": 0.6439,
+      "step": 7268
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727959243294359e-05,
+      "loss": 0.5397,
+      "step": 7269
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7278788103694944e-05,
+      "loss": 0.6946,
+      "step": 7270
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7277983674283776e-05,
+      "loss": 0.6111,
+      "step": 7271
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7277179144721156e-05,
+      "loss": 0.677,
+      "step": 7272
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7276374515018158e-05,
+      "loss": 0.7519,
+      "step": 7273
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7275569785185854e-05,
+      "loss": 0.5335,
+      "step": 7274
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7274764955235315e-05,
+      "loss": 0.6455,
+      "step": 7275
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727396002517762e-05,
+      "loss": 0.6205,
+      "step": 7276
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7273154995023844e-05,
+      "loss": 0.6951,
+      "step": 7277
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7272349864785062e-05,
+      "loss": 0.6506,
+      "step": 7278
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7271544634472356e-05,
+      "loss": 0.6865,
+      "step": 7279
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727073930409681e-05,
+      "loss": 0.6193,
+      "step": 7280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72699338736695e-05,
+      "loss": 0.6972,
+      "step": 7281
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726912834320151e-05,
+      "loss": 0.6721,
+      "step": 7282
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7268322712703932e-05,
+      "loss": 0.6304,
+      "step": 7283
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7267516982187843e-05,
+      "loss": 0.6385,
+      "step": 7284
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7266711151664334e-05,
+      "loss": 0.6601,
+      "step": 7285
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7265905221144497e-05,
+      "loss": 0.6363,
+      "step": 7286
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7265099190639417e-05,
+      "loss": 0.5981,
+      "step": 7287
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7264293060160194e-05,
+      "loss": 0.6828,
+      "step": 7288
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726348682971791e-05,
+      "loss": 0.6579,
+      "step": 7289
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7262680499323664e-05,
+      "loss": 0.5782,
+      "step": 7290
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7261874068988557e-05,
+      "loss": 0.6664,
+      "step": 7291
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726106753872368e-05,
+      "loss": 0.6259,
+      "step": 7292
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7260260908540135e-05,
+      "loss": 0.7164,
+      "step": 7293
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725945417844902e-05,
+      "loss": 0.6055,
+      "step": 7294
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7258647348461436e-05,
+      "loss": 0.5994,
+      "step": 7295
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7257840418588486e-05,
+      "loss": 0.6387,
+      "step": 7296
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7257033388841277e-05,
+      "loss": 0.6795,
+      "step": 7297
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7256226259230907e-05,
+      "loss": 0.6973,
+      "step": 7298
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725541902976849e-05,
+      "loss": 0.689,
+      "step": 7299
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7254611700465137e-05,
+      "loss": 0.6241,
+      "step": 7300
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7253804271331946e-05,
+      "loss": 0.6928,
+      "step": 7301
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725299674238004e-05,
+      "loss": 0.622,
+      "step": 7302
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7252189113620522e-05,
+      "loss": 0.6452,
+      "step": 7303
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725138138506451e-05,
+      "loss": 0.6836,
+      "step": 7304
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7250573556723118e-05,
+      "loss": 0.5749,
+      "step": 7305
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7249765628607463e-05,
+      "loss": 0.616,
+      "step": 7306
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7248957600728664e-05,
+      "loss": 0.5956,
+      "step": 7307
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724814947309784e-05,
+      "loss": 0.6036,
+      "step": 7308
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7247341245726107e-05,
+      "loss": 0.7372,
+      "step": 7309
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724653291862459e-05,
+      "loss": 0.6894,
+      "step": 7310
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7245724491804418e-05,
+      "loss": 0.6401,
+      "step": 7311
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7244915965276706e-05,
+      "loss": 0.709,
+      "step": 7312
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7244107339052587e-05,
+      "loss": 0.7223,
+      "step": 7313
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7243298613143182e-05,
+      "loss": 0.7194,
+      "step": 7314
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724248978755963e-05,
+      "loss": 0.71,
+      "step": 7315
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724168086231305e-05,
+      "loss": 0.5864,
+      "step": 7316
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7240871837414577e-05,
+      "loss": 0.6544,
+      "step": 7317
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7240062712875346e-05,
+      "loss": 0.6933,
+      "step": 7318
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723925348870649e-05,
+      "loss": 0.6131,
+      "step": 7319
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723844416491915e-05,
+      "loss": 0.6919,
+      "step": 7320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7237634741524455e-05,
+      "loss": 0.5593,
+      "step": 7321
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7236825218533547e-05,
+      "loss": 0.6493,
+      "step": 7322
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7236015595957562e-05,
+      "loss": 0.6288,
+      "step": 7323
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7235205873807646e-05,
+      "loss": 0.6966,
+      "step": 7324
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7234396052094943e-05,
+      "loss": 0.6538,
+      "step": 7325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7233586130830587e-05,
+      "loss": 0.67,
+      "step": 7326
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7232776110025734e-05,
+      "loss": 0.6718,
+      "step": 7327
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723196598969153e-05,
+      "loss": 0.5405,
+      "step": 7328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7231155769839113e-05,
+      "loss": 0.7358,
+      "step": 7329
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723034545047964e-05,
+      "loss": 0.617,
+      "step": 7330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7229535031624264e-05,
+      "loss": 0.6436,
+      "step": 7331
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7228724513284133e-05,
+      "loss": 0.7022,
+      "step": 7332
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7227913895470395e-05,
+      "loss": 0.6811,
+      "step": 7333
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722710317819422e-05,
+      "loss": 0.6222,
+      "step": 7334
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722629236146675e-05,
+      "loss": 0.6085,
+      "step": 7335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7225481445299143e-05,
+      "loss": 0.6273,
+      "step": 7336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7224670429702568e-05,
+      "loss": 0.7123,
+      "step": 7337
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7223859314688177e-05,
+      "loss": 0.7024,
+      "step": 7338
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7223048100267138e-05,
+      "loss": 0.6973,
+      "step": 7339
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7222236786450607e-05,
+      "loss": 0.6604,
+      "step": 7340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7221425373249757e-05,
+      "loss": 0.654,
+      "step": 7341
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7220613860675742e-05,
+      "loss": 0.7102,
+      "step": 7342
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7219802248739742e-05,
+      "loss": 0.6831,
+      "step": 7343
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7218990537452913e-05,
+      "loss": 0.6886,
+      "step": 7344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7218178726826433e-05,
+      "loss": 0.6421,
+      "step": 7345
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7217366816871472e-05,
+      "loss": 0.6224,
+      "step": 7346
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.72165548075992e-05,
+      "loss": 0.6405,
+      "step": 7347
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7215742699020797e-05,
+      "loss": 0.679,
+      "step": 7348
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7214930491147432e-05,
+      "loss": 0.755,
+      "step": 7349
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7214118183990285e-05,
+      "loss": 0.6835,
+      "step": 7350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7213305777560532e-05,
+      "loss": 0.6241,
+      "step": 7351
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7212493271869352e-05,
+      "loss": 0.6846,
+      "step": 7352
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7211680666927927e-05,
+      "loss": 0.6243,
+      "step": 7353
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721086796274744e-05,
+      "loss": 0.695,
+      "step": 7354
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7210055159339078e-05,
+      "loss": 0.6961,
+      "step": 7355
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7209242256714015e-05,
+      "loss": 0.6938,
+      "step": 7356
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720842925488345e-05,
+      "loss": 0.6757,
+      "step": 7357
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720761615385856e-05,
+      "loss": 0.5809,
+      "step": 7358
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720680295365054e-05,
+      "loss": 0.6536,
+      "step": 7359
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7205989654270577e-05,
+      "loss": 0.5925,
+      "step": 7360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720517625572987e-05,
+      "loss": 0.6667,
+      "step": 7361
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.72043627580396e-05,
+      "loss": 0.5962,
+      "step": 7362
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7203549161210973e-05,
+      "loss": 0.6964,
+      "step": 7363
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7202735465255177e-05,
+      "loss": 0.6678,
+      "step": 7364
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7201921670183414e-05,
+      "loss": 0.7318,
+      "step": 7365
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720110777600688e-05,
+      "loss": 0.6096,
+      "step": 7366
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7200293782736776e-05,
+      "loss": 0.6913,
+      "step": 7367
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7199479690384303e-05,
+      "loss": 0.7482,
+      "step": 7368
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7198665498960664e-05,
+      "loss": 0.6406,
+      "step": 7369
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719785120847706e-05,
+      "loss": 0.5836,
+      "step": 7370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71970368189447e-05,
+      "loss": 0.6712,
+      "step": 7371
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719622233037479e-05,
+      "loss": 0.591,
+      "step": 7372
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7195407742778537e-05,
+      "loss": 0.6685,
+      "step": 7373
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719459305616715e-05,
+      "loss": 0.6963,
+      "step": 7374
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7193778270551845e-05,
+      "loss": 0.6203,
+      "step": 7375
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7192963385943825e-05,
+      "loss": 0.5479,
+      "step": 7376
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7192148402354313e-05,
+      "loss": 0.7158,
+      "step": 7377
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7191333319794514e-05,
+      "loss": 0.708,
+      "step": 7378
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7190518138275652e-05,
+      "loss": 0.5963,
+      "step": 7379
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7189702857808946e-05,
+      "loss": 0.5729,
+      "step": 7380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718888747840561e-05,
+      "loss": 0.686,
+      "step": 7381
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7188072000076862e-05,
+      "loss": 0.6588,
+      "step": 7382
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7187256422833928e-05,
+      "loss": 0.6797,
+      "step": 7383
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7186440746688032e-05,
+      "loss": 0.7199,
+      "step": 7384
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7185624971650395e-05,
+      "loss": 0.6441,
+      "step": 7385
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7184809097732247e-05,
+      "loss": 0.5946,
+      "step": 7386
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718399312494481e-05,
+      "loss": 0.6803,
+      "step": 7387
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7183177053299317e-05,
+      "loss": 0.6411,
+      "step": 7388
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7182360882806998e-05,
+      "loss": 0.6334,
+      "step": 7389
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7181544613479077e-05,
+      "loss": 0.6988,
+      "step": 7390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7180728245326796e-05,
+      "loss": 0.5867,
+      "step": 7391
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7179911778361386e-05,
+      "loss": 0.6438,
+      "step": 7392
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7179095212594078e-05,
+      "loss": 0.6209,
+      "step": 7393
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717827854803611e-05,
+      "loss": 0.6607,
+      "step": 7394
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7177461784698725e-05,
+      "loss": 0.6543,
+      "step": 7395
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7176644922593156e-05,
+      "loss": 0.7433,
+      "step": 7396
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717582796173065e-05,
+      "loss": 0.7048,
+      "step": 7397
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717501090212245e-05,
+      "loss": 0.7253,
+      "step": 7398
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717419374377979e-05,
+      "loss": 0.6314,
+      "step": 7399
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717337648671392e-05,
+      "loss": 0.7193,
+      "step": 7400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7172559130936088e-05,
+      "loss": 0.6255,
+      "step": 7401
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717174167645754e-05,
+      "loss": 0.5674,
+      "step": 7402
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7170924123289522e-05,
+      "loss": 0.6867,
+      "step": 7403
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7170106471443292e-05,
+      "loss": 0.6945,
+      "step": 7404
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7169288720930093e-05,
+      "loss": 0.6761,
+      "step": 7405
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7168470871761184e-05,
+      "loss": 0.6192,
+      "step": 7406
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7167652923947815e-05,
+      "loss": 0.6713,
+      "step": 7407
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7166834877501247e-05,
+      "loss": 0.7456,
+      "step": 7408
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7166016732432732e-05,
+      "loss": 0.6567,
+      "step": 7409
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7165198488753526e-05,
+      "loss": 0.6599,
+      "step": 7410
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7164380146474897e-05,
+      "loss": 0.7087,
+      "step": 7411
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7163561705608104e-05,
+      "loss": 0.687,
+      "step": 7412
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7162743166164403e-05,
+      "loss": 0.6244,
+      "step": 7413
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7161924528155066e-05,
+      "loss": 0.6352,
+      "step": 7414
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716110579159135e-05,
+      "loss": 0.6381,
+      "step": 7415
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716028695648453e-05,
+      "loss": 0.6181,
+      "step": 7416
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715946802284587e-05,
+      "loss": 0.5844,
+      "step": 7417
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715864899068664e-05,
+      "loss": 0.6789,
+      "step": 7418
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7157829860018104e-05,
+      "loss": 0.6715,
+      "step": 7419
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7157010630851544e-05,
+      "loss": 0.6538,
+      "step": 7420
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7156191303198227e-05,
+      "loss": 0.7334,
+      "step": 7421
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7155371877069433e-05,
+      "loss": 0.5578,
+      "step": 7422
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7154552352476432e-05,
+      "loss": 0.6708,
+      "step": 7423
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7153732729430506e-05,
+      "loss": 0.5642,
+      "step": 7424
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715291300794293e-05,
+      "loss": 0.6765,
+      "step": 7425
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715209318802499e-05,
+      "loss": 0.7453,
+      "step": 7426
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7151273269687954e-05,
+      "loss": 0.7429,
+      "step": 7427
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7150453252943126e-05,
+      "loss": 0.6509,
+      "step": 7428
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714963313780177e-05,
+      "loss": 0.6342,
+      "step": 7429
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7148812924275183e-05,
+      "loss": 0.6777,
+      "step": 7430
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7147992612374647e-05,
+      "loss": 0.6913,
+      "step": 7431
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7147172202111454e-05,
+      "loss": 0.7006,
+      "step": 7432
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714635169349689e-05,
+      "loss": 0.6282,
+      "step": 7433
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7145531086542246e-05,
+      "loss": 0.6274,
+      "step": 7434
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714471038125882e-05,
+      "loss": 0.6073,
+      "step": 7435
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7143889577657896e-05,
+      "loss": 0.6134,
+      "step": 7436
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7143068675750777e-05,
+      "loss": 0.6287,
+      "step": 7437
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7142247675548754e-05,
+      "loss": 0.6837,
+      "step": 7438
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7141426577063133e-05,
+      "loss": 0.6642,
+      "step": 7439
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7140605380305203e-05,
+      "loss": 0.6048,
+      "step": 7440
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7139784085286268e-05,
+      "loss": 0.6492,
+      "step": 7441
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7138962692017633e-05,
+      "loss": 0.7338,
+      "step": 7442
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7138141200510595e-05,
+      "loss": 0.6981,
+      "step": 7443
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7137319610776466e-05,
+      "loss": 0.5891,
+      "step": 7444
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7136497922826546e-05,
+      "loss": 0.6473,
+      "step": 7445
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7135676136672144e-05,
+      "loss": 0.6825,
+      "step": 7446
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7134854252324567e-05,
+      "loss": 0.7014,
+      "step": 7447
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713403226979513e-05,
+      "loss": 0.7132,
+      "step": 7448
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7133210189095136e-05,
+      "loss": 0.6899,
+      "step": 7449
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7132388010235908e-05,
+      "loss": 0.7144,
+      "step": 7450
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713156573322875e-05,
+      "loss": 0.6748,
+      "step": 7451
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713074335808498e-05,
+      "loss": 0.6058,
+      "step": 7452
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7129920884815917e-05,
+      "loss": 0.6513,
+      "step": 7453
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712909831343288e-05,
+      "loss": 0.7371,
+      "step": 7454
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712827564394718e-05,
+      "loss": 0.5869,
+      "step": 7455
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712745287637015e-05,
+      "loss": 0.666,
+      "step": 7456
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7126630010713105e-05,
+      "loss": 0.6315,
+      "step": 7457
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7125807046987364e-05,
+      "loss": 0.6252,
+      "step": 7458
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712498398520426e-05,
+      "loss": 0.6106,
+      "step": 7459
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712416082537512e-05,
+      "loss": 0.6654,
+      "step": 7460
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712333756751126e-05,
+      "loss": 0.6643,
+      "step": 7461
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712251421162402e-05,
+      "loss": 0.7304,
+      "step": 7462
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7121690757724722e-05,
+      "loss": 0.6203,
+      "step": 7463
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7120867205824704e-05,
+      "loss": 0.7018,
+      "step": 7464
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71200435559353e-05,
+      "loss": 0.6784,
+      "step": 7465
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7119219808067838e-05,
+      "loss": 0.6093,
+      "step": 7466
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7118395962233655e-05,
+      "loss": 0.6875,
+      "step": 7467
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711757201844409e-05,
+      "loss": 0.6672,
+      "step": 7468
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7116747976710476e-05,
+      "loss": 0.5994,
+      "step": 7469
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7115923837044158e-05,
+      "loss": 0.6844,
+      "step": 7470
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711509959945648e-05,
+      "loss": 0.6175,
+      "step": 7471
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7114275263958778e-05,
+      "loss": 0.7091,
+      "step": 7472
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7113450830562396e-05,
+      "loss": 0.6992,
+      "step": 7473
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711262629927868e-05,
+      "loss": 0.7141,
+      "step": 7474
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7111801670118977e-05,
+      "loss": 0.6719,
+      "step": 7475
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711097694309463e-05,
+      "loss": 0.6625,
+      "step": 7476
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7110152118216997e-05,
+      "loss": 0.5743,
+      "step": 7477
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710932719549742e-05,
+      "loss": 0.6134,
+      "step": 7478
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710850217494726e-05,
+      "loss": 0.696,
+      "step": 7479
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7107677056577856e-05,
+      "loss": 0.7162,
+      "step": 7480
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7106851840400574e-05,
+      "loss": 0.654,
+      "step": 7481
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710602652642677e-05,
+      "loss": 0.6495,
+      "step": 7482
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710520111466779e-05,
+      "loss": 0.6939,
+      "step": 7483
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7104375605135e-05,
+      "loss": 0.6631,
+      "step": 7484
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7103549997839764e-05,
+      "loss": 0.7169,
+      "step": 7485
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7102724292793435e-05,
+      "loss": 0.6424,
+      "step": 7486
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7101898490007382e-05,
+      "loss": 0.5806,
+      "step": 7487
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710107258949296e-05,
+      "loss": 0.5537,
+      "step": 7488
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7100246591261545e-05,
+      "loss": 0.7264,
+      "step": 7489
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7099420495324497e-05,
+      "loss": 0.7317,
+      "step": 7490
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7098594301693183e-05,
+      "loss": 0.5755,
+      "step": 7491
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7097768010378976e-05,
+      "loss": 0.6196,
+      "step": 7492
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709694162139324e-05,
+      "loss": 0.6786,
+      "step": 7493
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7096115134747358e-05,
+      "loss": 0.6585,
+      "step": 7494
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7095288550452695e-05,
+      "loss": 0.6553,
+      "step": 7495
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7094461868520625e-05,
+      "loss": 0.6364,
+      "step": 7496
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7093635088962526e-05,
+      "loss": 0.5784,
+      "step": 7497
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7092808211789776e-05,
+      "loss": 0.7062,
+      "step": 7498
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7091981237013754e-05,
+      "loss": 0.585,
+      "step": 7499
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7091154164645835e-05,
+      "loss": 0.5858,
+      "step": 7500
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7090326994697406e-05,
+      "loss": 0.6657,
+      "step": 7501
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708949972717985e-05,
+      "loss": 0.6657,
+      "step": 7502
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7088672362104543e-05,
+      "loss": 0.703,
+      "step": 7503
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708784489948288e-05,
+      "loss": 0.6935,
+      "step": 7504
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708701733932624e-05,
+      "loss": 0.7153,
+      "step": 7505
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7086189681646016e-05,
+      "loss": 0.7708,
+      "step": 7506
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7085361926453596e-05,
+      "loss": 0.6409,
+      "step": 7507
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708453407376037e-05,
+      "loss": 0.7021,
+      "step": 7508
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708370612357773e-05,
+      "loss": 0.6884,
+      "step": 7509
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7082878075917068e-05,
+      "loss": 0.7044,
+      "step": 7510
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7082049930789782e-05,
+      "loss": 0.6969,
+      "step": 7511
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7081221688207262e-05,
+      "loss": 0.6707,
+      "step": 7512
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7080393348180914e-05,
+      "loss": 0.6938,
+      "step": 7513
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7079564910722134e-05,
+      "loss": 0.5893,
+      "step": 7514
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7078736375842317e-05,
+      "loss": 0.661,
+      "step": 7515
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7077907743552867e-05,
+      "loss": 0.7343,
+      "step": 7516
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7077079013865186e-05,
+      "loss": 0.6627,
+      "step": 7517
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7076250186790685e-05,
+      "loss": 0.5994,
+      "step": 7518
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7075421262340757e-05,
+      "loss": 0.7399,
+      "step": 7519
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707459224052682e-05,
+      "loss": 0.6746,
+      "step": 7520
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7073763121360273e-05,
+      "loss": 0.6981,
+      "step": 7521
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7072933904852535e-05,
+      "loss": 0.6402,
+      "step": 7522
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707210459101501e-05,
+      "loss": 0.6888,
+      "step": 7523
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707127517985911e-05,
+      "loss": 0.6176,
+      "step": 7524
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7070445671396252e-05,
+      "loss": 0.7256,
+      "step": 7525
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7069616065637847e-05,
+      "loss": 0.7373,
+      "step": 7526
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7068786362595316e-05,
+      "loss": 0.5876,
+      "step": 7527
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7067956562280074e-05,
+      "loss": 0.6379,
+      "step": 7528
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7067126664703534e-05,
+      "loss": 0.6593,
+      "step": 7529
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7066296669877123e-05,
+      "loss": 0.7474,
+      "step": 7530
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7065466577812264e-05,
+      "loss": 0.6648,
+      "step": 7531
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7064636388520373e-05,
+      "loss": 0.6561,
+      "step": 7532
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706380610201288e-05,
+      "loss": 0.7311,
+      "step": 7533
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706297571830121e-05,
+      "loss": 0.604,
+      "step": 7534
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7062145237396786e-05,
+      "loss": 0.6441,
+      "step": 7535
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7061314659311036e-05,
+      "loss": 0.7051,
+      "step": 7536
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7060483984055394e-05,
+      "loss": 0.668,
+      "step": 7537
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705965321164129e-05,
+      "loss": 0.6586,
+      "step": 7538
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705882234208015e-05,
+      "loss": 0.5917,
+      "step": 7539
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7057991375383417e-05,
+      "loss": 0.6837,
+      "step": 7540
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7057160311562517e-05,
+      "loss": 0.6918,
+      "step": 7541
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7056329150628893e-05,
+      "loss": 0.598,
+      "step": 7542
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705549789259398e-05,
+      "loss": 0.7279,
+      "step": 7543
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7054666537469213e-05,
+      "loss": 0.6837,
+      "step": 7544
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7053835085266038e-05,
+      "loss": 0.67,
+      "step": 7545
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7053003535995895e-05,
+      "loss": 0.6624,
+      "step": 7546
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7052171889670224e-05,
+      "loss": 0.6773,
+      "step": 7547
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7051340146300473e-05,
+      "loss": 0.5857,
+      "step": 7548
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7050508305898086e-05,
+      "loss": 0.6366,
+      "step": 7549
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704967636847451e-05,
+      "loss": 0.6241,
+      "step": 7550
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7048844334041187e-05,
+      "loss": 0.6425,
+      "step": 7551
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704801220260958e-05,
+      "loss": 0.6084,
+      "step": 7552
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7047179974191127e-05,
+      "loss": 0.6144,
+      "step": 7553
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7046347648797288e-05,
+      "loss": 0.6923,
+      "step": 7554
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7045515226439512e-05,
+      "loss": 0.6879,
+      "step": 7555
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7044682707129258e-05,
+      "loss": 0.706,
+      "step": 7556
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704385009087798e-05,
+      "loss": 0.6693,
+      "step": 7557
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7043017377697137e-05,
+      "loss": 0.6929,
+      "step": 7558
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704218456759818e-05,
+      "loss": 0.6898,
+      "step": 7559
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704135166059258e-05,
+      "loss": 0.6983,
+      "step": 7560
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7040518656691793e-05,
+      "loss": 0.7502,
+      "step": 7561
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7039685555907284e-05,
+      "loss": 0.5779,
+      "step": 7562
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7038852358250514e-05,
+      "loss": 0.6719,
+      "step": 7563
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703801906373295e-05,
+      "loss": 0.7083,
+      "step": 7564
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7037185672366065e-05,
+      "loss": 0.7049,
+      "step": 7565
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7036352184161318e-05,
+      "loss": 0.7091,
+      "step": 7566
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7035518599130184e-05,
+      "loss": 0.5621,
+      "step": 7567
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7034684917284126e-05,
+      "loss": 0.6664,
+      "step": 7568
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703385113863463e-05,
+      "loss": 0.7113,
+      "step": 7569
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7033017263193158e-05,
+      "loss": 0.6276,
+      "step": 7570
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7032183290971192e-05,
+      "loss": 0.675,
+      "step": 7571
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70313492219802e-05,
+      "loss": 0.6291,
+      "step": 7572
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7030515056231668e-05,
+      "loss": 0.6699,
+      "step": 7573
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702968079373707e-05,
+      "loss": 0.6505,
+      "step": 7574
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702884643450789e-05,
+      "loss": 0.6901,
+      "step": 7575
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7028011978555602e-05,
+      "loss": 0.6886,
+      "step": 7576
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7027177425891695e-05,
+      "loss": 0.738,
+      "step": 7577
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7026342776527654e-05,
+      "loss": 0.5806,
+      "step": 7578
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702550803047496e-05,
+      "loss": 0.63,
+      "step": 7579
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702467318774511e-05,
+      "loss": 0.7171,
+      "step": 7580
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7023838248349574e-05,
+      "loss": 0.7334,
+      "step": 7581
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702300321229986e-05,
+      "loss": 0.699,
+      "step": 7582
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7022168079607446e-05,
+      "loss": 0.5823,
+      "step": 7583
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7021332850283833e-05,
+      "loss": 0.6548,
+      "step": 7584
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7020497524340505e-05,
+      "loss": 0.6589,
+      "step": 7585
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7019662101788967e-05,
+      "loss": 0.6509,
+      "step": 7586
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7018826582640708e-05,
+      "loss": 0.6461,
+      "step": 7587
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701799096690723e-05,
+      "loss": 0.6221,
+      "step": 7588
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701715525460003e-05,
+      "loss": 0.6282,
+      "step": 7589
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7016319445730608e-05,
+      "loss": 0.6682,
+      "step": 7590
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7015483540310462e-05,
+      "loss": 0.721,
+      "step": 7591
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70146475383511e-05,
+      "loss": 0.6785,
+      "step": 7592
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7013811439864026e-05,
+      "loss": 0.7315,
+      "step": 7593
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.701297524486074e-05,
+      "loss": 0.609,
+      "step": 7594
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7012138953352753e-05,
+      "loss": 0.7069,
+      "step": 7595
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7011302565351578e-05,
+      "loss": 0.5808,
+      "step": 7596
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7010466080868712e-05,
+      "loss": 0.6585,
+      "step": 7597
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7009629499915677e-05,
+      "loss": 0.6409,
+      "step": 7598
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7008792822503978e-05,
+      "loss": 0.691,
+      "step": 7599
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7007956048645136e-05,
+      "loss": 0.5877,
+      "step": 7600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7007119178350658e-05,
+      "loss": 0.7047,
+      "step": 7601
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700628221163206e-05,
+      "loss": 0.671,
+      "step": 7602
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700544514850087e-05,
+      "loss": 0.6303,
+      "step": 7603
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7004607988968594e-05,
+      "loss": 0.6455,
+      "step": 7604
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7003770733046757e-05,
+      "loss": 0.7329,
+      "step": 7605
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7002933380746883e-05,
+      "loss": 0.6809,
+      "step": 7606
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700209593208049e-05,
+      "loss": 0.6215,
+      "step": 7607
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7001258387059106e-05,
+      "loss": 0.659,
+      "step": 7608
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7000420745694256e-05,
+      "loss": 0.667,
+      "step": 7609
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699958300799746e-05,
+      "loss": 0.6029,
+      "step": 7610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699874517398026e-05,
+      "loss": 0.6894,
+      "step": 7611
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699790724365417e-05,
+      "loss": 0.6425,
+      "step": 7612
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699706921703073e-05,
+      "loss": 0.699,
+      "step": 7613
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699623109412147e-05,
+      "loss": 0.7271,
+      "step": 7614
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699539287493792e-05,
+      "loss": 0.6116,
+      "step": 7615
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699455455949162e-05,
+      "loss": 0.637,
+      "step": 7616
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69937161477941e-05,
+      "loss": 0.7064,
+      "step": 7617
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6992877639856903e-05,
+      "loss": 0.6619,
+      "step": 7618
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699203903569157e-05,
+      "loss": 0.6317,
+      "step": 7619
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6991200335309628e-05,
+      "loss": 0.6281,
+      "step": 7620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699036153872263e-05,
+      "loss": 0.683,
+      "step": 7621
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6989522645942112e-05,
+      "loss": 0.6287,
+      "step": 7622
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6988683656979624e-05,
+      "loss": 0.6403,
+      "step": 7623
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698784457184671e-05,
+      "loss": 0.74,
+      "step": 7624
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698700539055491e-05,
+      "loss": 0.6628,
+      "step": 7625
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698616611311578e-05,
+      "loss": 0.624,
+      "step": 7626
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6985326739540867e-05,
+      "loss": 0.7302,
+      "step": 7627
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6984487269841716e-05,
+      "loss": 0.672,
+      "step": 7628
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6983647704029886e-05,
+      "loss": 0.617,
+      "step": 7629
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6982808042116926e-05,
+      "loss": 0.6309,
+      "step": 7630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698196828411439e-05,
+      "loss": 0.6014,
+      "step": 7631
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698112843003384e-05,
+      "loss": 0.6633,
+      "step": 7632
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6980288479886822e-05,
+      "loss": 0.6793,
+      "step": 7633
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6979448433684908e-05,
+      "loss": 0.5825,
+      "step": 7634
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6978608291439648e-05,
+      "loss": 0.6006,
+      "step": 7635
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69777680531626e-05,
+      "loss": 0.6685,
+      "step": 7636
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697692771886534e-05,
+      "loss": 0.7172,
+      "step": 7637
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6976087288559425e-05,
+      "loss": 0.6508,
+      "step": 7638
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6975246762256414e-05,
+      "loss": 0.6331,
+      "step": 7639
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697440613996788e-05,
+      "loss": 0.6932,
+      "step": 7640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6973565421705392e-05,
+      "loss": 0.719,
+      "step": 7641
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6972724607480513e-05,
+      "loss": 0.676,
+      "step": 7642
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697188369730482e-05,
+      "loss": 0.6243,
+      "step": 7643
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697104269118988e-05,
+      "loss": 0.6477,
+      "step": 7644
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6970201589147266e-05,
+      "loss": 0.6165,
+      "step": 7645
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6969360391188554e-05,
+      "loss": 0.6943,
+      "step": 7646
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696851909732532e-05,
+      "loss": 0.641,
+      "step": 7647
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696767770756914e-05,
+      "loss": 0.6947,
+      "step": 7648
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696683622193159e-05,
+      "loss": 0.7126,
+      "step": 7649
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6965994640424254e-05,
+      "loss": 0.6109,
+      "step": 7650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6965152963058714e-05,
+      "loss": 0.5634,
+      "step": 7651
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6964311189846544e-05,
+      "loss": 0.6654,
+      "step": 7652
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6963469320799336e-05,
+      "loss": 0.6168,
+      "step": 7653
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696262735592867e-05,
+      "loss": 0.5755,
+      "step": 7654
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6961785295246134e-05,
+      "loss": 0.6294,
+      "step": 7655
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6960943138763312e-05,
+      "loss": 0.5554,
+      "step": 7656
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6960100886491798e-05,
+      "loss": 0.6693,
+      "step": 7657
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6959258538443185e-05,
+      "loss": 0.6489,
+      "step": 7658
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6958416094629054e-05,
+      "loss": 0.7242,
+      "step": 7659
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6957573555061003e-05,
+      "loss": 0.6584,
+      "step": 7660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695673091975063e-05,
+      "loss": 0.7398,
+      "step": 7661
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6955888188709527e-05,
+      "loss": 0.6547,
+      "step": 7662
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6955045361949287e-05,
+      "loss": 0.6302,
+      "step": 7663
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6954202439481517e-05,
+      "loss": 0.7114,
+      "step": 7664
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6953359421317804e-05,
+      "loss": 0.7198,
+      "step": 7665
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695251630746976e-05,
+      "loss": 0.6293,
+      "step": 7666
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695167309794898e-05,
+      "loss": 0.6478,
+      "step": 7667
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6950829792767073e-05,
+      "loss": 0.6631,
+      "step": 7668
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694998639193564e-05,
+      "loss": 0.6269,
+      "step": 7669
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6949142895466283e-05,
+      "loss": 0.6497,
+      "step": 7670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6948299303370616e-05,
+      "loss": 0.6532,
+      "step": 7671
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694745561566025e-05,
+      "loss": 0.562,
+      "step": 7672
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6946611832346787e-05,
+      "loss": 0.6607,
+      "step": 7673
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694576795344184e-05,
+      "loss": 0.6499,
+      "step": 7674
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694492397895702e-05,
+      "loss": 0.5736,
+      "step": 7675
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6944079908903952e-05,
+      "loss": 0.6773,
+      "step": 7676
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6943235743294237e-05,
+      "loss": 0.5925,
+      "step": 7677
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6942391482139496e-05,
+      "loss": 0.7146,
+      "step": 7678
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6941547125451355e-05,
+      "loss": 0.6209,
+      "step": 7679
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694070267324142e-05,
+      "loss": 0.6293,
+      "step": 7680
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693985812552132e-05,
+      "loss": 0.5836,
+      "step": 7681
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693901348230267e-05,
+      "loss": 0.696,
+      "step": 7682
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6938168743597105e-05,
+      "loss": 0.6062,
+      "step": 7683
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693732390941624e-05,
+      "loss": 0.6788,
+      "step": 7684
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6936478979771697e-05,
+      "loss": 0.5929,
+      "step": 7685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6935633954675112e-05,
+      "loss": 0.6309,
+      "step": 7686
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693478883413811e-05,
+      "loss": 0.7109,
+      "step": 7687
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693394361817232e-05,
+      "loss": 0.6906,
+      "step": 7688
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693309830678937e-05,
+      "loss": 0.6259,
+      "step": 7689
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6932252900000898e-05,
+      "loss": 0.6599,
+      "step": 7690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6931407397818532e-05,
+      "loss": 0.651,
+      "step": 7691
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6930561800253916e-05,
+      "loss": 0.6914,
+      "step": 7692
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6929716107318673e-05,
+      "loss": 0.5711,
+      "step": 7693
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6928870319024454e-05,
+      "loss": 0.7415,
+      "step": 7694
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6928024435382887e-05,
+      "loss": 0.6117,
+      "step": 7695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6927178456405617e-05,
+      "loss": 0.6759,
+      "step": 7696
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6926332382104286e-05,
+      "loss": 0.6114,
+      "step": 7697
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692548621249053e-05,
+      "loss": 0.7412,
+      "step": 7698
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6924639947576006e-05,
+      "loss": 0.631,
+      "step": 7699
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6923793587372347e-05,
+      "loss": 0.676,
+      "step": 7700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6922947131891208e-05,
+      "loss": 0.7235,
+      "step": 7701
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6922100581144228e-05,
+      "loss": 0.644,
+      "step": 7702
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6921253935143068e-05,
+      "loss": 0.7418,
+      "step": 7703
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6920407193899368e-05,
+      "loss": 0.6738,
+      "step": 7704
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6919560357424787e-05,
+      "loss": 0.6247,
+      "step": 7705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6918713425730974e-05,
+      "loss": 0.7161,
+      "step": 7706
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6917866398829583e-05,
+      "loss": 0.6184,
+      "step": 7707
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6917019276732274e-05,
+      "loss": 0.6526,
+      "step": 7708
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69161720594507e-05,
+      "loss": 0.646,
+      "step": 7709
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6915324746996526e-05,
+      "loss": 0.6,
+      "step": 7710
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69144773393814e-05,
+      "loss": 0.6564,
+      "step": 7711
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6913629836616996e-05,
+      "loss": 0.6707,
+      "step": 7712
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691278223871497e-05,
+      "loss": 0.6907,
+      "step": 7713
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6911934545686986e-05,
+      "loss": 0.6719,
+      "step": 7714
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691108675754471e-05,
+      "loss": 0.6745,
+      "step": 7715
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6910238874299805e-05,
+      "loss": 0.627,
+      "step": 7716
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690939089596394e-05,
+      "loss": 0.663,
+      "step": 7717
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690854282254879e-05,
+      "loss": 0.6918,
+      "step": 7718
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690769465406602e-05,
+      "loss": 0.5827,
+      "step": 7719
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6906846390527296e-05,
+      "loss": 0.7143,
+      "step": 7720
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69059980319443e-05,
+      "loss": 0.582,
+      "step": 7721
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6905149578328705e-05,
+      "loss": 0.6804,
+      "step": 7722
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6904301029692184e-05,
+      "loss": 0.6684,
+      "step": 7723
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6903452386046413e-05,
+      "loss": 0.597,
+      "step": 7724
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690260364740307e-05,
+      "loss": 0.6164,
+      "step": 7725
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6901754813773838e-05,
+      "loss": 0.6,
+      "step": 7726
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6900905885170392e-05,
+      "loss": 0.6493,
+      "step": 7727
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6900056861604422e-05,
+      "loss": 0.6207,
+      "step": 7728
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6899207743087603e-05,
+      "loss": 0.6549,
+      "step": 7729
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6898358529631625e-05,
+      "loss": 0.6674,
+      "step": 7730
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689750922124817e-05,
+      "loss": 0.7429,
+      "step": 7731
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6896659817948934e-05,
+      "loss": 0.5808,
+      "step": 7732
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6895810319745592e-05,
+      "loss": 0.6302,
+      "step": 7733
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6894960726649844e-05,
+      "loss": 0.6474,
+      "step": 7734
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6894111038673378e-05,
+      "loss": 0.7098,
+      "step": 7735
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6893261255827887e-05,
+      "loss": 0.6353,
+      "step": 7736
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6892411378125063e-05,
+      "loss": 0.5983,
+      "step": 7737
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68915614055766e-05,
+      "loss": 0.6538,
+      "step": 7738
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68907113381942e-05,
+      "loss": 0.7126,
+      "step": 7739
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6889861175989556e-05,
+      "loss": 0.6593,
+      "step": 7740
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6889010918974367e-05,
+      "loss": 0.6539,
+      "step": 7741
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6888160567160335e-05,
+      "loss": 0.6715,
+      "step": 7742
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6887310120559162e-05,
+      "loss": 0.6564,
+      "step": 7743
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688645957918255e-05,
+      "loss": 0.6329,
+      "step": 7744
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6885608943042202e-05,
+      "loss": 0.5818,
+      "step": 7745
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6884758212149827e-05,
+      "loss": 0.6492,
+      "step": 7746
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6883907386517128e-05,
+      "loss": 0.6085,
+      "step": 7747
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6883056466155815e-05,
+      "loss": 0.6555,
+      "step": 7748
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6882205451077597e-05,
+      "loss": 0.6658,
+      "step": 7749
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6881354341294184e-05,
+      "loss": 0.7003,
+      "step": 7750
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688050313681729e-05,
+      "loss": 0.6051,
+      "step": 7751
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6879651837658623e-05,
+      "loss": 0.5797,
+      "step": 7752
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6878800443829904e-05,
+      "loss": 0.6279,
+      "step": 7753
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687794895534285e-05,
+      "loss": 0.6483,
+      "step": 7754
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6877097372209174e-05,
+      "loss": 0.7117,
+      "step": 7755
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6876245694440595e-05,
+      "loss": 0.6862,
+      "step": 7756
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687539392204883e-05,
+      "loss": 0.6756,
+      "step": 7757
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687454205504561e-05,
+      "loss": 0.6142,
+      "step": 7758
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6873690093442645e-05,
+      "loss": 0.6336,
+      "step": 7759
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6872838037251665e-05,
+      "loss": 0.7228,
+      "step": 7760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6871985886484394e-05,
+      "loss": 0.6507,
+      "step": 7761
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6871133641152566e-05,
+      "loss": 0.6549,
+      "step": 7762
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6870281301267893e-05,
+      "loss": 0.6481,
+      "step": 7763
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686942886684212e-05,
+      "loss": 0.5974,
+      "step": 7764
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6868576337886963e-05,
+      "loss": 0.7235,
+      "step": 7765
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686772371441416e-05,
+      "loss": 0.6348,
+      "step": 7766
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6866870996435448e-05,
+      "loss": 0.6712,
+      "step": 7767
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686601818396256e-05,
+      "loss": 0.6524,
+      "step": 7768
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686516527700722e-05,
+      "loss": 0.6842,
+      "step": 7769
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686431227558118e-05,
+      "loss": 0.7055,
+      "step": 7770
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686345917969617e-05,
+      "loss": 0.5746,
+      "step": 7771
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6862605989363933e-05,
+      "loss": 0.702,
+      "step": 7772
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6861752704596206e-05,
+      "loss": 0.6831,
+      "step": 7773
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6860899325404732e-05,
+      "loss": 0.588,
+      "step": 7774
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6860045851801256e-05,
+      "loss": 0.5944,
+      "step": 7775
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685919228379752e-05,
+      "loss": 0.6733,
+      "step": 7776
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685833862140527e-05,
+      "loss": 0.7106,
+      "step": 7777
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685748486463626e-05,
+      "loss": 0.6815,
+      "step": 7778
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6856631013502226e-05,
+      "loss": 0.5194,
+      "step": 7779
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6855777068014925e-05,
+      "loss": 0.6901,
+      "step": 7780
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6854923028186112e-05,
+      "loss": 0.5829,
+      "step": 7781
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6854068894027532e-05,
+      "loss": 0.6428,
+      "step": 7782
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6853214665550942e-05,
+      "loss": 0.6369,
+      "step": 7783
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6852360342768096e-05,
+      "loss": 0.6335,
+      "step": 7784
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685150592569075e-05,
+      "loss": 0.6698,
+      "step": 7785
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6850651414330666e-05,
+      "loss": 0.6615,
+      "step": 7786
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6849796808699595e-05,
+      "loss": 0.7072,
+      "step": 7787
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6848942108809302e-05,
+      "loss": 0.7002,
+      "step": 7788
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6848087314671547e-05,
+      "loss": 0.5713,
+      "step": 7789
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6847232426298097e-05,
+      "loss": 0.7328,
+      "step": 7790
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684637744370071e-05,
+      "loss": 0.6557,
+      "step": 7791
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6845522366891152e-05,
+      "loss": 0.6745,
+      "step": 7792
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684466719588119e-05,
+      "loss": 0.7399,
+      "step": 7793
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6843811930682597e-05,
+      "loss": 0.681,
+      "step": 7794
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684295657130714e-05,
+      "loss": 0.6454,
+      "step": 7795
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6842101117766584e-05,
+      "loss": 0.6675,
+      "step": 7796
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6841245570072703e-05,
+      "loss": 0.556,
+      "step": 7797
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6840389928237276e-05,
+      "loss": 0.6805,
+      "step": 7798
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683953419227207e-05,
+      "loss": 0.5666,
+      "step": 7799
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6838678362188864e-05,
+      "loss": 0.5877,
+      "step": 7800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683782243799944e-05,
+      "loss": 0.6663,
+      "step": 7801
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6836966419715564e-05,
+      "loss": 0.6442,
+      "step": 7802
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6836110307349026e-05,
+      "loss": 0.626,
+      "step": 7803
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68352541009116e-05,
+      "loss": 0.6607,
+      "step": 7804
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6834397800415077e-05,
+      "loss": 0.6702,
+      "step": 7805
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6833541405871235e-05,
+      "loss": 0.672,
+      "step": 7806
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6832684917291855e-05,
+      "loss": 0.7259,
+      "step": 7807
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683182833468873e-05,
+      "loss": 0.6379,
+      "step": 7808
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683097165807364e-05,
+      "loss": 0.6948,
+      "step": 7809
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683011488745838e-05,
+      "loss": 0.6319,
+      "step": 7810
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6829258022854738e-05,
+      "loss": 0.6886,
+      "step": 7811
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6828401064274502e-05,
+      "loss": 0.6533,
+      "step": 7812
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682754401172947e-05,
+      "loss": 0.6314,
+      "step": 7813
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6826686865231434e-05,
+      "loss": 0.687,
+      "step": 7814
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6825829624792187e-05,
+      "loss": 0.6947,
+      "step": 7815
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6824972290423527e-05,
+      "loss": 0.622,
+      "step": 7816
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6824114862137254e-05,
+      "loss": 0.6589,
+      "step": 7817
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682325733994516e-05,
+      "loss": 0.5862,
+      "step": 7818
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6822399723859053e-05,
+      "loss": 0.642,
+      "step": 7819
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6821542013890727e-05,
+      "loss": 0.6275,
+      "step": 7820
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6820684210051993e-05,
+      "loss": 0.7245,
+      "step": 7821
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6819826312354647e-05,
+      "loss": 0.6929,
+      "step": 7822
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6818968320810502e-05,
+      "loss": 0.617,
+      "step": 7823
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6818110235431362e-05,
+      "loss": 0.6517,
+      "step": 7824
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681725205622903e-05,
+      "loss": 0.745,
+      "step": 7825
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6816393783215325e-05,
+      "loss": 0.7258,
+      "step": 7826
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681553541640205e-05,
+      "loss": 0.6531,
+      "step": 7827
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681467695580102e-05,
+      "loss": 0.7181,
+      "step": 7828
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6813818401424042e-05,
+      "loss": 0.645,
+      "step": 7829
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681295975328294e-05,
+      "loss": 0.6515,
+      "step": 7830
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681210101138952e-05,
+      "loss": 0.7214,
+      "step": 7831
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681124217575561e-05,
+      "loss": 0.64,
+      "step": 7832
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6810383246393022e-05,
+      "loss": 0.6811,
+      "step": 7833
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6809524223313577e-05,
+      "loss": 0.7006,
+      "step": 7834
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6808665106529096e-05,
+      "loss": 0.6695,
+      "step": 7835
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6807805896051397e-05,
+      "loss": 0.6987,
+      "step": 7836
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6806946591892308e-05,
+      "loss": 0.6068,
+      "step": 7837
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6806087194063657e-05,
+      "loss": 0.7214,
+      "step": 7838
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6805227702577262e-05,
+      "loss": 0.6038,
+      "step": 7839
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6804368117444954e-05,
+      "loss": 0.7158,
+      "step": 7840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6803508438678564e-05,
+      "loss": 0.6485,
+      "step": 7841
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6802648666289915e-05,
+      "loss": 0.6793,
+      "step": 7842
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6801788800290846e-05,
+      "loss": 0.6191,
+      "step": 7843
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680092884069319e-05,
+      "loss": 0.5914,
+      "step": 7844
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6800068787508775e-05,
+      "loss": 0.6091,
+      "step": 7845
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679920864074944e-05,
+      "loss": 0.6278,
+      "step": 7846
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6798348400427012e-05,
+      "loss": 0.6801,
+      "step": 7847
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6797488066553344e-05,
+      "loss": 0.6243,
+      "step": 7848
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6796627639140266e-05,
+      "loss": 0.5859,
+      "step": 7849
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6795767118199618e-05,
+      "loss": 0.6768,
+      "step": 7850
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6794906503743244e-05,
+      "loss": 0.6162,
+      "step": 7851
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6794045795782987e-05,
+      "loss": 0.7085,
+      "step": 7852
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679318499433069e-05,
+      "loss": 0.6455,
+      "step": 7853
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6792324099398196e-05,
+      "loss": 0.7476,
+      "step": 7854
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6791463110997358e-05,
+      "loss": 0.7322,
+      "step": 7855
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6790602029140014e-05,
+      "loss": 0.6235,
+      "step": 7856
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6789740853838026e-05,
+      "loss": 0.6002,
+      "step": 7857
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678887958510323e-05,
+      "loss": 0.6539,
+      "step": 7858
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6788018222947496e-05,
+      "loss": 0.6574,
+      "step": 7859
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678715676738266e-05,
+      "loss": 0.6141,
+      "step": 7860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6786295218420587e-05,
+      "loss": 0.6161,
+      "step": 7861
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6785433576073125e-05,
+      "loss": 0.6564,
+      "step": 7862
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6784571840352137e-05,
+      "loss": 0.6861,
+      "step": 7863
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678371001126948e-05,
+      "loss": 0.67,
+      "step": 7864
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6782848088837008e-05,
+      "loss": 0.6736,
+      "step": 7865
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6781986073066588e-05,
+      "loss": 0.6262,
+      "step": 7866
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678112396397008e-05,
+      "loss": 0.7308,
+      "step": 7867
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6780261761559353e-05,
+      "loss": 0.6892,
+      "step": 7868
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6779399465846262e-05,
+      "loss": 0.6588,
+      "step": 7869
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677853707684268e-05,
+      "loss": 0.6661,
+      "step": 7870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677767459456047e-05,
+      "loss": 0.6782,
+      "step": 7871
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6776812019011503e-05,
+      "loss": 0.6389,
+      "step": 7872
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6775949350207646e-05,
+      "loss": 0.6057,
+      "step": 7873
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6775086588160773e-05,
+      "loss": 0.6705,
+      "step": 7874
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6774223732882758e-05,
+      "loss": 0.6585,
+      "step": 7875
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6773360784385467e-05,
+      "loss": 0.7039,
+      "step": 7876
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6772497742680783e-05,
+      "loss": 0.6729,
+      "step": 7877
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6771634607780584e-05,
+      "loss": 0.7,
+      "step": 7878
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6770771379696735e-05,
+      "loss": 0.6831,
+      "step": 7879
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6769908058441125e-05,
+      "loss": 0.6131,
+      "step": 7880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676904464402563e-05,
+      "loss": 0.6643,
+      "step": 7881
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676818113646214e-05,
+      "loss": 0.5798,
+      "step": 7882
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6767317535762523e-05,
+      "loss": 0.6254,
+      "step": 7883
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6766453841938672e-05,
+      "loss": 0.6166,
+      "step": 7884
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676559005500247e-05,
+      "loss": 0.5853,
+      "step": 7885
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676472617496581e-05,
+      "loss": 0.718,
+      "step": 7886
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676386220184057e-05,
+      "loss": 0.5661,
+      "step": 7887
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676299813563864e-05,
+      "loss": 0.7196,
+      "step": 7888
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6762133976371914e-05,
+      "loss": 0.6474,
+      "step": 7889
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6761269724052285e-05,
+      "loss": 0.6706,
+      "step": 7890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6760405378691646e-05,
+      "loss": 0.7193,
+      "step": 7891
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675954094030188e-05,
+      "loss": 0.7456,
+      "step": 7892
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67586764088949e-05,
+      "loss": 0.6711,
+      "step": 7893
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675781178448259e-05,
+      "loss": 0.6716,
+      "step": 7894
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6756947067076855e-05,
+      "loss": 0.6604,
+      "step": 7895
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675608225668959e-05,
+      "loss": 0.7398,
+      "step": 7896
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6755217353332694e-05,
+      "loss": 0.7093,
+      "step": 7897
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675435235701807e-05,
+      "loss": 0.7018,
+      "step": 7898
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6753487267757624e-05,
+      "loss": 0.6059,
+      "step": 7899
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675262208556326e-05,
+      "loss": 0.6466,
+      "step": 7900
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6751756810446888e-05,
+      "loss": 0.7122,
+      "step": 7901
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6750891442420404e-05,
+      "loss": 0.5994,
+      "step": 7902
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6750025981495723e-05,
+      "loss": 0.6954,
+      "step": 7903
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6749160427684753e-05,
+      "loss": 0.6597,
+      "step": 7904
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6748294780999404e-05,
+      "loss": 0.6485,
+      "step": 7905
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674742904145159e-05,
+      "loss": 0.7162,
+      "step": 7906
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6746563209053226e-05,
+      "loss": 0.6157,
+      "step": 7907
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6745697283816224e-05,
+      "loss": 0.6288,
+      "step": 7908
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67448312657525e-05,
+      "loss": 0.697,
+      "step": 7909
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6743965154873966e-05,
+      "loss": 0.6605,
+      "step": 7910
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6743098951192555e-05,
+      "loss": 0.6804,
+      "step": 7911
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674223265472017e-05,
+      "loss": 0.6848,
+      "step": 7912
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674136626546874e-05,
+      "loss": 0.6956,
+      "step": 7913
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674049978345019e-05,
+      "loss": 0.5919,
+      "step": 7914
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6739633208676438e-05,
+      "loss": 0.6865,
+      "step": 7915
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673876654115941e-05,
+      "loss": 0.684,
+      "step": 7916
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6737899780911036e-05,
+      "loss": 0.7156,
+      "step": 7917
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673703292794324e-05,
+      "loss": 0.6604,
+      "step": 7918
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6736165982267953e-05,
+      "loss": 0.6459,
+      "step": 7919
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6735298943897097e-05,
+      "loss": 0.5817,
+      "step": 7920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6734431812842613e-05,
+      "loss": 0.6675,
+      "step": 7921
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6733564589116433e-05,
+      "loss": 0.5987,
+      "step": 7922
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6732697272730482e-05,
+      "loss": 0.6237,
+      "step": 7923
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6731829863696705e-05,
+      "loss": 0.6409,
+      "step": 7924
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6730962362027034e-05,
+      "loss": 0.6795,
+      "step": 7925
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6730094767733402e-05,
+      "loss": 0.6197,
+      "step": 7926
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6729227080827756e-05,
+      "loss": 0.6722,
+      "step": 7927
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6728359301322032e-05,
+      "loss": 0.6808,
+      "step": 7928
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6727491429228173e-05,
+      "loss": 0.6694,
+      "step": 7929
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6726623464558116e-05,
+      "loss": 0.6578,
+      "step": 7930
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6725755407323814e-05,
+      "loss": 0.7126,
+      "step": 7931
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6724887257537207e-05,
+      "loss": 0.6813,
+      "step": 7932
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6724019015210245e-05,
+      "loss": 0.6205,
+      "step": 7933
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672315068035487e-05,
+      "loss": 0.5553,
+      "step": 7934
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6722282252983034e-05,
+      "loss": 0.633,
+      "step": 7935
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6721413733106685e-05,
+      "loss": 0.6441,
+      "step": 7936
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6720545120737784e-05,
+      "loss": 0.6039,
+      "step": 7937
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671967641588827e-05,
+      "loss": 0.6667,
+      "step": 7938
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671880761857011e-05,
+      "loss": 0.6596,
+      "step": 7939
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6717938728795247e-05,
+      "loss": 0.616,
+      "step": 7940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671706974657565e-05,
+      "loss": 0.7013,
+      "step": 7941
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671620067192327e-05,
+      "loss": 0.6757,
+      "step": 7942
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6715331504850065e-05,
+      "loss": 0.6907,
+      "step": 7943
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6714462245368e-05,
+      "loss": 0.5886,
+      "step": 7944
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6713592893489032e-05,
+      "loss": 0.6646,
+      "step": 7945
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671272344922513e-05,
+      "loss": 0.6357,
+      "step": 7946
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6711853912588256e-05,
+      "loss": 0.6991,
+      "step": 7947
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671098428359037e-05,
+      "loss": 0.6149,
+      "step": 7948
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6710114562243448e-05,
+      "loss": 0.5868,
+      "step": 7949
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6709244748559448e-05,
+      "loss": 0.6621,
+      "step": 7950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670837484255035e-05,
+      "loss": 0.7095,
+      "step": 7951
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6707504844228116e-05,
+      "loss": 0.6203,
+      "step": 7952
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6706634753604725e-05,
+      "loss": 0.6072,
+      "step": 7953
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6705764570692148e-05,
+      "loss": 0.704,
+      "step": 7954
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6704894295502355e-05,
+      "loss": 0.612,
+      "step": 7955
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6704023928047323e-05,
+      "loss": 0.655,
+      "step": 7956
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6703153468339036e-05,
+      "loss": 0.7152,
+      "step": 7957
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6702282916389464e-05,
+      "loss": 0.6899,
+      "step": 7958
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670141227221059e-05,
+      "loss": 0.6617,
+      "step": 7959
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6700541535814397e-05,
+      "loss": 0.5676,
+      "step": 7960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6699670707212863e-05,
+      "loss": 0.6797,
+      "step": 7961
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6698799786417973e-05,
+      "loss": 0.7392,
+      "step": 7962
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6697928773441713e-05,
+      "loss": 0.6582,
+      "step": 7963
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669705766829607e-05,
+      "loss": 0.6151,
+      "step": 7964
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6696186470993026e-05,
+      "loss": 0.6353,
+      "step": 7965
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6695315181544572e-05,
+      "loss": 0.7151,
+      "step": 7966
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6694443799962698e-05,
+      "loss": 0.6988,
+      "step": 7967
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6693572326259397e-05,
+      "loss": 0.6004,
+      "step": 7968
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6692700760446658e-05,
+      "loss": 0.7392,
+      "step": 7969
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6691829102536477e-05,
+      "loss": 0.5932,
+      "step": 7970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6690957352540848e-05,
+      "loss": 0.6504,
+      "step": 7971
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669008551047176e-05,
+      "loss": 0.5954,
+      "step": 7972
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6689213576341226e-05,
+      "loss": 0.5998,
+      "step": 7973
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668834155016123e-05,
+      "loss": 0.6847,
+      "step": 7974
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668746943194378e-05,
+      "loss": 0.7226,
+      "step": 7975
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6686597221700874e-05,
+      "loss": 0.6868,
+      "step": 7976
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6685724919444514e-05,
+      "loss": 0.6129,
+      "step": 7977
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6684852525186704e-05,
+      "loss": 0.689,
+      "step": 7978
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668398003893945e-05,
+      "loss": 0.629,
+      "step": 7979
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668310746071476e-05,
+      "loss": 0.6259,
+      "step": 7980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668223479052464e-05,
+      "loss": 0.6038,
+      "step": 7981
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668136202838109e-05,
+      "loss": 0.6236,
+      "step": 7982
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6680489174296135e-05,
+      "loss": 0.7203,
+      "step": 7983
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6679616228281775e-05,
+      "loss": 0.5558,
+      "step": 7984
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6678743190350032e-05,
+      "loss": 0.7025,
+      "step": 7985
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6677870060512908e-05,
+      "loss": 0.6993,
+      "step": 7986
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6676996838782425e-05,
+      "loss": 0.6576,
+      "step": 7987
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6676123525170603e-05,
+      "loss": 0.6366,
+      "step": 7988
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667525011968945e-05,
+      "loss": 0.6853,
+      "step": 7989
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6674376622350995e-05,
+      "loss": 0.6014,
+      "step": 7990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6673503033167253e-05,
+      "loss": 0.7443,
+      "step": 7991
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6672629352150237e-05,
+      "loss": 0.6056,
+      "step": 7992
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6671755579311984e-05,
+      "loss": 0.6475,
+      "step": 7993
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6670881714664515e-05,
+      "loss": 0.5786,
+      "step": 7994
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667000775821985e-05,
+      "loss": 0.5737,
+      "step": 7995
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6669133709990016e-05,
+      "loss": 0.6552,
+      "step": 7996
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6668259569987043e-05,
+      "loss": 0.7555,
+      "step": 7997
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6667385338222954e-05,
+      "loss": 0.6165,
+      "step": 7998
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6666511014709788e-05,
+      "loss": 0.678,
+      "step": 7999
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6665636599459576e-05,
+      "loss": 0.5868,
+      "step": 8000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6664762092484346e-05,
+      "loss": 0.5883,
+      "step": 8001
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666388749379613e-05,
+      "loss": 0.5834,
+      "step": 8002
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6663012803406972e-05,
+      "loss": 0.6033,
+      "step": 8003
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66621380213289e-05,
+      "loss": 0.5844,
+      "step": 8004
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6661263147573954e-05,
+      "loss": 0.7203,
+      "step": 8005
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6660388182154176e-05,
+      "loss": 0.6867,
+      "step": 8006
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66595131250816e-05,
+      "loss": 0.6755,
+      "step": 8007
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6658637976368276e-05,
+      "loss": 0.5806,
+      "step": 8008
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6657762736026245e-05,
+      "loss": 0.6773,
+      "step": 8009
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6656887404067544e-05,
+      "loss": 0.5577,
+      "step": 8010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6656011980504226e-05,
+      "loss": 0.6874,
+      "step": 8011
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6655136465348332e-05,
+      "loss": 0.641,
+      "step": 8012
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6654260858611913e-05,
+      "loss": 0.7025,
+      "step": 8013
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665338516030702e-05,
+      "loss": 0.6855,
+      "step": 8014
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66525093704457e-05,
+      "loss": 0.6693,
+      "step": 8015
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6651633489040003e-05,
+      "loss": 0.6581,
+      "step": 8016
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6650757516101988e-05,
+      "loss": 0.7246,
+      "step": 8017
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6649881451643706e-05,
+      "loss": 0.6321,
+      "step": 8018
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6649005295677206e-05,
+      "loss": 0.6819,
+      "step": 8019
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664812904821456e-05,
+      "loss": 0.6876,
+      "step": 8020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664725270926781e-05,
+      "loss": 0.674,
+      "step": 8021
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6646376278849024e-05,
+      "loss": 0.6454,
+      "step": 8022
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664549975697026e-05,
+      "loss": 0.5992,
+      "step": 8023
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664462314364358e-05,
+      "loss": 0.6613,
+      "step": 8024
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664374643888105e-05,
+      "loss": 0.6207,
+      "step": 8025
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664286964269473e-05,
+      "loss": 0.7067,
+      "step": 8026
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6641992755096685e-05,
+      "loss": 0.5945,
+      "step": 8027
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6641115776098983e-05,
+      "loss": 0.6241,
+      "step": 8028
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6640238705713693e-05,
+      "loss": 0.6726,
+      "step": 8029
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6639361543952885e-05,
+      "loss": 0.6374,
+      "step": 8030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663848429082863e-05,
+      "loss": 0.6206,
+      "step": 8031
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6637606946352996e-05,
+      "loss": 0.5542,
+      "step": 8032
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6636729510538056e-05,
+      "loss": 0.7534,
+      "step": 8033
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6635851983395887e-05,
+      "loss": 0.5982,
+      "step": 8034
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6634974364938566e-05,
+      "loss": 0.6975,
+      "step": 8035
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6634096655178168e-05,
+      "loss": 0.6005,
+      "step": 8036
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6633218854126768e-05,
+      "loss": 0.62,
+      "step": 8037
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6632340961796455e-05,
+      "loss": 0.6035,
+      "step": 8038
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6631462978199297e-05,
+      "loss": 0.6632,
+      "step": 8039
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663058490334738e-05,
+      "loss": 0.6019,
+      "step": 8040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6629706737252792e-05,
+      "loss": 0.6205,
+      "step": 8041
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6628828479927613e-05,
+      "loss": 0.6739,
+      "step": 8042
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6627950131383928e-05,
+      "loss": 0.7236,
+      "step": 8043
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6627071691633826e-05,
+      "loss": 0.6674,
+      "step": 8044
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6626193160689395e-05,
+      "loss": 0.6933,
+      "step": 8045
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6625314538562725e-05,
+      "loss": 0.5849,
+      "step": 8046
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6624435825265903e-05,
+      "loss": 0.6891,
+      "step": 8047
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6623557020811024e-05,
+      "loss": 0.6145,
+      "step": 8048
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662267812521018e-05,
+      "loss": 0.6724,
+      "step": 8049
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6621799138475468e-05,
+      "loss": 0.7328,
+      "step": 8050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662092006061898e-05,
+      "loss": 0.6749,
+      "step": 8051
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6620040891652816e-05,
+      "loss": 0.7297,
+      "step": 8052
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6619161631589072e-05,
+      "loss": 0.678,
+      "step": 8053
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661828228043985e-05,
+      "loss": 0.6592,
+      "step": 8054
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6617402838217242e-05,
+      "loss": 0.6492,
+      "step": 8055
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661652330493336e-05,
+      "loss": 0.6881,
+      "step": 8056
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6615643680600306e-05,
+      "loss": 0.6091,
+      "step": 8057
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6614763965230175e-05,
+      "loss": 0.5764,
+      "step": 8058
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6613884158835086e-05,
+      "loss": 0.6298,
+      "step": 8059
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6613004261427137e-05,
+      "loss": 0.6757,
+      "step": 8060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661212427301844e-05,
+      "loss": 0.6452,
+      "step": 8061
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66112441936211e-05,
+      "loss": 0.6164,
+      "step": 8062
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6610364023247234e-05,
+      "loss": 0.6104,
+      "step": 8063
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660948376190895e-05,
+      "loss": 0.6595,
+      "step": 8064
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660860340961836e-05,
+      "loss": 0.6195,
+      "step": 8065
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6607722966387584e-05,
+      "loss": 0.6569,
+      "step": 8066
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660684243222873e-05,
+      "loss": 0.6466,
+      "step": 8067
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660596180715392e-05,
+      "loss": 0.7124,
+      "step": 8068
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6605081091175273e-05,
+      "loss": 0.6719,
+      "step": 8069
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6604200284304904e-05,
+      "loss": 0.6083,
+      "step": 8070
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6603319386554937e-05,
+      "loss": 0.6827,
+      "step": 8071
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660243839793749e-05,
+      "loss": 0.5756,
+      "step": 8072
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6601557318464696e-05,
+      "loss": 0.6227,
+      "step": 8073
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6600676148148667e-05,
+      "loss": 0.6492,
+      "step": 8074
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6599794887001535e-05,
+      "loss": 0.6336,
+      "step": 8075
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659891353503543e-05,
+      "loss": 0.5673,
+      "step": 8076
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659803209226247e-05,
+      "loss": 0.6731,
+      "step": 8077
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6597150558694798e-05,
+      "loss": 0.6896,
+      "step": 8078
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6596268934344536e-05,
+      "loss": 0.7341,
+      "step": 8079
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6595387219223814e-05,
+      "loss": 0.648,
+      "step": 8080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6594505413344772e-05,
+      "loss": 0.673,
+      "step": 8081
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659362351671954e-05,
+      "loss": 0.6556,
+      "step": 8082
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6592741529360254e-05,
+      "loss": 0.5834,
+      "step": 8083
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6591859451279053e-05,
+      "loss": 0.6621,
+      "step": 8084
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659097728248807e-05,
+      "loss": 0.6099,
+      "step": 8085
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6590095022999454e-05,
+      "loss": 0.7814,
+      "step": 8086
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6589212672825334e-05,
+      "loss": 0.6785,
+      "step": 8087
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6588330231977862e-05,
+      "loss": 0.6395,
+      "step": 8088
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6587447700469178e-05,
+      "loss": 0.6105,
+      "step": 8089
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6586565078311424e-05,
+      "loss": 0.673,
+      "step": 8090
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6585682365516743e-05,
+      "loss": 0.7008,
+      "step": 8091
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6584799562097293e-05,
+      "loss": 0.6509,
+      "step": 8092
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658391666806521e-05,
+      "loss": 0.5806,
+      "step": 8093
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658303368343265e-05,
+      "loss": 0.7276,
+      "step": 8094
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658215060821176e-05,
+      "loss": 0.6644,
+      "step": 8095
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6581267442414697e-05,
+      "loss": 0.6387,
+      "step": 8096
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658038418605361e-05,
+      "loss": 0.6082,
+      "step": 8097
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6579500839140656e-05,
+      "loss": 0.5977,
+      "step": 8098
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6578617401687988e-05,
+      "loss": 0.6311,
+      "step": 8099
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6577733873707762e-05,
+      "loss": 0.7057,
+      "step": 8100
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6576850255212142e-05,
+      "loss": 0.6231,
+      "step": 8101
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657596654621328e-05,
+      "loss": 0.589,
+      "step": 8102
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6575082746723342e-05,
+      "loss": 0.6694,
+      "step": 8103
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657419885675449e-05,
+      "loss": 0.6997,
+      "step": 8104
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6573314876318885e-05,
+      "loss": 0.6804,
+      "step": 8105
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6572430805428687e-05,
+      "loss": 0.69,
+      "step": 8106
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657154664409607e-05,
+      "loss": 0.6843,
+      "step": 8107
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6570662392333196e-05,
+      "loss": 0.6817,
+      "step": 8108
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6569778050152233e-05,
+      "loss": 0.682,
+      "step": 8109
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6568893617565353e-05,
+      "loss": 0.5896,
+      "step": 8110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6568009094584723e-05,
+      "loss": 0.5986,
+      "step": 8111
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6567124481222516e-05,
+      "loss": 0.6761,
+      "step": 8112
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656623977749091e-05,
+      "loss": 0.5872,
+      "step": 8113
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6565354983402074e-05,
+      "loss": 0.6881,
+      "step": 8114
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656447009896818e-05,
+      "loss": 0.6517,
+      "step": 8115
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6563585124201414e-05,
+      "loss": 0.6808,
+      "step": 8116
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6562700059113948e-05,
+      "loss": 0.6669,
+      "step": 8117
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656181490371796e-05,
+      "loss": 0.7274,
+      "step": 8118
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6560929658025635e-05,
+      "loss": 0.5894,
+      "step": 8119
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6560044322049154e-05,
+      "loss": 0.7037,
+      "step": 8120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65591588958007e-05,
+      "loss": 0.5695,
+      "step": 8121
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6558273379292456e-05,
+      "loss": 0.6656,
+      "step": 8122
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6557387772536606e-05,
+      "loss": 0.5679,
+      "step": 8123
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655650207554534e-05,
+      "loss": 0.6822,
+      "step": 8124
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6555616288330844e-05,
+      "loss": 0.7139,
+      "step": 8125
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6554730410905303e-05,
+      "loss": 0.703,
+      "step": 8126
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655384444328092e-05,
+      "loss": 0.6931,
+      "step": 8127
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6552958385469872e-05,
+      "loss": 0.5837,
+      "step": 8128
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6552072237484363e-05,
+      "loss": 0.6966,
+      "step": 8129
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655118599933658e-05,
+      "loss": 0.5832,
+      "step": 8130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6550299671038722e-05,
+      "loss": 0.6746,
+      "step": 8131
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654941325260299e-05,
+      "loss": 0.7025,
+      "step": 8132
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654852674404157e-05,
+      "loss": 0.5568,
+      "step": 8133
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654764014536667e-05,
+      "loss": 0.6554,
+      "step": 8134
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6546753456590488e-05,
+      "loss": 0.6927,
+      "step": 8135
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6545866677725227e-05,
+      "loss": 0.653,
+      "step": 8136
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6544979808783092e-05,
+      "loss": 0.6028,
+      "step": 8137
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6544092849776278e-05,
+      "loss": 0.6259,
+      "step": 8138
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6543205800716997e-05,
+      "loss": 0.6796,
+      "step": 8139
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6542318661617457e-05,
+      "loss": 0.7226,
+      "step": 8140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654143143248986e-05,
+      "loss": 0.6353,
+      "step": 8141
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6540544113346424e-05,
+      "loss": 0.661,
+      "step": 8142
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6539656704199353e-05,
+      "loss": 0.6243,
+      "step": 8143
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6538769205060856e-05,
+      "loss": 0.5988,
+      "step": 8144
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6537881615943153e-05,
+      "loss": 0.6803,
+      "step": 8145
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6536993936858455e-05,
+      "loss": 0.6519,
+      "step": 8146
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6536106167818974e-05,
+      "loss": 0.5869,
+      "step": 8147
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6535218308836927e-05,
+      "loss": 0.5722,
+      "step": 8148
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6534330359924537e-05,
+      "loss": 0.6207,
+      "step": 8149
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6533442321094017e-05,
+      "loss": 0.7007,
+      "step": 8150
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6532554192357593e-05,
+      "loss": 0.7038,
+      "step": 8151
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653166597372748e-05,
+      "loss": 0.7101,
+      "step": 8152
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6530777665215907e-05,
+      "loss": 0.6355,
+      "step": 8153
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6529889266835093e-05,
+      "loss": 0.703,
+      "step": 8154
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6529000778597266e-05,
+      "loss": 0.6392,
+      "step": 8155
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6528112200514652e-05,
+      "loss": 0.592,
+      "step": 8156
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652722353259948e-05,
+      "loss": 0.6656,
+      "step": 8157
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6526334774863973e-05,
+      "loss": 0.6624,
+      "step": 8158
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6525445927320367e-05,
+      "loss": 0.6683,
+      "step": 8159
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6524556989980888e-05,
+      "loss": 0.7081,
+      "step": 8160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6523667962857776e-05,
+      "loss": 0.6532,
+      "step": 8161
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652277884596326e-05,
+      "loss": 0.7296,
+      "step": 8162
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652188963930957e-05,
+      "loss": 0.7555,
+      "step": 8163
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6521000342908956e-05,
+      "loss": 0.726,
+      "step": 8164
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6520110956773643e-05,
+      "loss": 0.6405,
+      "step": 8165
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6519221480915875e-05,
+      "loss": 0.622,
+      "step": 8166
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6518331915347888e-05,
+      "loss": 0.5784,
+      "step": 8167
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651744226008193e-05,
+      "loss": 0.644,
+      "step": 8168
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651655251513024e-05,
+      "loss": 0.6769,
+      "step": 8169
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651566268050506e-05,
+      "loss": 0.6654,
+      "step": 8170
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651477275621864e-05,
+      "loss": 0.6677,
+      "step": 8171
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6513882742283217e-05,
+      "loss": 0.6058,
+      "step": 8172
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6512992638711046e-05,
+      "loss": 0.751,
+      "step": 8173
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6512102445514376e-05,
+      "loss": 0.6388,
+      "step": 8174
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651121216270545e-05,
+      "loss": 0.7026,
+      "step": 8175
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6510321790296527e-05,
+      "loss": 0.5986,
+      "step": 8176
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6509431328299853e-05,
+      "loss": 0.6775,
+      "step": 8177
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6508540776727687e-05,
+      "loss": 0.6299,
+      "step": 8178
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650765013559228e-05,
+      "loss": 0.6121,
+      "step": 8179
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6506759404905886e-05,
+      "loss": 0.6352,
+      "step": 8180
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6505868584680768e-05,
+      "loss": 0.7334,
+      "step": 8181
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6504977674929184e-05,
+      "loss": 0.7045,
+      "step": 8182
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6504086675663388e-05,
+      "loss": 0.6015,
+      "step": 8183
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6503195586895646e-05,
+      "loss": 0.5709,
+      "step": 8184
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6502304408638215e-05,
+      "loss": 0.613,
+      "step": 8185
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6501413140903365e-05,
+      "loss": 0.6725,
+      "step": 8186
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650052178370336e-05,
+      "loss": 0.6807,
+      "step": 8187
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649963033705046e-05,
+      "loss": 0.674,
+      "step": 8188
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6498738800956936e-05,
+      "loss": 0.6344,
+      "step": 8189
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6497847175435057e-05,
+      "loss": 0.71,
+      "step": 8190
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6496955460497087e-05,
+      "loss": 0.6796,
+      "step": 8191
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6496063656155304e-05,
+      "loss": 0.739,
+      "step": 8192
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649517176242198e-05,
+      "loss": 0.682,
+      "step": 8193
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6494279779309385e-05,
+      "loss": 0.6204,
+      "step": 8194
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649338770682979e-05,
+      "loss": 0.6754,
+      "step": 8195
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6492495544995474e-05,
+      "loss": 0.6343,
+      "step": 8196
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649160329381872e-05,
+      "loss": 0.6872,
+      "step": 8197
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6490710953311796e-05,
+      "loss": 0.6386,
+      "step": 8198
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6489818523486987e-05,
+      "loss": 0.6133,
+      "step": 8199
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6488926004356573e-05,
+      "loss": 0.7038,
+      "step": 8200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6488033395932838e-05,
+      "loss": 0.6931,
+      "step": 8201
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648714069822806e-05,
+      "loss": 0.6549,
+      "step": 8202
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6486247911254525e-05,
+      "loss": 0.6297,
+      "step": 8203
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648535503502452e-05,
+      "loss": 0.6944,
+      "step": 8204
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648446206955033e-05,
+      "loss": 0.6353,
+      "step": 8205
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6483569014844247e-05,
+      "loss": 0.6475,
+      "step": 8206
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6482675870918557e-05,
+      "loss": 0.6897,
+      "step": 8207
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648178263778555e-05,
+      "loss": 0.6381,
+      "step": 8208
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648088931545752e-05,
+      "loss": 0.6483,
+      "step": 8209
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6479995903946756e-05,
+      "loss": 0.7232,
+      "step": 8210
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6479102403265557e-05,
+      "loss": 0.7385,
+      "step": 8211
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6478208813426217e-05,
+      "loss": 0.6697,
+      "step": 8212
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6477315134441025e-05,
+      "loss": 0.6285,
+      "step": 8213
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647642136632229e-05,
+      "loss": 0.691,
+      "step": 8214
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6475527509082304e-05,
+      "loss": 0.6562,
+      "step": 8215
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6474633562733372e-05,
+      "loss": 0.652,
+      "step": 8216
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6473739527287792e-05,
+      "loss": 0.7028,
+      "step": 8217
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647284540275787e-05,
+      "loss": 0.62,
+      "step": 8218
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6471951189155905e-05,
+      "loss": 0.6393,
+      "step": 8219
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6471056886494205e-05,
+      "loss": 0.679,
+      "step": 8220
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6470162494785077e-05,
+      "loss": 0.6807,
+      "step": 8221
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646926801404083e-05,
+      "loss": 0.6889,
+      "step": 8222
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6468373444273767e-05,
+      "loss": 0.6495,
+      "step": 8223
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64674787854962e-05,
+      "loss": 0.6689,
+      "step": 8224
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646658403772045e-05,
+      "loss": 0.6078,
+      "step": 8225
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6465689200958818e-05,
+      "loss": 0.7095,
+      "step": 8226
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646479427522362e-05,
+      "loss": 0.6455,
+      "step": 8227
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646389926052717e-05,
+      "loss": 0.647,
+      "step": 8228
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646300415688179e-05,
+      "loss": 0.6412,
+      "step": 8229
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646210896429979e-05,
+      "loss": 0.5975,
+      "step": 8230
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64612136827935e-05,
+      "loss": 0.6239,
+      "step": 8231
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646031831237523e-05,
+      "loss": 0.7137,
+      "step": 8232
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64594228530573e-05,
+      "loss": 0.6155,
+      "step": 8233
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6458527304852036e-05,
+      "loss": 0.6638,
+      "step": 8234
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6457631667771764e-05,
+      "loss": 0.636,
+      "step": 8235
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6456735941828802e-05,
+      "loss": 0.644,
+      "step": 8236
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6455840127035483e-05,
+      "loss": 0.6852,
+      "step": 8237
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645494422340413e-05,
+      "loss": 0.6695,
+      "step": 8238
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6454048230947074e-05,
+      "loss": 0.7099,
+      "step": 8239
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6453152149676643e-05,
+      "loss": 0.6847,
+      "step": 8240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645225597960517e-05,
+      "loss": 0.6466,
+      "step": 8241
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6451359720744984e-05,
+      "loss": 0.6803,
+      "step": 8242
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6450463373108418e-05,
+      "loss": 0.6612,
+      "step": 8243
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644956693670781e-05,
+      "loss": 0.6955,
+      "step": 8244
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6448670411555493e-05,
+      "loss": 0.6554,
+      "step": 8245
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6447773797663806e-05,
+      "loss": 0.718,
+      "step": 8246
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6446877095045083e-05,
+      "loss": 0.5938,
+      "step": 8247
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644598030371167e-05,
+      "loss": 0.5838,
+      "step": 8248
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6445083423675902e-05,
+      "loss": 0.6908,
+      "step": 8249
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6444186454950123e-05,
+      "loss": 0.6042,
+      "step": 8250
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6443289397546676e-05,
+      "loss": 0.6088,
+      "step": 8251
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6442392251477904e-05,
+      "loss": 0.5843,
+      "step": 8252
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6441495016756156e-05,
+      "loss": 0.5891,
+      "step": 8253
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6440597693393776e-05,
+      "loss": 0.6829,
+      "step": 8254
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6439700281403113e-05,
+      "loss": 0.6939,
+      "step": 8255
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6438802780796517e-05,
+      "loss": 0.6657,
+      "step": 8256
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6437905191586338e-05,
+      "loss": 0.6794,
+      "step": 8257
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6437007513784918e-05,
+      "loss": 0.6735,
+      "step": 8258
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643610974740463e-05,
+      "loss": 0.656,
+      "step": 8259
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643521189245781e-05,
+      "loss": 0.6418,
+      "step": 8260
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643431394895682e-05,
+      "loss": 0.6088,
+      "step": 8261
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6433415916914018e-05,
+      "loss": 0.7177,
+      "step": 8262
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643251779634176e-05,
+      "loss": 0.714,
+      "step": 8263
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6431619587252407e-05,
+      "loss": 0.6757,
+      "step": 8264
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6430721289658315e-05,
+      "loss": 0.5981,
+      "step": 8265
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6429822903571848e-05,
+      "loss": 0.6827,
+      "step": 8266
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6428924429005366e-05,
+      "loss": 0.6407,
+      "step": 8267
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6428025865971236e-05,
+      "loss": 0.6258,
+      "step": 8268
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6427127214481825e-05,
+      "loss": 0.6296,
+      "step": 8269
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6426228474549492e-05,
+      "loss": 0.6293,
+      "step": 8270
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6425329646186612e-05,
+      "loss": 0.6576,
+      "step": 8271
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6424430729405546e-05,
+      "loss": 0.6952,
+      "step": 8272
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642353172421867e-05,
+      "loss": 0.6621,
+      "step": 8273
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6422632630638355e-05,
+      "loss": 0.5646,
+      "step": 8274
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642173344867697e-05,
+      "loss": 0.6505,
+      "step": 8275
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642083417834689e-05,
+      "loss": 0.6829,
+      "step": 8276
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6419934819660493e-05,
+      "loss": 0.6414,
+      "step": 8277
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6419035372630148e-05,
+      "loss": 0.7169,
+      "step": 8278
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641813583726824e-05,
+      "loss": 0.5994,
+      "step": 8279
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6417236213587138e-05,
+      "loss": 0.6602,
+      "step": 8280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641633650159923e-05,
+      "loss": 0.6971,
+      "step": 8281
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6415436701316894e-05,
+      "loss": 0.6406,
+      "step": 8282
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6414536812752513e-05,
+      "loss": 0.5678,
+      "step": 8283
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6413636835918466e-05,
+      "loss": 0.633,
+      "step": 8284
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641273677082714e-05,
+      "loss": 0.6538,
+      "step": 8285
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6411836617490925e-05,
+      "loss": 0.6861,
+      "step": 8286
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64109363759222e-05,
+      "loss": 0.5952,
+      "step": 8287
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6410036046133364e-05,
+      "loss": 0.6886,
+      "step": 8288
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6409135628136795e-05,
+      "loss": 0.7144,
+      "step": 8289
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640823512194489e-05,
+      "loss": 0.6617,
+      "step": 8290
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6407334527570037e-05,
+      "loss": 0.6715,
+      "step": 8291
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6406433845024632e-05,
+      "loss": 0.6682,
+      "step": 8292
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6405533074321067e-05,
+      "loss": 0.6986,
+      "step": 8293
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640463221547174e-05,
+      "loss": 0.6263,
+      "step": 8294
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6403731268489046e-05,
+      "loss": 0.6692,
+      "step": 8295
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640283023338538e-05,
+      "loss": 0.666,
+      "step": 8296
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640192911017315e-05,
+      "loss": 0.674,
+      "step": 8297
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6401027898864745e-05,
+      "loss": 0.6344,
+      "step": 8298
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6400126599472576e-05,
+      "loss": 0.6548,
+      "step": 8299
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639922521200904e-05,
+      "loss": 0.6413,
+      "step": 8300
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6398323736486536e-05,
+      "loss": 0.6221,
+      "step": 8301
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6397422172917484e-05,
+      "loss": 0.657,
+      "step": 8302
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6396520521314277e-05,
+      "loss": 0.6881,
+      "step": 8303
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6395618781689328e-05,
+      "loss": 0.6059,
+      "step": 8304
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6394716954055043e-05,
+      "loss": 0.5798,
+      "step": 8305
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6393815038423837e-05,
+      "loss": 0.6599,
+      "step": 8306
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6392913034808117e-05,
+      "loss": 0.6566,
+      "step": 8307
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6392010943220293e-05,
+      "loss": 0.6591,
+      "step": 8308
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6391108763672787e-05,
+      "loss": 0.7277,
+      "step": 8309
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6390206496178007e-05,
+      "loss": 0.6939,
+      "step": 8310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.638930414074837e-05,
+      "loss": 0.6096,
+      "step": 8311
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6388401697396293e-05,
+      "loss": 0.5886,
+      "step": 8312
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63874991661342e-05,
+      "loss": 0.6424,
+      "step": 8313
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63865965469745e-05,
+      "loss": 0.6231,
+      "step": 8314
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6385693839929623e-05,
+      "loss": 0.6115,
+      "step": 8315
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6384791045011986e-05,
+      "loss": 0.6909,
+      "step": 8316
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6383888162234014e-05,
+      "loss": 0.6174,
+      "step": 8317
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6382985191608134e-05,
+      "loss": 0.6407,
+      "step": 8318
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6382082133146767e-05,
+      "loss": 0.7379,
+      "step": 8319
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6381178986862344e-05,
+      "loss": 0.6146,
+      "step": 8320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.638027575276729e-05,
+      "loss": 0.7163,
+      "step": 8321
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6379372430874037e-05,
+      "loss": 0.6274,
+      "step": 8322
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637846902119501e-05,
+      "loss": 0.6876,
+      "step": 8323
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6377565523742648e-05,
+      "loss": 0.5908,
+      "step": 8324
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637666193852938e-05,
+      "loss": 0.6281,
+      "step": 8325
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637575826556764e-05,
+      "loss": 0.578,
+      "step": 8326
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6374854504869864e-05,
+      "loss": 0.7266,
+      "step": 8327
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637395065644849e-05,
+      "loss": 0.6175,
+      "step": 8328
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6373046720315954e-05,
+      "loss": 0.7063,
+      "step": 8329
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6372142696484693e-05,
+      "loss": 0.6803,
+      "step": 8330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637123858496715e-05,
+      "loss": 0.6645,
+      "step": 8331
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6370334385775766e-05,
+      "loss": 0.6926,
+      "step": 8332
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6369430098922987e-05,
+      "loss": 0.6108,
+      "step": 8333
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6368525724421248e-05,
+      "loss": 0.6152,
+      "step": 8334
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6367621262283003e-05,
+      "loss": 0.6715,
+      "step": 8335
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636671671252069e-05,
+      "loss": 0.7135,
+      "step": 8336
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6365812075146765e-05,
+      "loss": 0.6261,
+      "step": 8337
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6364907350173672e-05,
+      "loss": 0.6604,
+      "step": 8338
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6364002537613858e-05,
+      "loss": 0.6697,
+      "step": 8339
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6363097637479782e-05,
+      "loss": 0.647,
+      "step": 8340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636219264978389e-05,
+      "loss": 0.5953,
+      "step": 8341
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6361287574538633e-05,
+      "loss": 0.6113,
+      "step": 8342
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6360382411756472e-05,
+      "loss": 0.6781,
+      "step": 8343
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635947716144986e-05,
+      "loss": 0.6843,
+      "step": 8344
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6358571823631257e-05,
+      "loss": 0.5961,
+      "step": 8345
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6357666398313115e-05,
+      "loss": 0.6293,
+      "step": 8346
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6356760885507897e-05,
+      "loss": 0.5855,
+      "step": 8347
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6355855285228063e-05,
+      "loss": 0.6965,
+      "step": 8348
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635494959748608e-05,
+      "loss": 0.6308,
+      "step": 8349
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6354043822294402e-05,
+      "loss": 0.7216,
+      "step": 8350
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6353137959665502e-05,
+      "loss": 0.6934,
+      "step": 8351
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635223200961184e-05,
+      "loss": 0.7244,
+      "step": 8352
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6351325972145882e-05,
+      "loss": 0.6762,
+      "step": 8353
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63504198472801e-05,
+      "loss": 0.6837,
+      "step": 8354
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6349513635026955e-05,
+      "loss": 0.6894,
+      "step": 8355
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6348607335398932e-05,
+      "loss": 0.6098,
+      "step": 8356
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6347700948408486e-05,
+      "loss": 0.6236,
+      "step": 8357
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6346794474068102e-05,
+      "loss": 0.6898,
+      "step": 8358
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634588791239025e-05,
+      "loss": 0.6398,
+      "step": 8359
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6344981263387404e-05,
+      "loss": 0.6023,
+      "step": 8360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634407452707204e-05,
+      "loss": 0.624,
+      "step": 8361
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6343167703456634e-05,
+      "loss": 0.6062,
+      "step": 8362
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6342260792553667e-05,
+      "loss": 0.6998,
+      "step": 8363
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634135379437562e-05,
+      "loss": 0.6882,
+      "step": 8364
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6340446708934975e-05,
+      "loss": 0.5734,
+      "step": 8365
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6339539536244207e-05,
+      "loss": 0.6294,
+      "step": 8366
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633863227631581e-05,
+      "loss": 0.7257,
+      "step": 8367
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633772492916226e-05,
+      "loss": 0.6943,
+      "step": 8368
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633681749479605e-05,
+      "loss": 0.7021,
+      "step": 8369
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633590997322966e-05,
+      "loss": 0.6448,
+      "step": 8370
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633500236447558e-05,
+      "loss": 0.6086,
+      "step": 8371
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6334094668546304e-05,
+      "loss": 0.5998,
+      "step": 8372
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6333186885454317e-05,
+      "loss": 0.6565,
+      "step": 8373
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6332279015212114e-05,
+      "loss": 0.5923,
+      "step": 8374
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633137105783219e-05,
+      "loss": 0.5719,
+      "step": 8375
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6330463013327034e-05,
+      "loss": 0.6252,
+      "step": 8376
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632955488170915e-05,
+      "loss": 0.7298,
+      "step": 8377
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6328646662991025e-05,
+      "loss": 0.7583,
+      "step": 8378
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632773835718516e-05,
+      "loss": 0.6911,
+      "step": 8379
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632682996430406e-05,
+      "loss": 0.6776,
+      "step": 8380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6325921484360215e-05,
+      "loss": 0.6872,
+      "step": 8381
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6325012917366135e-05,
+      "loss": 0.6765,
+      "step": 8382
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632410426333432e-05,
+      "loss": 0.669,
+      "step": 8383
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6323195522277272e-05,
+      "loss": 0.7153,
+      "step": 8384
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6322286694207495e-05,
+      "loss": 0.5994,
+      "step": 8385
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6321377779137503e-05,
+      "loss": 0.5881,
+      "step": 8386
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6320468777079796e-05,
+      "loss": 0.653,
+      "step": 8387
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6319559688046886e-05,
+      "loss": 0.5295,
+      "step": 8388
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6318650512051278e-05,
+      "loss": 0.6374,
+      "step": 8389
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631774124910549e-05,
+      "loss": 0.6753,
+      "step": 8390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6316831899222033e-05,
+      "loss": 0.6837,
+      "step": 8391
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6315922462413417e-05,
+      "loss": 0.6217,
+      "step": 8392
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6315012938692153e-05,
+      "loss": 0.6617,
+      "step": 8393
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631410332807077e-05,
+      "loss": 0.6852,
+      "step": 8394
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631319363056177e-05,
+      "loss": 0.7026,
+      "step": 8395
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6312283846177684e-05,
+      "loss": 0.581,
+      "step": 8396
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6311373974931026e-05,
+      "loss": 0.7302,
+      "step": 8397
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6310464016834316e-05,
+      "loss": 0.6267,
+      "step": 8398
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6309553971900073e-05,
+      "loss": 0.5969,
+      "step": 8399
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630864384014083e-05,
+      "loss": 0.6288,
+      "step": 8400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6307733621569094e-05,
+      "loss": 0.5981,
+      "step": 8401
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630682331619741e-05,
+      "loss": 0.6815,
+      "step": 8402
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630591292403829e-05,
+      "loss": 0.6392,
+      "step": 8403
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630500244510427e-05,
+      "loss": 0.6654,
+      "step": 8404
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6304091879407875e-05,
+      "loss": 0.6996,
+      "step": 8405
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6303181226961636e-05,
+      "loss": 0.6495,
+      "step": 8406
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6302270487778088e-05,
+      "loss": 0.5953,
+      "step": 8407
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630135966186976e-05,
+      "loss": 0.6776,
+      "step": 8408
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6300448749249182e-05,
+      "loss": 0.7011,
+      "step": 8409
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6299537749928897e-05,
+      "loss": 0.7112,
+      "step": 8410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6298626663921434e-05,
+      "loss": 0.6111,
+      "step": 8411
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629771549123934e-05,
+      "loss": 0.6897,
+      "step": 8412
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629680423189514e-05,
+      "loss": 0.6812,
+      "step": 8413
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6295892885901384e-05,
+      "loss": 0.7258,
+      "step": 8414
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629498145327061e-05,
+      "loss": 0.6551,
+      "step": 8415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629406993401536e-05,
+      "loss": 0.6651,
+      "step": 8416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6293158328148177e-05,
+      "loss": 0.6098,
+      "step": 8417
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6292246635681607e-05,
+      "loss": 0.5535,
+      "step": 8418
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6291334856628195e-05,
+      "loss": 0.5953,
+      "step": 8419
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6290422991000484e-05,
+      "loss": 0.7268,
+      "step": 8420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6289511038811027e-05,
+      "loss": 0.6873,
+      "step": 8421
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6288599000072373e-05,
+      "loss": 0.6951,
+      "step": 8422
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6287686874797066e-05,
+      "loss": 0.6614,
+      "step": 8423
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6286774662997665e-05,
+      "loss": 0.6743,
+      "step": 8424
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628586236468672e-05,
+      "loss": 0.6195,
+      "step": 8425
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628494997987679e-05,
+      "loss": 0.6691,
+      "step": 8426
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6284037508580425e-05,
+      "loss": 0.6285,
+      "step": 8427
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628312495081018e-05,
+      "loss": 0.6173,
+      "step": 8428
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6282212306578616e-05,
+      "loss": 0.5943,
+      "step": 8429
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628129957589829e-05,
+      "loss": 0.6255,
+      "step": 8430
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6280386758781763e-05,
+      "loss": 0.6024,
+      "step": 8431
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6279473855241596e-05,
+      "loss": 0.6928,
+      "step": 8432
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627856086529035e-05,
+      "loss": 0.6463,
+      "step": 8433
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627764778894059e-05,
+      "loss": 0.692,
+      "step": 8434
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627673462620488e-05,
+      "loss": 0.5793,
+      "step": 8435
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627582137709579e-05,
+      "loss": 0.6018,
+      "step": 8436
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6274908041625877e-05,
+      "loss": 0.6953,
+      "step": 8437
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627399461980772e-05,
+      "loss": 0.7278,
+      "step": 8438
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6273081111653884e-05,
+      "loss": 0.6257,
+      "step": 8439
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6272167517176938e-05,
+      "loss": 0.6295,
+      "step": 8440
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627125383638946e-05,
+      "loss": 0.6789,
+      "step": 8441
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6270340069304014e-05,
+      "loss": 0.6935,
+      "step": 8442
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626942621593318e-05,
+      "loss": 0.687,
+      "step": 8443
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6268512276289535e-05,
+      "loss": 0.6856,
+      "step": 8444
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626759825038565e-05,
+      "loss": 0.6097,
+      "step": 8445
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6266684138234108e-05,
+      "loss": 0.5897,
+      "step": 8446
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6265769939847485e-05,
+      "loss": 0.6449,
+      "step": 8447
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626485565523836e-05,
+      "loss": 0.6372,
+      "step": 8448
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6263941284419322e-05,
+      "loss": 0.6497,
+      "step": 8449
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6263026827402943e-05,
+      "loss": 0.5732,
+      "step": 8450
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6262112284201815e-05,
+      "loss": 0.6897,
+      "step": 8451
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6261197654828515e-05,
+      "loss": 0.6472,
+      "step": 8452
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6260282939295637e-05,
+      "loss": 0.714,
+      "step": 8453
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6259368137615767e-05,
+      "loss": 0.7121,
+      "step": 8454
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6258453249801487e-05,
+      "loss": 0.6277,
+      "step": 8455
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6257538275865396e-05,
+      "loss": 0.6407,
+      "step": 8456
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6256623215820074e-05,
+      "loss": 0.6957,
+      "step": 8457
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6255708069678122e-05,
+      "loss": 0.5993,
+      "step": 8458
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625479283745213e-05,
+      "loss": 0.6454,
+      "step": 8459
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6253877519154693e-05,
+      "loss": 0.6353,
+      "step": 8460
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6252962114798407e-05,
+      "loss": 0.7412,
+      "step": 8461
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625204662439587e-05,
+      "loss": 0.7002,
+      "step": 8462
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6251131047959675e-05,
+      "loss": 0.6116,
+      "step": 8463
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625021538550242e-05,
+      "loss": 0.6908,
+      "step": 8464
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6249299637036716e-05,
+      "loss": 0.7269,
+      "step": 8465
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6248383802575154e-05,
+      "loss": 0.6957,
+      "step": 8466
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624746788213034e-05,
+      "loss": 0.6457,
+      "step": 8467
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6246551875714883e-05,
+      "loss": 0.5917,
+      "step": 8468
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624563578334138e-05,
+      "loss": 0.6993,
+      "step": 8469
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624471960502244e-05,
+      "loss": 0.6009,
+      "step": 8470
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6243803340770676e-05,
+      "loss": 0.5504,
+      "step": 8471
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6242886990598688e-05,
+      "loss": 0.5769,
+      "step": 8472
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624197055451909e-05,
+      "loss": 0.6903,
+      "step": 8473
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6241054032544493e-05,
+      "loss": 0.6251,
+      "step": 8474
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624013742468751e-05,
+      "loss": 0.637,
+      "step": 8475
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623922073096075e-05,
+      "loss": 0.6557,
+      "step": 8476
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623830395137683e-05,
+      "loss": 0.672,
+      "step": 8477
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6237387085948365e-05,
+      "loss": 0.6833,
+      "step": 8478
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6236470134687977e-05,
+      "loss": 0.6834,
+      "step": 8479
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6235553097608278e-05,
+      "loss": 0.6194,
+      "step": 8480
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623463597472189e-05,
+      "loss": 0.6465,
+      "step": 8481
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6233718766041435e-05,
+      "loss": 0.6188,
+      "step": 8482
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623280147157953e-05,
+      "loss": 0.6442,
+      "step": 8483
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6231884091348803e-05,
+      "loss": 0.6413,
+      "step": 8484
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623096662536187e-05,
+      "loss": 0.6368,
+      "step": 8485
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6230049073631364e-05,
+      "loss": 0.657,
+      "step": 8486
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622913143616991e-05,
+      "loss": 0.7309,
+      "step": 8487
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622821371299013e-05,
+      "loss": 0.6018,
+      "step": 8488
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622729590410466e-05,
+      "loss": 0.6125,
+      "step": 8489
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6226378009526126e-05,
+      "loss": 0.6753,
+      "step": 8490
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622546002926716e-05,
+      "loss": 0.6935,
+      "step": 8491
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6224541963340392e-05,
+      "loss": 0.6169,
+      "step": 8492
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6223623811758458e-05,
+      "loss": 0.6054,
+      "step": 8493
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622270557453399e-05,
+      "loss": 0.6438,
+      "step": 8494
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622178725167963e-05,
+      "loss": 0.6059,
+      "step": 8495
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6220868843208006e-05,
+      "loss": 0.695,
+      "step": 8496
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6219950349131765e-05,
+      "loss": 0.7146,
+      "step": 8497
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6219031769463537e-05,
+      "loss": 0.577,
+      "step": 8498
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621811310421597e-05,
+      "loss": 0.6519,
+      "step": 8499
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6217194353401703e-05,
+      "loss": 0.6979,
+      "step": 8500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621627551703338e-05,
+      "loss": 0.6527,
+      "step": 8501
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621535659512364e-05,
+      "loss": 0.5981,
+      "step": 8502
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6214437587685136e-05,
+      "loss": 0.6338,
+      "step": 8503
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621351849473051e-05,
+      "loss": 0.6771,
+      "step": 8504
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6212599316272408e-05,
+      "loss": 0.6427,
+      "step": 8505
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6211680052323483e-05,
+      "loss": 0.6226,
+      "step": 8506
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621076070289638e-05,
+      "loss": 0.6916,
+      "step": 8507
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6209841268003753e-05,
+      "loss": 0.6062,
+      "step": 8508
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6208921747658257e-05,
+      "loss": 0.5723,
+      "step": 8509
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6208002141872533e-05,
+      "loss": 0.7058,
+      "step": 8510
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620708245065925e-05,
+      "loss": 0.5863,
+      "step": 8511
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6206162674031062e-05,
+      "loss": 0.7449,
+      "step": 8512
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620524281200062e-05,
+      "loss": 0.6387,
+      "step": 8513
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6204322864580583e-05,
+      "loss": 0.652,
+      "step": 8514
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6203402831783613e-05,
+      "loss": 0.7164,
+      "step": 8515
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6202482713622368e-05,
+      "loss": 0.56,
+      "step": 8516
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6201562510109513e-05,
+      "loss": 0.682,
+      "step": 8517
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6200642221257705e-05,
+      "loss": 0.7315,
+      "step": 8518
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6199721847079616e-05,
+      "loss": 0.6899,
+      "step": 8519
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6198801387587905e-05,
+      "loss": 0.6518,
+      "step": 8520
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6197880842795243e-05,
+      "loss": 0.6075,
+      "step": 8521
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619696021271429e-05,
+      "loss": 0.709,
+      "step": 8522
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619603949735772e-05,
+      "loss": 0.5844,
+      "step": 8523
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6195118696738202e-05,
+      "loss": 0.6189,
+      "step": 8524
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6194197810868408e-05,
+      "loss": 0.5873,
+      "step": 8525
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619327683976101e-05,
+      "loss": 0.6346,
+      "step": 8526
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6192355783428682e-05,
+      "loss": 0.7094,
+      "step": 8527
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6191434641884097e-05,
+      "loss": 0.6027,
+      "step": 8528
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619051341513993e-05,
+      "loss": 0.6228,
+      "step": 8529
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618959210320886e-05,
+      "loss": 0.6757,
+      "step": 8530
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6188670706103562e-05,
+      "loss": 0.6366,
+      "step": 8531
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6187749223836718e-05,
+      "loss": 0.6189,
+      "step": 8532
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6186827656421006e-05,
+      "loss": 0.6225,
+      "step": 8533
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6185906003869114e-05,
+      "loss": 0.7288,
+      "step": 8534
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618498426619372e-05,
+      "loss": 0.6693,
+      "step": 8535
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6184062443407503e-05,
+      "loss": 0.6279,
+      "step": 8536
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6183140535523154e-05,
+      "loss": 0.6038,
+      "step": 8537
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618221854255336e-05,
+      "loss": 0.6402,
+      "step": 8538
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6181296464510804e-05,
+      "loss": 0.62,
+      "step": 8539
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618037430140818e-05,
+      "loss": 0.5809,
+      "step": 8540
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6179452053258174e-05,
+      "loss": 0.6408,
+      "step": 8541
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6178529720073483e-05,
+      "loss": 0.7149,
+      "step": 8542
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6177607301866786e-05,
+      "loss": 0.6979,
+      "step": 8543
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6176684798650788e-05,
+      "loss": 0.6809,
+      "step": 8544
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617576221043818e-05,
+      "loss": 0.599,
+      "step": 8545
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6174839537241658e-05,
+      "loss": 0.6166,
+      "step": 8546
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6173916779073917e-05,
+      "loss": 0.6553,
+      "step": 8547
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617299393594766e-05,
+      "loss": 0.6668,
+      "step": 8548
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6172071007875575e-05,
+      "loss": 0.7271,
+      "step": 8549
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6171147994870375e-05,
+      "loss": 0.6036,
+      "step": 8550
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6170224896944755e-05,
+      "loss": 0.6702,
+      "step": 8551
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6169301714111418e-05,
+      "loss": 0.5761,
+      "step": 8552
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616837844638307e-05,
+      "loss": 0.6706,
+      "step": 8553
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6167455093772413e-05,
+      "loss": 0.6617,
+      "step": 8554
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6166531656292156e-05,
+      "loss": 0.711,
+      "step": 8555
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6165608133955006e-05,
+      "loss": 0.6139,
+      "step": 8556
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616468452677367e-05,
+      "loss": 0.6861,
+      "step": 8557
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6163760834760856e-05,
+      "loss": 0.6871,
+      "step": 8558
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616283705792928e-05,
+      "loss": 0.6755,
+      "step": 8559
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6161913196291646e-05,
+      "loss": 0.5978,
+      "step": 8560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6160989249860675e-05,
+      "loss": 0.7336,
+      "step": 8561
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6160065218649083e-05,
+      "loss": 0.6492,
+      "step": 8562
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6159141102669576e-05,
+      "loss": 0.6453,
+      "step": 8563
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6158216901934873e-05,
+      "loss": 0.6892,
+      "step": 8564
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6157292616457697e-05,
+      "loss": 0.7277,
+      "step": 8565
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6156368246250766e-05,
+      "loss": 0.6927,
+      "step": 8566
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6155443791326798e-05,
+      "loss": 0.6148,
+      "step": 8567
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615451925169851e-05,
+      "loss": 0.7234,
+      "step": 8568
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6153594627378636e-05,
+      "loss": 0.6298,
+      "step": 8569
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615266991837989e-05,
+      "loss": 0.5827,
+      "step": 8570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6151745124715003e-05,
+      "loss": 0.694,
+      "step": 8571
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6150820246396693e-05,
+      "loss": 0.6941,
+      "step": 8572
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6149895283437694e-05,
+      "loss": 0.6244,
+      "step": 8573
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6148970235850732e-05,
+      "loss": 0.7056,
+      "step": 8574
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6148045103648537e-05,
+      "loss": 0.6295,
+      "step": 8575
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6147119886843837e-05,
+      "loss": 0.6476,
+      "step": 8576
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6146194585449368e-05,
+      "loss": 0.6345,
+      "step": 8577
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6145269199477856e-05,
+      "loss": 0.6301,
+      "step": 8578
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6144343728942048e-05,
+      "loss": 0.6559,
+      "step": 8579
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614341817385467e-05,
+      "loss": 0.6653,
+      "step": 8580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6142492534228455e-05,
+      "loss": 0.6201,
+      "step": 8581
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614156681007615e-05,
+      "loss": 0.6579,
+      "step": 8582
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6140641001410487e-05,
+      "loss": 0.6589,
+      "step": 8583
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613971510824421e-05,
+      "loss": 0.6312,
+      "step": 8584
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6138789130590056e-05,
+      "loss": 0.6591,
+      "step": 8585
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613786306846077e-05,
+      "loss": 0.6396,
+      "step": 8586
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6136936921869097e-05,
+      "loss": 0.6714,
+      "step": 8587
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6136010690827776e-05,
+      "loss": 0.7076,
+      "step": 8588
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613508437534956e-05,
+      "loss": 0.6118,
+      "step": 8589
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6134157975447192e-05,
+      "loss": 0.6128,
+      "step": 8590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6133231491133418e-05,
+      "loss": 0.6863,
+      "step": 8591
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613230492242099e-05,
+      "loss": 0.6439,
+      "step": 8592
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6131378269322656e-05,
+      "loss": 0.7031,
+      "step": 8593
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613045153185117e-05,
+      "loss": 0.5955,
+      "step": 8594
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612952471001929e-05,
+      "loss": 0.7069,
+      "step": 8595
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612859780383976e-05,
+      "loss": 0.641,
+      "step": 8596
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612767081332534e-05,
+      "loss": 0.6272,
+      "step": 8597
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6126743738488782e-05,
+      "loss": 0.6438,
+      "step": 8598
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6125816579342853e-05,
+      "loss": 0.6331,
+      "step": 8599
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6124889335900302e-05,
+      "loss": 0.5808,
+      "step": 8600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612396200817389e-05,
+      "loss": 0.6278,
+      "step": 8601
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6123034596176382e-05,
+      "loss": 0.6251,
+      "step": 8602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6122107099920535e-05,
+      "loss": 0.642,
+      "step": 8603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612117951941912e-05,
+      "loss": 0.607,
+      "step": 8604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6120251854684892e-05,
+      "loss": 0.652,
+      "step": 8605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6119324105730622e-05,
+      "loss": 0.6643,
+      "step": 8606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6118396272569077e-05,
+      "loss": 0.6252,
+      "step": 8607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6117468355213023e-05,
+      "loss": 0.6877,
+      "step": 8608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6116540353675226e-05,
+      "loss": 0.6428,
+      "step": 8609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6115612267968463e-05,
+      "loss": 0.6879,
+      "step": 8610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61146840981055e-05,
+      "loss": 0.5687,
+      "step": 8611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611375584409911e-05,
+      "loss": 0.6371,
+      "step": 8612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6112827505962066e-05,
+      "loss": 0.6274,
+      "step": 8613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6111899083707148e-05,
+      "loss": 0.6834,
+      "step": 8614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6110970577347127e-05,
+      "loss": 0.6233,
+      "step": 8615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6110041986894782e-05,
+      "loss": 0.6362,
+      "step": 8616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6109113312362888e-05,
+      "loss": 0.6988,
+      "step": 8617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610818455376423e-05,
+      "loss": 0.6692,
+      "step": 8618
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6107255711111584e-05,
+      "loss": 0.7014,
+      "step": 8619
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6106326784417734e-05,
+      "loss": 0.6174,
+      "step": 8620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6105397773695458e-05,
+      "loss": 0.6082,
+      "step": 8621
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610446867895755e-05,
+      "loss": 0.6051,
+      "step": 8622
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6103539500216783e-05,
+      "loss": 0.6996,
+      "step": 8623
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6102610237485956e-05,
+      "loss": 0.7075,
+      "step": 8624
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6101680890777846e-05,
+      "loss": 0.6263,
+      "step": 8625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6100751460105244e-05,
+      "loss": 0.6857,
+      "step": 8626
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6099821945480946e-05,
+      "loss": 0.6662,
+      "step": 8627
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6098892346917737e-05,
+      "loss": 0.6395,
+      "step": 8628
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6097962664428405e-05,
+      "loss": 0.619,
+      "step": 8629
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6097032898025756e-05,
+      "loss": 0.5942,
+      "step": 8630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6096103047722573e-05,
+      "loss": 0.623,
+      "step": 8631
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6095173113531655e-05,
+      "loss": 0.6659,
+      "step": 8632
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60942430954658e-05,
+      "loss": 0.589,
+      "step": 8633
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6093312993537806e-05,
+      "loss": 0.6514,
+      "step": 8634
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609238280776047e-05,
+      "loss": 0.6636,
+      "step": 8635
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6091452538146593e-05,
+      "loss": 0.6594,
+      "step": 8636
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609052218470898e-05,
+      "loss": 0.6524,
+      "step": 8637
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6089591747460428e-05,
+      "loss": 0.5926,
+      "step": 8638
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608866122641374e-05,
+      "loss": 0.6599,
+      "step": 8639
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608773062158173e-05,
+      "loss": 0.5818,
+      "step": 8640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6086799932977193e-05,
+      "loss": 0.6646,
+      "step": 8641
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608586916061294e-05,
+      "loss": 0.7236,
+      "step": 8642
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608493830450178e-05,
+      "loss": 0.5673,
+      "step": 8643
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6084007364656523e-05,
+      "loss": 0.6621,
+      "step": 8644
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608307634108998e-05,
+      "loss": 0.6834,
+      "step": 8645
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608214523381496e-05,
+      "loss": 0.708,
+      "step": 8646
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608121404284428e-05,
+      "loss": 0.5995,
+      "step": 8647
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6080282768190748e-05,
+      "loss": 0.704,
+      "step": 8648
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6079351409867185e-05,
+      "loss": 0.7241,
+      "step": 8649
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6078419967886402e-05,
+      "loss": 0.6477,
+      "step": 8650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6077488442261218e-05,
+      "loss": 0.6046,
+      "step": 8651
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607655683300446e-05,
+      "loss": 0.7482,
+      "step": 8652
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6075625140128934e-05,
+      "loss": 0.6123,
+      "step": 8653
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607469336364747e-05,
+      "loss": 0.6395,
+      "step": 8654
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607376150357289e-05,
+      "loss": 0.66,
+      "step": 8655
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607282955991801e-05,
+      "loss": 0.5857,
+      "step": 8656
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6071897532695662e-05,
+      "loss": 0.6872,
+      "step": 8657
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6070965421918668e-05,
+      "loss": 0.6486,
+      "step": 8658
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6070033227599858e-05,
+      "loss": 0.5846,
+      "step": 8659
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6069100949752052e-05,
+      "loss": 0.5783,
+      "step": 8660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6068168588388088e-05,
+      "loss": 0.6178,
+      "step": 8661
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606723614352079e-05,
+      "loss": 0.6902,
+      "step": 8662
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6066303615162994e-05,
+      "loss": 0.559,
+      "step": 8663
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606537100332753e-05,
+      "loss": 0.7291,
+      "step": 8664
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6064438308027227e-05,
+      "loss": 0.5757,
+      "step": 8665
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606350552927493e-05,
+      "loss": 0.67,
+      "step": 8666
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606257266708347e-05,
+      "loss": 0.7019,
+      "step": 8667
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6061639721465675e-05,
+      "loss": 0.6385,
+      "step": 8668
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6060706692434398e-05,
+      "loss": 0.575,
+      "step": 8669
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6059773580002468e-05,
+      "loss": 0.6367,
+      "step": 8670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6058840384182732e-05,
+      "loss": 0.6389,
+      "step": 8671
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6057907104988026e-05,
+      "loss": 0.7081,
+      "step": 8672
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6056973742431194e-05,
+      "loss": 0.6241,
+      "step": 8673
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6056040296525086e-05,
+      "loss": 0.7006,
+      "step": 8674
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6055106767282537e-05,
+      "loss": 0.5728,
+      "step": 8675
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60541731547164e-05,
+      "loss": 0.7142,
+      "step": 8676
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6053239458839524e-05,
+      "loss": 0.6867,
+      "step": 8677
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6052305679664752e-05,
+      "loss": 0.7379,
+      "step": 8678
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6051371817204934e-05,
+      "loss": 0.6111,
+      "step": 8679
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6050437871472927e-05,
+      "loss": 0.6692,
+      "step": 8680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6049503842481576e-05,
+      "loss": 0.6737,
+      "step": 8681
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6048569730243735e-05,
+      "loss": 0.6608,
+      "step": 8682
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604763553477226e-05,
+      "loss": 0.632,
+      "step": 8683
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6046701256080008e-05,
+      "loss": 0.6375,
+      "step": 8684
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6045766894179836e-05,
+      "loss": 0.6618,
+      "step": 8685
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6044832449084594e-05,
+      "loss": 0.6088,
+      "step": 8686
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6043897920807148e-05,
+      "loss": 0.6315,
+      "step": 8687
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6042963309360357e-05,
+      "loss": 0.5804,
+      "step": 8688
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604202861475708e-05,
+      "loss": 0.7132,
+      "step": 8689
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604109383701018e-05,
+      "loss": 0.5951,
+      "step": 8690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604015897613252e-05,
+      "loss": 0.683,
+      "step": 8691
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6039224032136965e-05,
+      "loss": 0.6681,
+      "step": 8692
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6038289005036385e-05,
+      "loss": 0.647,
+      "step": 8693
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603735389484364e-05,
+      "loss": 0.6761,
+      "step": 8694
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60364187015716e-05,
+      "loss": 0.6471,
+      "step": 8695
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6035483425233134e-05,
+      "loss": 0.6699,
+      "step": 8696
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6034548065841113e-05,
+      "loss": 0.6445,
+      "step": 8697
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.603361262340841e-05,
+      "loss": 0.6144,
+      "step": 8698
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6032677097947895e-05,
+      "loss": 0.6808,
+      "step": 8699
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6031741489472442e-05,
+      "loss": 0.7457,
+      "step": 8700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6030805797994923e-05,
+      "loss": 0.7401,
+      "step": 8701
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6029870023528222e-05,
+      "loss": 0.6634,
+      "step": 8702
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602893416608521e-05,
+      "loss": 0.6652,
+      "step": 8703
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6027998225678763e-05,
+      "loss": 0.6525,
+      "step": 8704
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6027062202321765e-05,
+      "loss": 0.5874,
+      "step": 8705
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.60261260960271e-05,
+      "loss": 0.6392,
+      "step": 8706
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6025189906807642e-05,
+      "loss": 0.6738,
+      "step": 8707
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6024253634676275e-05,
+      "loss": 0.6736,
+      "step": 8708
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6023317279645887e-05,
+      "loss": 0.6558,
+      "step": 8709
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602238084172936e-05,
+      "loss": 0.6086,
+      "step": 8710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602144432093958e-05,
+      "loss": 0.6959,
+      "step": 8711
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6020507717289434e-05,
+      "loss": 0.6169,
+      "step": 8712
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6019571030791815e-05,
+      "loss": 0.6485,
+      "step": 8713
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601863426145961e-05,
+      "loss": 0.7298,
+      "step": 8714
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6017697409305703e-05,
+      "loss": 0.6589,
+      "step": 8715
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6016760474342997e-05,
+      "loss": 0.6286,
+      "step": 8716
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601582345658438e-05,
+      "loss": 0.5139,
+      "step": 8717
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6014886356042744e-05,
+      "loss": 0.5928,
+      "step": 8718
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6013949172730987e-05,
+      "loss": 0.6034,
+      "step": 8719
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6013011906662008e-05,
+      "loss": 0.6468,
+      "step": 8720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6012074557848697e-05,
+      "loss": 0.7013,
+      "step": 8721
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601113712630396e-05,
+      "loss": 0.6462,
+      "step": 8722
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6010199612040693e-05,
+      "loss": 0.6853,
+      "step": 8723
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.60092620150718e-05,
+      "loss": 0.6316,
+      "step": 8724
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6008324335410183e-05,
+      "loss": 0.6092,
+      "step": 8725
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6007386573068743e-05,
+      "loss": 0.5834,
+      "step": 8726
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600644872806038e-05,
+      "loss": 0.6589,
+      "step": 8727
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600551080039801e-05,
+      "loss": 0.5708,
+      "step": 8728
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6004572790094535e-05,
+      "loss": 0.5892,
+      "step": 8729
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600363469716286e-05,
+      "loss": 0.6451,
+      "step": 8730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6002696521615897e-05,
+      "loss": 0.7229,
+      "step": 8731
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6001758263466554e-05,
+      "loss": 0.6671,
+      "step": 8732
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6000819922727745e-05,
+      "loss": 0.6252,
+      "step": 8733
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599988149941238e-05,
+      "loss": 0.5785,
+      "step": 8734
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599894299353338e-05,
+      "loss": 0.7154,
+      "step": 8735
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5998004405103644e-05,
+      "loss": 0.7152,
+      "step": 8736
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5997065734136104e-05,
+      "loss": 0.6871,
+      "step": 8737
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5996126980643665e-05,
+      "loss": 0.684,
+      "step": 8738
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599518814463925e-05,
+      "loss": 0.6714,
+      "step": 8739
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599424922613578e-05,
+      "loss": 0.6874,
+      "step": 8740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5993310225146178e-05,
+      "loss": 0.7199,
+      "step": 8741
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5992371141683357e-05,
+      "loss": 0.6216,
+      "step": 8742
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5991431975760243e-05,
+      "loss": 0.6448,
+      "step": 8743
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599049272738976e-05,
+      "loss": 0.6662,
+      "step": 8744
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5989553396584836e-05,
+      "loss": 0.6341,
+      "step": 8745
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5988613983358394e-05,
+      "loss": 0.6635,
+      "step": 8746
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5987674487723358e-05,
+      "loss": 0.6735,
+      "step": 8747
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5986734909692664e-05,
+      "loss": 0.6396,
+      "step": 8748
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5985795249279237e-05,
+      "loss": 0.6729,
+      "step": 8749
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5984855506496006e-05,
+      "loss": 0.674,
+      "step": 8750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5983915681355907e-05,
+      "loss": 0.6492,
+      "step": 8751
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5982975773871866e-05,
+      "loss": 0.6347,
+      "step": 8752
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5982035784056823e-05,
+      "loss": 0.6527,
+      "step": 8753
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5981095711923713e-05,
+      "loss": 0.6757,
+      "step": 8754
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598015555748547e-05,
+      "loss": 0.5965,
+      "step": 8755
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597921532075503e-05,
+      "loss": 0.6276,
+      "step": 8756
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5978275001745336e-05,
+      "loss": 0.7265,
+      "step": 8757
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5977334600469324e-05,
+      "loss": 0.6756,
+      "step": 8758
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5976394116939934e-05,
+      "loss": 0.7252,
+      "step": 8759
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5975453551170113e-05,
+      "loss": 0.5925,
+      "step": 8760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5974512903172797e-05,
+      "loss": 0.6167,
+      "step": 8761
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5973572172960935e-05,
+      "loss": 0.7355,
+      "step": 8762
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5972631360547473e-05,
+      "loss": 0.7086,
+      "step": 8763
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5971690465945354e-05,
+      "loss": 0.6176,
+      "step": 8764
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5970749489167528e-05,
+      "loss": 0.6448,
+      "step": 8765
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596980843022694e-05,
+      "loss": 0.6115,
+      "step": 8766
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5968867289136545e-05,
+      "loss": 0.6958,
+      "step": 8767
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5967926065909288e-05,
+      "loss": 0.6742,
+      "step": 8768
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596698476055813e-05,
+      "loss": 0.6065,
+      "step": 8769
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5966043373096018e-05,
+      "loss": 0.6065,
+      "step": 8770
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5965101903535903e-05,
+      "loss": 0.6243,
+      "step": 8771
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5964160351890743e-05,
+      "loss": 0.7094,
+      "step": 8772
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5963218718173502e-05,
+      "loss": 0.6223,
+      "step": 8773
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596227700239713e-05,
+      "loss": 0.6416,
+      "step": 8774
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5961335204574586e-05,
+      "loss": 0.6685,
+      "step": 8775
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596039332471883e-05,
+      "loss": 0.6245,
+      "step": 8776
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5959451362842827e-05,
+      "loss": 0.7206,
+      "step": 8777
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5958509318959537e-05,
+      "loss": 0.7066,
+      "step": 8778
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5957567193081924e-05,
+      "loss": 0.6035,
+      "step": 8779
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595662498522295e-05,
+      "loss": 0.6404,
+      "step": 8780
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595568269539558e-05,
+      "loss": 0.6301,
+      "step": 8781
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5954740323612784e-05,
+      "loss": 0.6293,
+      "step": 8782
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595379786988753e-05,
+      "loss": 0.6865,
+      "step": 8783
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5952855334232785e-05,
+      "loss": 0.6758,
+      "step": 8784
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595191271666152e-05,
+      "loss": 0.683,
+      "step": 8785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595097001718671e-05,
+      "loss": 0.6079,
+      "step": 8786
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595002723582132e-05,
+      "loss": 0.6161,
+      "step": 8787
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594908437257833e-05,
+      "loss": 0.6237,
+      "step": 8788
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5948141427470708e-05,
+      "loss": 0.6997,
+      "step": 8789
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5947198400511438e-05,
+      "loss": 0.6899,
+      "step": 8790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594625529171349e-05,
+      "loss": 0.6426,
+      "step": 8791
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594531210108984e-05,
+      "loss": 0.6256,
+      "step": 8792
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5944368828653476e-05,
+      "loss": 0.6828,
+      "step": 8793
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5943425474417376e-05,
+      "loss": 0.6709,
+      "step": 8794
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5942482038394515e-05,
+      "loss": 0.6978,
+      "step": 8795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594153852059788e-05,
+      "loss": 0.6019,
+      "step": 8796
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5940594921040457e-05,
+      "loss": 0.645,
+      "step": 8797
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5939651239735225e-05,
+      "loss": 0.6894,
+      "step": 8798
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5938707476695173e-05,
+      "loss": 0.6535,
+      "step": 8799
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5937763631933288e-05,
+      "loss": 0.6275,
+      "step": 8800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593681970546256e-05,
+      "loss": 0.6646,
+      "step": 8801
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5935875697295973e-05,
+      "loss": 0.6224,
+      "step": 8802
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5934931607446524e-05,
+      "loss": 0.6154,
+      "step": 8803
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5933987435927202e-05,
+      "loss": 0.6501,
+      "step": 8804
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5933043182750996e-05,
+      "loss": 0.6503,
+      "step": 8805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5932098847930905e-05,
+      "loss": 0.5932,
+      "step": 8806
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5931154431479917e-05,
+      "loss": 0.6898,
+      "step": 8807
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5930209933411036e-05,
+      "loss": 0.578,
+      "step": 8808
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5929265353737253e-05,
+      "loss": 0.5765,
+      "step": 8809
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5928320692471572e-05,
+      "loss": 0.6643,
+      "step": 8810
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5927375949626988e-05,
+      "loss": 0.6902,
+      "step": 8811
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5926431125216503e-05,
+      "loss": 0.6142,
+      "step": 8812
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5925486219253117e-05,
+      "loss": 0.7456,
+      "step": 8813
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5924541231749836e-05,
+      "loss": 0.6086,
+      "step": 8814
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592359616271966e-05,
+      "loss": 0.5856,
+      "step": 8815
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5922651012175596e-05,
+      "loss": 0.6109,
+      "step": 8816
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592170578013065e-05,
+      "loss": 0.6421,
+      "step": 8817
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5920760466597832e-05,
+      "loss": 0.674,
+      "step": 8818
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5919815071590145e-05,
+      "loss": 0.6929,
+      "step": 8819
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59188695951206e-05,
+      "loss": 0.6849,
+      "step": 8820
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591792403720221e-05,
+      "loss": 0.6294,
+      "step": 8821
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5916978397847987e-05,
+      "loss": 0.57,
+      "step": 8822
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591603267707094e-05,
+      "loss": 0.6744,
+      "step": 8823
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5915086874884086e-05,
+      "loss": 0.6917,
+      "step": 8824
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5914140991300436e-05,
+      "loss": 0.5777,
+      "step": 8825
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5913195026333016e-05,
+      "loss": 0.6666,
+      "step": 8826
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591224897999483e-05,
+      "loss": 0.7432,
+      "step": 8827
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5911302852298907e-05,
+      "loss": 0.6432,
+      "step": 8828
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5910356643258263e-05,
+      "loss": 0.6576,
+      "step": 8829
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5909410352885914e-05,
+      "loss": 0.5997,
+      "step": 8830
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590846398119489e-05,
+      "loss": 0.6373,
+      "step": 8831
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590751752819821e-05,
+      "loss": 0.6816,
+      "step": 8832
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5906570993908896e-05,
+      "loss": 0.6423,
+      "step": 8833
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5905624378339976e-05,
+      "loss": 0.6898,
+      "step": 8834
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5904677681504475e-05,
+      "loss": 0.643,
+      "step": 8835
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590373090341542e-05,
+      "loss": 0.6162,
+      "step": 8836
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590278404408584e-05,
+      "loss": 0.6843,
+      "step": 8837
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5901837103528767e-05,
+      "loss": 0.6709,
+      "step": 8838
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590089008175723e-05,
+      "loss": 0.604,
+      "step": 8839
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589994297878426e-05,
+      "loss": 0.6746,
+      "step": 8840
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589899579462289e-05,
+      "loss": 0.6136,
+      "step": 8841
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5898048529286153e-05,
+      "loss": 0.6977,
+      "step": 8842
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5897101182787088e-05,
+      "loss": 0.6903,
+      "step": 8843
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589615375513873e-05,
+      "loss": 0.5599,
+      "step": 8844
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5895206246354115e-05,
+      "loss": 0.6844,
+      "step": 8845
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589425865644628e-05,
+      "loss": 0.6961,
+      "step": 8846
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5893310985428264e-05,
+      "loss": 0.6197,
+      "step": 8847
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5892363233313117e-05,
+      "loss": 0.6733,
+      "step": 8848
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589141540011387e-05,
+      "loss": 0.6296,
+      "step": 8849
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5890467485843573e-05,
+      "loss": 0.6015,
+      "step": 8850
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5889519490515265e-05,
+      "loss": 0.6999,
+      "step": 8851
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5888571414141997e-05,
+      "loss": 0.6214,
+      "step": 8852
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588762325673681e-05,
+      "loss": 0.6327,
+      "step": 8853
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5886675018312754e-05,
+      "loss": 0.6947,
+      "step": 8854
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5885726698882877e-05,
+      "loss": 0.5667,
+      "step": 8855
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5884778298460232e-05,
+      "loss": 0.6239,
+      "step": 8856
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5883829817057862e-05,
+      "loss": 0.6489,
+      "step": 8857
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588288125468883e-05,
+      "loss": 0.695,
+      "step": 8858
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5881932611366178e-05,
+      "loss": 0.614,
+      "step": 8859
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5880983887102965e-05,
+      "loss": 0.6408,
+      "step": 8860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5880035081912246e-05,
+      "loss": 0.715,
+      "step": 8861
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587908619580708e-05,
+      "loss": 0.6346,
+      "step": 8862
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587813722880052e-05,
+      "loss": 0.5807,
+      "step": 8863
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5877188180905626e-05,
+      "loss": 0.6324,
+      "step": 8864
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5876239052135463e-05,
+      "loss": 0.656,
+      "step": 8865
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5875289842503083e-05,
+      "loss": 0.7126,
+      "step": 8866
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5874340552021554e-05,
+      "loss": 0.6781,
+      "step": 8867
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5873391180703937e-05,
+      "loss": 0.6397,
+      "step": 8868
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5872441728563293e-05,
+      "loss": 0.7309,
+      "step": 8869
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5871492195612698e-05,
+      "loss": 0.6796,
+      "step": 8870
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5870542581865205e-05,
+      "loss": 0.706,
+      "step": 8871
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586959288733389e-05,
+      "loss": 0.659,
+      "step": 8872
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5868643112031824e-05,
+      "loss": 0.6559,
+      "step": 8873
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5867693255972064e-05,
+      "loss": 0.67,
+      "step": 8874
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5866743319167696e-05,
+      "loss": 0.622,
+      "step": 8875
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586579330163178e-05,
+      "loss": 0.6048,
+      "step": 8876
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5864843203377397e-05,
+      "loss": 0.6164,
+      "step": 8877
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5863893024417615e-05,
+      "loss": 0.6073,
+      "step": 8878
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5862942764765518e-05,
+      "loss": 0.6457,
+      "step": 8879
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5861992424434173e-05,
+      "loss": 0.6921,
+      "step": 8880
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586104200343666e-05,
+      "loss": 0.6518,
+      "step": 8881
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586009150178606e-05,
+      "loss": 0.5604,
+      "step": 8882
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5859140919495452e-05,
+      "loss": 0.6015,
+      "step": 8883
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5858190256577917e-05,
+      "loss": 0.6138,
+      "step": 8884
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585723951304654e-05,
+      "loss": 0.7324,
+      "step": 8885
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5856288688914396e-05,
+      "loss": 0.6345,
+      "step": 8886
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5855337784194576e-05,
+      "loss": 0.6844,
+      "step": 8887
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5854386798900165e-05,
+      "loss": 0.6719,
+      "step": 8888
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5853435733044246e-05,
+      "loss": 0.637,
+      "step": 8889
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5852484586639905e-05,
+      "loss": 0.6328,
+      "step": 8890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5851533359700238e-05,
+      "loss": 0.6669,
+      "step": 8891
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585058205223833e-05,
+      "loss": 0.6733,
+      "step": 8892
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5849630664267266e-05,
+      "loss": 0.6104,
+      "step": 8893
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584867919580015e-05,
+      "loss": 0.5728,
+      "step": 8894
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584772764685007e-05,
+      "loss": 0.6327,
+      "step": 8895
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5846776017430115e-05,
+      "loss": 0.6405,
+      "step": 8896
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584582430755339e-05,
+      "loss": 0.6503,
+      "step": 8897
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584487251723298e-05,
+      "loss": 0.5864,
+      "step": 8898
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584392064648199e-05,
+      "loss": 0.6817,
+      "step": 8899
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584296869531352e-05,
+      "loss": 0.6042,
+      "step": 8900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5842016663740664e-05,
+      "loss": 0.6462,
+      "step": 8901
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5841064551776528e-05,
+      "loss": 0.6238,
+      "step": 8902
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5840112359434207e-05,
+      "loss": 0.7042,
+      "step": 8903
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5839160086726812e-05,
+      "loss": 0.5788,
+      "step": 8904
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5838207733667443e-05,
+      "loss": 0.6433,
+      "step": 8905
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5837255300269204e-05,
+      "loss": 0.7215,
+      "step": 8906
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5836302786545205e-05,
+      "loss": 0.6533,
+      "step": 8907
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5835350192508553e-05,
+      "loss": 0.6768,
+      "step": 8908
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5834397518172352e-05,
+      "loss": 0.6634,
+      "step": 8909
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5833444763549716e-05,
+      "loss": 0.7042,
+      "step": 8910
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583249192865375e-05,
+      "loss": 0.6729,
+      "step": 8911
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5831539013497574e-05,
+      "loss": 0.6693,
+      "step": 8912
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58305860180943e-05,
+      "loss": 0.6815,
+      "step": 8913
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5829632942457032e-05,
+      "loss": 0.6401,
+      "step": 8914
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5828679786598897e-05,
+      "loss": 0.6824,
+      "step": 8915
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5827726550533007e-05,
+      "loss": 0.6542,
+      "step": 8916
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5826773234272476e-05,
+      "loss": 0.5993,
+      "step": 8917
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582581983783043e-05,
+      "loss": 0.6993,
+      "step": 8918
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582486636121998e-05,
+      "loss": 0.5731,
+      "step": 8919
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582391280445425e-05,
+      "loss": 0.6614,
+      "step": 8920
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5822959167546365e-05,
+      "loss": 0.6173,
+      "step": 8921
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5822005450509447e-05,
+      "loss": 0.6324,
+      "step": 8922
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5821051653356617e-05,
+      "loss": 0.6732,
+      "step": 8923
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5820097776101e-05,
+      "loss": 0.6803,
+      "step": 8924
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5819143818755725e-05,
+      "loss": 0.6874,
+      "step": 8925
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5818189781333914e-05,
+      "loss": 0.6362,
+      "step": 8926
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5817235663848704e-05,
+      "loss": 0.61,
+      "step": 8927
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5816281466313218e-05,
+      "loss": 0.6935,
+      "step": 8928
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581532718874059e-05,
+      "loss": 0.6586,
+      "step": 8929
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5814372831143946e-05,
+      "loss": 0.7429,
+      "step": 8930
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581341839353643e-05,
+      "loss": 0.6387,
+      "step": 8931
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581246387593116e-05,
+      "loss": 0.6738,
+      "step": 8932
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5811509278341285e-05,
+      "loss": 0.5862,
+      "step": 8933
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5810554600779936e-05,
+      "loss": 0.7099,
+      "step": 8934
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5809599843260248e-05,
+      "loss": 0.6428,
+      "step": 8935
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580864500579536e-05,
+      "loss": 0.6407,
+      "step": 8936
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580769008839842e-05,
+      "loss": 0.5726,
+      "step": 8937
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5806735091082555e-05,
+      "loss": 0.6594,
+      "step": 8938
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5805780013860914e-05,
+      "loss": 0.7219,
+      "step": 8939
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580482485674664e-05,
+      "loss": 0.6133,
+      "step": 8940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5803869619752874e-05,
+      "loss": 0.6586,
+      "step": 8941
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5802914302892764e-05,
+      "loss": 0.6392,
+      "step": 8942
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580195890617945e-05,
+      "loss": 0.6919,
+      "step": 8943
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580100342962609e-05,
+      "loss": 0.6952,
+      "step": 8944
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580004787324582e-05,
+      "loss": 0.6251,
+      "step": 8945
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.57990922370518e-05,
+      "loss": 0.5786,
+      "step": 8946
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579813652105717e-05,
+      "loss": 0.6536,
+      "step": 8947
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5797180725275092e-05,
+      "loss": 0.6472,
+      "step": 8948
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5796224849718708e-05,
+      "loss": 0.7254,
+      "step": 8949
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579526889440118e-05,
+      "loss": 0.5977,
+      "step": 8950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579431285933566e-05,
+      "loss": 0.6343,
+      "step": 8951
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5793356744535303e-05,
+      "loss": 0.6246,
+      "step": 8952
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5792400550013267e-05,
+      "loss": 0.696,
+      "step": 8953
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579144427578271e-05,
+      "loss": 0.5755,
+      "step": 8954
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579048792185679e-05,
+      "loss": 0.6537,
+      "step": 8955
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578953148824867e-05,
+      "loss": 0.6615,
+      "step": 8956
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5788574974971506e-05,
+      "loss": 0.6557,
+      "step": 8957
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5787618382038467e-05,
+      "loss": 0.631,
+      "step": 8958
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578666170946271e-05,
+      "loss": 0.5777,
+      "step": 8959
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578570495725741e-05,
+      "loss": 0.683,
+      "step": 8960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578474812543572e-05,
+      "loss": 0.6761,
+      "step": 8961
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5783791214010818e-05,
+      "loss": 0.6904,
+      "step": 8962
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5782834222995864e-05,
+      "loss": 0.6924,
+      "step": 8963
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578187715240403e-05,
+      "loss": 0.6227,
+      "step": 8964
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5780920002248484e-05,
+      "loss": 0.6866,
+      "step": 8965
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5779962772542404e-05,
+      "loss": 0.6658,
+      "step": 8966
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5779005463298954e-05,
+      "loss": 0.6266,
+      "step": 8967
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577804807453131e-05,
+      "loss": 0.6622,
+      "step": 8968
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577709060625265e-05,
+      "loss": 0.6995,
+      "step": 8969
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5776133058476143e-05,
+      "loss": 0.6773,
+      "step": 8970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5775175431214976e-05,
+      "loss": 0.6603,
+      "step": 8971
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577421772448232e-05,
+      "loss": 0.6884,
+      "step": 8972
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5773259938291347e-05,
+      "loss": 0.7332,
+      "step": 8973
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.577230207265525e-05,
+      "loss": 0.7002,
+      "step": 8974
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5771344127587204e-05,
+      "loss": 0.6568,
+      "step": 8975
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5770386103100392e-05,
+      "loss": 0.6345,
+      "step": 8976
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5769427999208e-05,
+      "loss": 0.5954,
+      "step": 8977
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57684698159232e-05,
+      "loss": 0.615,
+      "step": 8978
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5767511553259195e-05,
+      "loss": 0.6003,
+      "step": 8979
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576655321122916e-05,
+      "loss": 0.6926,
+      "step": 8980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5765594789846284e-05,
+      "loss": 0.7258,
+      "step": 8981
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5764636289123763e-05,
+      "loss": 0.6918,
+      "step": 8982
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5763677709074777e-05,
+      "loss": 0.7061,
+      "step": 8983
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5762719049712526e-05,
+      "loss": 0.7126,
+      "step": 8984
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5761760311050193e-05,
+      "loss": 0.5977,
+      "step": 8985
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5760801493100975e-05,
+      "loss": 0.5734,
+      "step": 8986
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575984259587807e-05,
+      "loss": 0.6702,
+      "step": 8987
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5758883619394665e-05,
+      "loss": 0.6513,
+      "step": 8988
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5757924563663966e-05,
+      "loss": 0.6654,
+      "step": 8989
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5756965428699163e-05,
+      "loss": 0.5925,
+      "step": 8990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5756006214513458e-05,
+      "loss": 0.6076,
+      "step": 8991
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5755046921120046e-05,
+      "loss": 0.6449,
+      "step": 8992
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5754087548532136e-05,
+      "loss": 0.6004,
+      "step": 8993
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5753128096762926e-05,
+      "loss": 0.6987,
+      "step": 8994
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5752168565825616e-05,
+      "loss": 0.7019,
+      "step": 8995
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5751208955733415e-05,
+      "loss": 0.721,
+      "step": 8996
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5750249266499524e-05,
+      "loss": 0.651,
+      "step": 8997
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5749289498137147e-05,
+      "loss": 0.6727,
+      "step": 8998
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57483296506595e-05,
+      "loss": 0.6179,
+      "step": 8999
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5747369724079785e-05,
+      "loss": 0.6954,
+      "step": 9000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574640971841121e-05,
+      "loss": 0.6329,
+      "step": 9001
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574544963366699e-05,
+      "loss": 0.5598,
+      "step": 9002
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5744489469860334e-05,
+      "loss": 0.5793,
+      "step": 9003
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5743529227004456e-05,
+      "loss": 0.5683,
+      "step": 9004
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574256890511257e-05,
+      "loss": 0.6604,
+      "step": 9005
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5741608504197887e-05,
+      "loss": 0.7211,
+      "step": 9006
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5740648024273627e-05,
+      "loss": 0.6082,
+      "step": 9007
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5739687465353005e-05,
+      "loss": 0.5785,
+      "step": 9008
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573872682744924e-05,
+      "loss": 0.5921,
+      "step": 9009
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5737766110575554e-05,
+      "loss": 0.6539,
+      "step": 9010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5736805314745166e-05,
+      "loss": 0.5716,
+      "step": 9011
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5735844439971295e-05,
+      "loss": 0.6652,
+      "step": 9012
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5734883486267162e-05,
+      "loss": 0.5829,
+      "step": 9013
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5733922453645993e-05,
+      "loss": 0.6823,
+      "step": 9014
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5732961342121015e-05,
+      "loss": 0.6222,
+      "step": 9015
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573200015170545e-05,
+      "loss": 0.7463,
+      "step": 9016
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5731038882412527e-05,
+      "loss": 0.7317,
+      "step": 9017
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573007753425547e-05,
+      "loss": 0.6966,
+      "step": 9018
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5729116107247516e-05,
+      "loss": 0.6,
+      "step": 9019
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572815460140189e-05,
+      "loss": 0.7022,
+      "step": 9020
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5727193016731823e-05,
+      "loss": 0.6301,
+      "step": 9021
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5726231353250544e-05,
+      "loss": 0.6721,
+      "step": 9022
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5725269610971294e-05,
+      "loss": 0.6807,
+      "step": 9023
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5724307789907304e-05,
+      "loss": 0.6043,
+      "step": 9024
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5723345890071805e-05,
+      "loss": 0.6654,
+      "step": 9025
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572238391147804e-05,
+      "loss": 0.6228,
+      "step": 9026
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5721421854139248e-05,
+      "loss": 0.6291,
+      "step": 9027
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5720459718068658e-05,
+      "loss": 0.6574,
+      "step": 9028
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571949750327952e-05,
+      "loss": 0.6639,
+      "step": 9029
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5718535209785068e-05,
+      "loss": 0.6825,
+      "step": 9030
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5717572837598545e-05,
+      "loss": 0.6134,
+      "step": 9031
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57166103867332e-05,
+      "loss": 0.6155,
+      "step": 9032
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5715647857202273e-05,
+      "loss": 0.6853,
+      "step": 9033
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5714685249019006e-05,
+      "loss": 0.7152,
+      "step": 9034
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571372256219665e-05,
+      "loss": 0.6036,
+      "step": 9035
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5712759796748454e-05,
+      "loss": 0.6999,
+      "step": 9036
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5711796952687663e-05,
+      "loss": 0.6432,
+      "step": 9037
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5710834030027523e-05,
+      "loss": 0.6702,
+      "step": 9038
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570987102878129e-05,
+      "loss": 0.6794,
+      "step": 9039
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5708907948962216e-05,
+      "loss": 0.7062,
+      "step": 9040
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570794479058355e-05,
+      "loss": 0.7001,
+      "step": 9041
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570698155365855e-05,
+      "loss": 0.6205,
+      "step": 9042
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570601823820047e-05,
+      "loss": 0.6466,
+      "step": 9043
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5705054844222564e-05,
+      "loss": 0.5796,
+      "step": 9044
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570409137173809e-05,
+      "loss": 0.6155,
+      "step": 9045
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5703127820760312e-05,
+      "loss": 0.5703,
+      "step": 9046
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5702164191302478e-05,
+      "loss": 0.618,
+      "step": 9047
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5701200483377855e-05,
+      "loss": 0.6137,
+      "step": 9048
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5700236696999703e-05,
+      "loss": 0.624,
+      "step": 9049
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569927283218129e-05,
+      "loss": 0.6374,
+      "step": 9050
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569830888893587e-05,
+      "loss": 0.6916,
+      "step": 9051
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5697344867276716e-05,
+      "loss": 0.6859,
+      "step": 9052
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569638076721709e-05,
+      "loss": 0.6692,
+      "step": 9053
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569541658877026e-05,
+      "loss": 0.6897,
+      "step": 9054
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569445233194949e-05,
+      "loss": 0.628,
+      "step": 9055
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5693487996768056e-05,
+      "loss": 0.6934,
+      "step": 9056
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5692523583239222e-05,
+      "loss": 0.614,
+      "step": 9057
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569155909137626e-05,
+      "loss": 0.7511,
+      "step": 9058
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569059452119245e-05,
+      "loss": 0.6886,
+      "step": 9059
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5689629872701053e-05,
+      "loss": 0.6454,
+      "step": 9060
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5688665145915352e-05,
+      "loss": 0.5508,
+      "step": 9061
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5687700340848618e-05,
+      "loss": 0.6761,
+      "step": 9062
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5686735457514136e-05,
+      "loss": 0.6673,
+      "step": 9063
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5685770495925173e-05,
+      "loss": 0.5971,
+      "step": 9064
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568480545609501e-05,
+      "loss": 0.6262,
+      "step": 9065
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5683840338036934e-05,
+      "loss": 0.6861,
+      "step": 9066
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5682875141764218e-05,
+      "loss": 0.6567,
+      "step": 9067
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5681909867290144e-05,
+      "loss": 0.6719,
+      "step": 9068
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5680944514628e-05,
+      "loss": 0.606,
+      "step": 9069
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567997908379107e-05,
+      "loss": 0.6418,
+      "step": 9070
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5679013574792634e-05,
+      "loss": 0.6986,
+      "step": 9071
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567804798764598e-05,
+      "loss": 0.7282,
+      "step": 9072
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56770823223644e-05,
+      "loss": 0.611,
+      "step": 9073
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5676116578961176e-05,
+      "loss": 0.6364,
+      "step": 9074
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5675150757449603e-05,
+      "loss": 0.6752,
+      "step": 9075
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5674184857842967e-05,
+      "loss": 0.6807,
+      "step": 9076
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567321888015456e-05,
+      "loss": 0.6097,
+      "step": 9077
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5672252824397683e-05,
+      "loss": 0.5923,
+      "step": 9078
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5671286690585615e-05,
+      "loss": 0.6754,
+      "step": 9079
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567032047873166e-05,
+      "loss": 0.6784,
+      "step": 9080
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5669354188849116e-05,
+      "loss": 0.5515,
+      "step": 9081
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566838782095128e-05,
+      "loss": 0.556,
+      "step": 9082
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5667421375051443e-05,
+      "loss": 0.7154,
+      "step": 9083
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5666454851162908e-05,
+      "loss": 0.6989,
+      "step": 9084
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5665488249298976e-05,
+      "loss": 0.559,
+      "step": 9085
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5664521569472945e-05,
+      "loss": 0.7126,
+      "step": 9086
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5663554811698122e-05,
+      "loss": 0.5977,
+      "step": 9087
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5662587975987807e-05,
+      "loss": 0.5908,
+      "step": 9088
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566162106235531e-05,
+      "loss": 0.7541,
+      "step": 9089
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5660654070813928e-05,
+      "loss": 0.5821,
+      "step": 9090
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5659687001376974e-05,
+      "loss": 0.5837,
+      "step": 9091
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5658719854057752e-05,
+      "loss": 0.6594,
+      "step": 9092
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5657752628869576e-05,
+      "loss": 0.6429,
+      "step": 9093
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565678532582575e-05,
+      "loss": 0.7543,
+      "step": 9094
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5655817944939588e-05,
+      "loss": 0.6453,
+      "step": 9095
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56548504862244e-05,
+      "loss": 0.63,
+      "step": 9096
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5653882949693503e-05,
+      "loss": 0.6186,
+      "step": 9097
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5652915335360207e-05,
+      "loss": 0.6812,
+      "step": 9098
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565194764323783e-05,
+      "loss": 0.6245,
+      "step": 9099
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5650979873339686e-05,
+      "loss": 0.6167,
+      "step": 9100
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5650012025679095e-05,
+      "loss": 0.6409,
+      "step": 9101
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5649044100269374e-05,
+      "loss": 0.6261,
+      "step": 9102
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564807609712384e-05,
+      "loss": 0.6482,
+      "step": 9103
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564710801625582e-05,
+      "loss": 0.5932,
+      "step": 9104
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564613985767863e-05,
+      "loss": 0.6062,
+      "step": 9105
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5645171621405594e-05,
+      "loss": 0.6283,
+      "step": 9106
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5644203307450036e-05,
+      "loss": 0.7022,
+      "step": 9107
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564323491582528e-05,
+      "loss": 0.6338,
+      "step": 9108
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5642266446544655e-05,
+      "loss": 0.6533,
+      "step": 9109
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5641297899621486e-05,
+      "loss": 0.6745,
+      "step": 9110
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56403292750691e-05,
+      "loss": 0.6207,
+      "step": 9111
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563936057290083e-05,
+      "loss": 0.6631,
+      "step": 9112
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5638391793129998e-05,
+      "loss": 0.6682,
+      "step": 9113
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5637422935769942e-05,
+      "loss": 0.609,
+      "step": 9114
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5636454000833995e-05,
+      "loss": 0.5713,
+      "step": 9115
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5635484988335488e-05,
+      "loss": 0.696,
+      "step": 9116
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5634515898287753e-05,
+      "loss": 0.5804,
+      "step": 9117
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5633546730704134e-05,
+      "loss": 0.5802,
+      "step": 9118
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5632577485597958e-05,
+      "loss": 0.6297,
+      "step": 9119
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563160816298257e-05,
+      "loss": 0.6448,
+      "step": 9120
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56306387628713e-05,
+      "loss": 0.6833,
+      "step": 9121
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5629669285277497e-05,
+      "loss": 0.5761,
+      "step": 9122
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5628699730214495e-05,
+      "loss": 0.722,
+      "step": 9123
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562773009769564e-05,
+      "loss": 0.6641,
+      "step": 9124
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5626760387734273e-05,
+      "loss": 0.6038,
+      "step": 9125
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562579060034374e-05,
+      "loss": 0.67,
+      "step": 9126
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562482073553739e-05,
+      "loss": 0.6731,
+      "step": 9127
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5623850793328556e-05,
+      "loss": 0.7143,
+      "step": 9128
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56228807737306e-05,
+      "loss": 0.6382,
+      "step": 9129
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562191067675686e-05,
+      "loss": 0.6749,
+      "step": 9130
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562094050242069e-05,
+      "loss": 0.7031,
+      "step": 9131
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561997025073544e-05,
+      "loss": 0.6792,
+      "step": 9132
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561899992171446e-05,
+      "loss": 0.5649,
+      "step": 9133
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5618029515371106e-05,
+      "loss": 0.7018,
+      "step": 9134
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561705903171873e-05,
+      "loss": 0.6951,
+      "step": 9135
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561608847077068e-05,
+      "loss": 0.723,
+      "step": 9136
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5615117832540324e-05,
+      "loss": 0.6718,
+      "step": 9137
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5614147117041012e-05,
+      "loss": 0.5865,
+      "step": 9138
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5613176324286102e-05,
+      "loss": 0.5631,
+      "step": 9139
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5612205454288957e-05,
+      "loss": 0.7001,
+      "step": 9140
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5611234507062928e-05,
+      "loss": 0.6256,
+      "step": 9141
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5610263482621383e-05,
+      "loss": 0.6379,
+      "step": 9142
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5609292380977684e-05,
+      "loss": 0.6427,
+      "step": 9143
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5608321202145194e-05,
+      "loss": 0.686,
+      "step": 9144
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5607349946137275e-05,
+      "loss": 0.6312,
+      "step": 9145
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5606378612967295e-05,
+      "loss": 0.6283,
+      "step": 9146
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5605407202648618e-05,
+      "loss": 0.6417,
+      "step": 9147
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5604435715194612e-05,
+      "loss": 0.669,
+      "step": 9148
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5603464150618645e-05,
+      "loss": 0.6822,
+      "step": 9149
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560249250893409e-05,
+      "loss": 0.624,
+      "step": 9150
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5601520790154313e-05,
+      "loss": 0.6269,
+      "step": 9151
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560054899429269e-05,
+      "loss": 0.6904,
+      "step": 9152
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5599577121362592e-05,
+      "loss": 0.7277,
+      "step": 9153
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559860517137739e-05,
+      "loss": 0.6773,
+      "step": 9154
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5597633144350465e-05,
+      "loss": 0.5775,
+      "step": 9155
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559666104029519e-05,
+      "loss": 0.6396,
+      "step": 9156
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5595688859224937e-05,
+      "loss": 0.6848,
+      "step": 9157
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559471660115309e-05,
+      "loss": 0.7312,
+      "step": 9158
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559374426609303e-05,
+      "loss": 0.7192,
+      "step": 9159
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5592771854058128e-05,
+      "loss": 0.5933,
+      "step": 9160
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5591799365061774e-05,
+      "loss": 0.6886,
+      "step": 9161
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5590826799117348e-05,
+      "loss": 0.5748,
+      "step": 9162
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5589854156238235e-05,
+      "loss": 0.6154,
+      "step": 9163
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5588881436437813e-05,
+      "loss": 0.6839,
+      "step": 9164
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5587908639729472e-05,
+      "loss": 0.7214,
+      "step": 9165
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5586935766126597e-05,
+      "loss": 0.64,
+      "step": 9166
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5585962815642583e-05,
+      "loss": 0.6639,
+      "step": 9167
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5584989788290806e-05,
+      "loss": 0.5786,
+      "step": 9168
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5584016684084664e-05,
+      "loss": 0.7117,
+      "step": 9169
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5583043503037547e-05,
+      "loss": 0.6104,
+      "step": 9170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5582070245162847e-05,
+      "loss": 0.5228,
+      "step": 9171
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558109691047395e-05,
+      "loss": 0.5909,
+      "step": 9172
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558012349898426e-05,
+      "loss": 0.6819,
+      "step": 9173
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557915001070717e-05,
+      "loss": 0.6939,
+      "step": 9174
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557817644565607e-05,
+      "loss": 0.5807,
+      "step": 9175
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557720280384436e-05,
+      "loss": 0.6995,
+      "step": 9176
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5576229085285438e-05,
+      "loss": 0.5885,
+      "step": 9177
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5575255289992707e-05,
+      "loss": 0.6841,
+      "step": 9178
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557428141797956e-05,
+      "loss": 0.6107,
+      "step": 9179
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557330746925941e-05,
+      "loss": 0.7374,
+      "step": 9180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5572333443845647e-05,
+      "loss": 0.5693,
+      "step": 9181
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557135934175168e-05,
+      "loss": 0.6092,
+      "step": 9182
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5570385162990916e-05,
+      "loss": 0.5778,
+      "step": 9183
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5569410907576753e-05,
+      "loss": 0.6821,
+      "step": 9184
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5568436575522604e-05,
+      "loss": 0.6499,
+      "step": 9185
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5567462166841874e-05,
+      "loss": 0.6341,
+      "step": 9186
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5566487681547977e-05,
+      "loss": 0.6099,
+      "step": 9187
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556551311965431e-05,
+      "loss": 0.7293,
+      "step": 9188
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55645384811743e-05,
+      "loss": 0.647,
+      "step": 9189
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556356376612135e-05,
+      "loss": 0.6415,
+      "step": 9190
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556258897450887e-05,
+      "loss": 0.5785,
+      "step": 9191
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556161410635028e-05,
+      "loss": 0.6072,
+      "step": 9192
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5560639161658997e-05,
+      "loss": 0.5867,
+      "step": 9193
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5559664140448428e-05,
+      "loss": 0.6805,
+      "step": 9194
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5558689042732e-05,
+      "loss": 0.5465,
+      "step": 9195
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5557713868523123e-05,
+      "loss": 0.7046,
+      "step": 9196
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5556738617835217e-05,
+      "loss": 0.725,
+      "step": 9197
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555576329068171e-05,
+      "loss": 0.6975,
+      "step": 9198
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5554787887076016e-05,
+      "loss": 0.6143,
+      "step": 9199
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555381240703156e-05,
+      "loss": 0.7382,
+      "step": 9200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5552836850561764e-05,
+      "loss": 0.7001,
+      "step": 9201
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5551861217680057e-05,
+      "loss": 0.6879,
+      "step": 9202
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5550885508399857e-05,
+      "loss": 0.6475,
+      "step": 9203
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55499097227346e-05,
+      "loss": 0.6026,
+      "step": 9204
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5548933860697706e-05,
+      "loss": 0.6662,
+      "step": 9205
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5547957922302606e-05,
+      "loss": 0.6253,
+      "step": 9206
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554698190756273e-05,
+      "loss": 0.6522,
+      "step": 9207
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554600581649151e-05,
+      "loss": 0.6453,
+      "step": 9208
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5545029649102378e-05,
+      "loss": 0.6689,
+      "step": 9209
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5544053405408763e-05,
+      "loss": 0.6552,
+      "step": 9210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5543077085424104e-05,
+      "loss": 0.6448,
+      "step": 9211
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5542100689161833e-05,
+      "loss": 0.6597,
+      "step": 9212
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554112421663539e-05,
+      "loss": 0.6632,
+      "step": 9213
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5540147667858204e-05,
+      "loss": 0.5875,
+      "step": 9214
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553917104284372e-05,
+      "loss": 0.6503,
+      "step": 9215
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5538194341605372e-05,
+      "loss": 0.6153,
+      "step": 9216
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553721756415661e-05,
+      "loss": 0.6419,
+      "step": 9217
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5536240710510865e-05,
+      "loss": 0.6243,
+      "step": 9218
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5535263780681582e-05,
+      "loss": 0.6951,
+      "step": 9219
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5534286774682205e-05,
+      "loss": 0.6068,
+      "step": 9220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5533309692526183e-05,
+      "loss": 0.6603,
+      "step": 9221
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5532332534226957e-05,
+      "loss": 0.6463,
+      "step": 9222
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5531355299797973e-05,
+      "loss": 0.6644,
+      "step": 9223
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553037798925268e-05,
+      "loss": 0.6664,
+      "step": 9224
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552940060260453e-05,
+      "loss": 0.6359,
+      "step": 9225
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552842313986696e-05,
+      "loss": 0.6185,
+      "step": 9226
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5527445601053437e-05,
+      "loss": 0.7147,
+      "step": 9227
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5526467986177405e-05,
+      "loss": 0.6663,
+      "step": 9228
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5525490295252316e-05,
+      "loss": 0.6756,
+      "step": 9229
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5524512528291627e-05,
+      "loss": 0.6502,
+      "step": 9230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5523534685308792e-05,
+      "loss": 0.6816,
+      "step": 9231
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5522556766317266e-05,
+      "loss": 0.6221,
+      "step": 9232
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55215787713305e-05,
+      "loss": 0.6085,
+      "step": 9233
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5520600700361964e-05,
+      "loss": 0.6052,
+      "step": 9234
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5519622553425114e-05,
+      "loss": 0.6209,
+      "step": 9235
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5518644330533403e-05,
+      "loss": 0.6415,
+      "step": 9236
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55176660317003e-05,
+      "loss": 0.697,
+      "step": 9237
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5516687656939263e-05,
+      "loss": 0.6463,
+      "step": 9238
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5515709206263757e-05,
+      "loss": 0.6764,
+      "step": 9239
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5514730679687244e-05,
+      "loss": 0.6613,
+      "step": 9240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5513752077223193e-05,
+      "loss": 0.6704,
+      "step": 9241
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551277339888507e-05,
+      "loss": 0.6598,
+      "step": 9242
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5511794644686337e-05,
+      "loss": 0.6436,
+      "step": 9243
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5510815814640468e-05,
+      "loss": 0.6135,
+      "step": 9244
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.550983690876093e-05,
+      "loss": 0.5843,
+      "step": 9245
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55088579270612e-05,
+      "loss": 0.6607,
+      "step": 9246
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5507878869554738e-05,
+      "loss": 0.5619,
+      "step": 9247
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5506899736255027e-05,
+      "loss": 0.6897,
+      "step": 9248
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5505920527175534e-05,
+      "loss": 0.6901,
+      "step": 9249
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550494124232974e-05,
+      "loss": 0.6854,
+      "step": 9250
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550396188173112e-05,
+      "loss": 0.6092,
+      "step": 9251
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5502982445393144e-05,
+      "loss": 0.5834,
+      "step": 9252
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550200293332929e-05,
+      "loss": 0.621,
+      "step": 9253
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5501023345553046e-05,
+      "loss": 0.5616,
+      "step": 9254
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5500043682077887e-05,
+      "loss": 0.5993,
+      "step": 9255
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5499063942917295e-05,
+      "loss": 0.5969,
+      "step": 9256
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549808412808475e-05,
+      "loss": 0.6577,
+      "step": 9257
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5497104237593736e-05,
+      "loss": 0.6607,
+      "step": 9258
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549612427145774e-05,
+      "loss": 0.6503,
+      "step": 9259
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5495144229690243e-05,
+      "loss": 0.5905,
+      "step": 9260
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5494164112304734e-05,
+      "loss": 0.6636,
+      "step": 9261
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5493183919314696e-05,
+      "loss": 0.6998,
+      "step": 9262
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5492203650733623e-05,
+      "loss": 0.5858,
+      "step": 9263
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5491223306575e-05,
+      "loss": 0.6531,
+      "step": 9264
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5490242886852323e-05,
+      "loss": 0.6844,
+      "step": 9265
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5489262391579076e-05,
+      "loss": 0.7394,
+      "step": 9266
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548828182076876e-05,
+      "loss": 0.6368,
+      "step": 9267
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5487301174434858e-05,
+      "loss": 0.7041,
+      "step": 9268
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5486320452590872e-05,
+      "loss": 0.7089,
+      "step": 9269
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5485339655250294e-05,
+      "loss": 0.7062,
+      "step": 9270
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5484358782426626e-05,
+      "loss": 0.6869,
+      "step": 9271
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5483377834133362e-05,
+      "loss": 0.6579,
+      "step": 9272
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5482396810384002e-05,
+      "loss": 0.7357,
+      "step": 9273
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5481415711192042e-05,
+      "loss": 0.6381,
+      "step": 9274
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5480434536570986e-05,
+      "loss": 0.7166,
+      "step": 9275
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5479453286534334e-05,
+      "loss": 0.6444,
+      "step": 9276
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547847196109559e-05,
+      "loss": 0.6012,
+      "step": 9277
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547749056026826e-05,
+      "loss": 0.6857,
+      "step": 9278
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5476509084065846e-05,
+      "loss": 0.666,
+      "step": 9279
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5475527532501857e-05,
+      "loss": 0.6821,
+      "step": 9280
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5474545905589797e-05,
+      "loss": 0.6468,
+      "step": 9281
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5473564203343173e-05,
+      "loss": 0.6792,
+      "step": 9282
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54725824257755e-05,
+      "loss": 0.711,
+      "step": 9283
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5471600572900278e-05,
+      "loss": 0.7561,
+      "step": 9284
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5470618644731027e-05,
+      "loss": 0.6417,
+      "step": 9285
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5469636641281262e-05,
+      "loss": 0.68,
+      "step": 9286
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546865456256448e-05,
+      "loss": 0.6727,
+      "step": 9287
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5467672408594214e-05,
+      "loss": 0.6187,
+      "step": 9288
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546669017938397e-05,
+      "loss": 0.6121,
+      "step": 9289
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5465707874947266e-05,
+      "loss": 0.6838,
+      "step": 9290
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546472549529762e-05,
+      "loss": 0.6429,
+      "step": 9291
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5463743040448547e-05,
+      "loss": 0.6308,
+      "step": 9292
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5462760510413568e-05,
+      "loss": 0.6897,
+      "step": 9293
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5461777905206206e-05,
+      "loss": 0.6337,
+      "step": 9294
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5460795224839982e-05,
+      "loss": 0.6363,
+      "step": 9295
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5459812469328417e-05,
+      "loss": 0.6413,
+      "step": 9296
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5458829638685032e-05,
+      "loss": 0.6713,
+      "step": 9297
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5457846732923358e-05,
+      "loss": 0.6461,
+      "step": 9298
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5456863752056918e-05,
+      "loss": 0.6433,
+      "step": 9299
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5455880696099236e-05,
+      "loss": 0.6628,
+      "step": 9300
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5454897565063843e-05,
+      "loss": 0.5817,
+      "step": 9301
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5453914358964265e-05,
+      "loss": 0.6058,
+      "step": 9302
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545293107781403e-05,
+      "loss": 0.5634,
+      "step": 9303
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5451947721626676e-05,
+      "loss": 0.5916,
+      "step": 9304
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5450964290415732e-05,
+      "loss": 0.6061,
+      "step": 9305
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544998078419473e-05,
+      "loss": 0.6422,
+      "step": 9306
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5448997202977198e-05,
+      "loss": 0.6232,
+      "step": 9307
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5448013546776682e-05,
+      "loss": 0.715,
+      "step": 9308
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5447029815606713e-05,
+      "loss": 0.5801,
+      "step": 9309
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5446046009480823e-05,
+      "loss": 0.7153,
+      "step": 9310
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5445062128412555e-05,
+      "loss": 0.7106,
+      "step": 9311
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544407817241545e-05,
+      "loss": 0.6982,
+      "step": 9312
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5443094141503045e-05,
+      "loss": 0.644,
+      "step": 9313
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5442110035688883e-05,
+      "loss": 0.5849,
+      "step": 9314
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5441125854986504e-05,
+      "loss": 0.6712,
+      "step": 9315
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5440141599409452e-05,
+      "loss": 0.6386,
+      "step": 9316
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5439157268971276e-05,
+      "loss": 0.6937,
+      "step": 9317
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543817286368551e-05,
+      "loss": 0.6983,
+      "step": 9318
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543718838356571e-05,
+      "loss": 0.6032,
+      "step": 9319
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543620382862542e-05,
+      "loss": 0.7112,
+      "step": 9320
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5435219198878192e-05,
+      "loss": 0.6505,
+      "step": 9321
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5434234494337568e-05,
+      "loss": 0.6798,
+      "step": 9322
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5433249715017103e-05,
+      "loss": 0.6414,
+      "step": 9323
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5432264860930345e-05,
+      "loss": 0.5699,
+      "step": 9324
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5431279932090856e-05,
+      "loss": 0.6619,
+      "step": 9325
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5430294928512178e-05,
+      "loss": 0.6433,
+      "step": 9326
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542930985020787e-05,
+      "loss": 0.5497,
+      "step": 9327
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542832469719149e-05,
+      "loss": 0.75,
+      "step": 9328
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5427339469476594e-05,
+      "loss": 0.6754,
+      "step": 9329
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5426354167076733e-05,
+      "loss": 0.6789,
+      "step": 9330
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5425368790005474e-05,
+      "loss": 0.6221,
+      "step": 9331
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5424383338276375e-05,
+      "loss": 0.6452,
+      "step": 9332
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5423397811902988e-05,
+      "loss": 0.657,
+      "step": 9333
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542241221089889e-05,
+      "loss": 0.6423,
+      "step": 9334
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5421426535277628e-05,
+      "loss": 0.636,
+      "step": 9335
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542044078505278e-05,
+      "loss": 0.6932,
+      "step": 9336
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54194549602379e-05,
+      "loss": 0.5919,
+      "step": 9337
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5418469060846558e-05,
+      "loss": 0.6391,
+      "step": 9338
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5417483086892322e-05,
+      "loss": 0.5929,
+      "step": 9339
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541649703838876e-05,
+      "loss": 0.6124,
+      "step": 9340
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5415510915349433e-05,
+      "loss": 0.6089,
+      "step": 9341
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5414524717787925e-05,
+      "loss": 0.6679,
+      "step": 9342
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5413538445717795e-05,
+      "loss": 0.6009,
+      "step": 9343
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5412552099152617e-05,
+      "loss": 0.6965,
+      "step": 9344
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5411565678105972e-05,
+      "loss": 0.5732,
+      "step": 9345
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5410579182591427e-05,
+      "loss": 0.624,
+      "step": 9346
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5409592612622556e-05,
+      "loss": 0.714,
+      "step": 9347
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5408605968212937e-05,
+      "loss": 0.7164,
+      "step": 9348
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540761924937615e-05,
+      "loss": 0.6339,
+      "step": 9349
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540663245612577e-05,
+      "loss": 0.6446,
+      "step": 9350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5405645588475375e-05,
+      "loss": 0.6295,
+      "step": 9351
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5404658646438547e-05,
+      "loss": 0.5814,
+      "step": 9352
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540367163002887e-05,
+      "loss": 0.6363,
+      "step": 9353
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5402684539259918e-05,
+      "loss": 0.5925,
+      "step": 9354
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5401697374145285e-05,
+      "loss": 0.5493,
+      "step": 9355
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5400710134698548e-05,
+      "loss": 0.7033,
+      "step": 9356
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5399722820933295e-05,
+      "loss": 0.6926,
+      "step": 9357
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5398735432863107e-05,
+      "loss": 0.648,
+      "step": 9358
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539774797050158e-05,
+      "loss": 0.6529,
+      "step": 9359
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5396760433862295e-05,
+      "loss": 0.5864,
+      "step": 9360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5395772822958844e-05,
+      "loss": 0.6834,
+      "step": 9361
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5394785137804823e-05,
+      "loss": 0.5923,
+      "step": 9362
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5393797378413816e-05,
+      "loss": 0.6597,
+      "step": 9363
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5392809544799415e-05,
+      "loss": 0.6021,
+      "step": 9364
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539182163697522e-05,
+      "loss": 0.6471,
+      "step": 9365
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5390833654954818e-05,
+      "loss": 0.6943,
+      "step": 9366
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538984559875181e-05,
+      "loss": 0.7071,
+      "step": 9367
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538885746837979e-05,
+      "loss": 0.6859,
+      "step": 9368
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5387869263852356e-05,
+      "loss": 0.6727,
+      "step": 9369
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5386880985183108e-05,
+      "loss": 0.66,
+      "step": 9370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5385892632385643e-05,
+      "loss": 0.6527,
+      "step": 9371
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5384904205473564e-05,
+      "loss": 0.6165,
+      "step": 9372
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538391570446047e-05,
+      "loss": 0.6118,
+      "step": 9373
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5382927129359968e-05,
+      "loss": 0.6206,
+      "step": 9374
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5381938480185658e-05,
+      "loss": 0.696,
+      "step": 9375
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5380949756951144e-05,
+      "loss": 0.6305,
+      "step": 9376
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5379960959670036e-05,
+      "loss": 0.6944,
+      "step": 9377
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5378972088355934e-05,
+      "loss": 0.6005,
+      "step": 9378
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5377983143022453e-05,
+      "loss": 0.7014,
+      "step": 9379
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5376994123683194e-05,
+      "loss": 0.6074,
+      "step": 9380
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5376005030351776e-05,
+      "loss": 0.6954,
+      "step": 9381
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53750158630418e-05,
+      "loss": 0.6892,
+      "step": 9382
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5374026621766886e-05,
+      "loss": 0.5878,
+      "step": 9383
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5373037306540646e-05,
+      "loss": 0.6485,
+      "step": 9384
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537204791737669e-05,
+      "loss": 0.6702,
+      "step": 9385
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5371058454288633e-05,
+      "loss": 0.6418,
+      "step": 9386
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537006891729009e-05,
+      "loss": 0.6083,
+      "step": 9387
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5369079306394684e-05,
+      "loss": 0.6636,
+      "step": 9388
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5368089621616026e-05,
+      "loss": 0.7269,
+      "step": 9389
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5367099862967742e-05,
+      "loss": 0.6878,
+      "step": 9390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5366110030463444e-05,
+      "loss": 0.6306,
+      "step": 9391
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536512012411676e-05,
+      "loss": 0.7006,
+      "step": 9392
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536413014394131e-05,
+      "loss": 0.5812,
+      "step": 9393
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536314008995071e-05,
+      "loss": 0.6789,
+      "step": 9394
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5362149962158593e-05,
+      "loss": 0.6346,
+      "step": 9395
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5361159760578584e-05,
+      "loss": 0.6553,
+      "step": 9396
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5360169485224302e-05,
+      "loss": 0.625,
+      "step": 9397
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535917913610938e-05,
+      "loss": 0.7067,
+      "step": 9398
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5358188713247443e-05,
+      "loss": 0.6695,
+      "step": 9399
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5357198216652123e-05,
+      "loss": 0.7061,
+      "step": 9400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5356207646337045e-05,
+      "loss": 0.6423,
+      "step": 9401
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5355217002315844e-05,
+      "loss": 0.6442,
+      "step": 9402
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5354226284602153e-05,
+      "loss": 0.5692,
+      "step": 9403
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5353235493209606e-05,
+      "loss": 0.7029,
+      "step": 9404
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535224462815183e-05,
+      "loss": 0.6006,
+      "step": 9405
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535125368944247e-05,
+      "loss": 0.6817,
+      "step": 9406
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5350262677095153e-05,
+      "loss": 0.5848,
+      "step": 9407
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5349271591123524e-05,
+      "loss": 0.6815,
+      "step": 9408
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5348280431541215e-05,
+      "loss": 0.6323,
+      "step": 9409
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5347289198361866e-05,
+      "loss": 0.6802,
+      "step": 9410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5346297891599126e-05,
+      "loss": 0.6534,
+      "step": 9411
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5345306511266624e-05,
+      "loss": 0.5942,
+      "step": 9412
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5344315057378008e-05,
+      "loss": 0.5849,
+      "step": 9413
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5343323529946918e-05,
+      "loss": 0.6149,
+      "step": 9414
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5342331928987005e-05,
+      "loss": 0.6404,
+      "step": 9415
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534134025451191e-05,
+      "loss": 0.6627,
+      "step": 9416
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534034850653528e-05,
+      "loss": 0.5901,
+      "step": 9417
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533935668507076e-05,
+      "loss": 0.6197,
+      "step": 9418
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533836479013201e-05,
+      "loss": 0.7343,
+      "step": 9419
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5337372821732656e-05,
+      "loss": 0.643,
+      "step": 9420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5336380779886372e-05,
+      "loss": 0.6676,
+      "step": 9421
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5335388664606793e-05,
+      "loss": 0.6398,
+      "step": 9422
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5334396475907585e-05,
+      "loss": 0.6305,
+      "step": 9423
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533340421380239e-05,
+      "loss": 0.5813,
+      "step": 9424
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5332411878304864e-05,
+      "loss": 0.7155,
+      "step": 9425
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533141946942867e-05,
+      "loss": 0.6595,
+      "step": 9426
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5330426987187456e-05,
+      "loss": 0.6797,
+      "step": 9427
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5329434431594883e-05,
+      "loss": 0.6229,
+      "step": 9428
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5328441802664612e-05,
+      "loss": 0.6417,
+      "step": 9429
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53274491004103e-05,
+      "loss": 0.6313,
+      "step": 9430
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5326456324845605e-05,
+      "loss": 0.7081,
+      "step": 9431
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532546347598419e-05,
+      "loss": 0.6087,
+      "step": 9432
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532447055383972e-05,
+      "loss": 0.6311,
+      "step": 9433
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5323477558425858e-05,
+      "loss": 0.6733,
+      "step": 9434
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5322484489756267e-05,
+      "loss": 0.6343,
+      "step": 9435
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5321491347844608e-05,
+      "loss": 0.7168,
+      "step": 9436
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5320498132704558e-05,
+      "loss": 0.7063,
+      "step": 9437
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5319504844349773e-05,
+      "loss": 0.7005,
+      "step": 9438
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531851148279393e-05,
+      "loss": 0.5596,
+      "step": 9439
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5317518048050698e-05,
+      "loss": 0.6574,
+      "step": 9440
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531652454013374e-05,
+      "loss": 0.7,
+      "step": 9441
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531553095905674e-05,
+      "loss": 0.6214,
+      "step": 9442
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5314537304833356e-05,
+      "loss": 0.5851,
+      "step": 9443
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5313543577477274e-05,
+      "loss": 0.6158,
+      "step": 9444
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531254977700216e-05,
+      "loss": 0.6954,
+      "step": 9445
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5311555903421697e-05,
+      "loss": 0.6807,
+      "step": 9446
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5310561956749554e-05,
+      "loss": 0.6985,
+      "step": 9447
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5309567936999413e-05,
+      "loss": 0.7389,
+      "step": 9448
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5308573844184955e-05,
+      "loss": 0.645,
+      "step": 9449
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5307579678319852e-05,
+      "loss": 0.5705,
+      "step": 9450
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5306585439417792e-05,
+      "loss": 0.67,
+      "step": 9451
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5305591127492455e-05,
+      "loss": 0.7108,
+      "step": 9452
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5304596742557516e-05,
+      "loss": 0.6765,
+      "step": 9453
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530360228462667e-05,
+      "loss": 0.5761,
+      "step": 9454
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5302607753713597e-05,
+      "loss": 0.6725,
+      "step": 9455
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5301613149831978e-05,
+      "loss": 0.6799,
+      "step": 9456
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5300618472995507e-05,
+      "loss": 0.5814,
+      "step": 9457
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5299623723217867e-05,
+      "loss": 0.7116,
+      "step": 9458
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5298628900512747e-05,
+      "loss": 0.593,
+      "step": 9459
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5297634004893838e-05,
+      "loss": 0.6023,
+      "step": 9460
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529663903637483e-05,
+      "loss": 0.6954,
+      "step": 9461
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5295643994969417e-05,
+      "loss": 0.6331,
+      "step": 9462
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5294648880691284e-05,
+      "loss": 0.5659,
+      "step": 9463
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529365369355414e-05,
+      "loss": 0.6832,
+      "step": 9464
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529265843357166e-05,
+      "loss": 0.6754,
+      "step": 9465
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5291663100757554e-05,
+      "loss": 0.6387,
+      "step": 9466
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5290667695125512e-05,
+      "loss": 0.7003,
+      "step": 9467
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5289672216689236e-05,
+      "loss": 0.5794,
+      "step": 9468
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528867666546242e-05,
+      "loss": 0.6506,
+      "step": 9469
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5287681041458767e-05,
+      "loss": 0.6718,
+      "step": 9470
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5286685344691977e-05,
+      "loss": 0.6267,
+      "step": 9471
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528568957517575e-05,
+      "loss": 0.635,
+      "step": 9472
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5284693732923793e-05,
+      "loss": 0.5968,
+      "step": 9473
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5283697817949806e-05,
+      "loss": 0.6833,
+      "step": 9474
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5282701830267493e-05,
+      "loss": 0.7662,
+      "step": 9475
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528170576989056e-05,
+      "loss": 0.6835,
+      "step": 9476
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5280709636832718e-05,
+      "loss": 0.6741,
+      "step": 9477
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527971343110767e-05,
+      "loss": 0.575,
+      "step": 9478
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5278717152729124e-05,
+      "loss": 0.6549,
+      "step": 9479
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5277720801710795e-05,
+      "loss": 0.7161,
+      "step": 9480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5276724378066384e-05,
+      "loss": 0.6374,
+      "step": 9481
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5275727881809617e-05,
+      "loss": 0.636,
+      "step": 9482
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5274731312954198e-05,
+      "loss": 0.6033,
+      "step": 9483
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5273734671513838e-05,
+      "loss": 0.5587,
+      "step": 9484
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5272737957502254e-05,
+      "loss": 0.6895,
+      "step": 9485
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5271741170933167e-05,
+      "loss": 0.6013,
+      "step": 9486
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527074431182028e-05,
+      "loss": 0.6295,
+      "step": 9487
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526974738017733e-05,
+      "loss": 0.6566,
+      "step": 9488
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5268750376018022e-05,
+      "loss": 0.602,
+      "step": 9489
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526775329935608e-05,
+      "loss": 0.5709,
+      "step": 9490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5266756150205224e-05,
+      "loss": 0.6904,
+      "step": 9491
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5265758928579175e-05,
+      "loss": 0.6089,
+      "step": 9492
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5264761634491656e-05,
+      "loss": 0.6401,
+      "step": 9493
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5263764267956392e-05,
+      "loss": 0.6019,
+      "step": 9494
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526276682898711e-05,
+      "loss": 0.6504,
+      "step": 9495
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5261769317597523e-05,
+      "loss": 0.5836,
+      "step": 9496
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5260771733801375e-05,
+      "loss": 0.6612,
+      "step": 9497
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.525977407761238e-05,
+      "loss": 0.6129,
+      "step": 9498
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5258776349044276e-05,
+      "loss": 0.6438,
+      "step": 9499
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5257778548110786e-05,
+      "loss": 0.6406,
+      "step": 9500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5256780674825641e-05,
+      "loss": 0.5935,
+      "step": 9501
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.525578272920258e-05,
+      "loss": 0.5655,
+      "step": 9502
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5254784711255329e-05,
+      "loss": 0.6885,
+      "step": 9503
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.525378662099762e-05,
+      "loss": 0.6092,
+      "step": 9504
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5252788458443197e-05,
+      "loss": 0.6561,
+      "step": 9505
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5251790223605784e-05,
+      "loss": 0.6938,
+      "step": 9506
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5250791916499126e-05,
+      "loss": 0.562,
+      "step": 9507
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5249793537136957e-05,
+      "loss": 0.563,
+      "step": 9508
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5248795085533013e-05,
+      "loss": 0.6677,
+      "step": 9509
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5247796561701038e-05,
+      "loss": 0.6806,
+      "step": 9510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5246797965654773e-05,
+      "loss": 0.6732,
+      "step": 9511
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5245799297407957e-05,
+      "loss": 0.6418,
+      "step": 9512
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5244800556974334e-05,
+      "loss": 0.5666,
+      "step": 9513
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5243801744367645e-05,
+      "loss": 0.6235,
+      "step": 9514
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5242802859601637e-05,
+      "loss": 0.6218,
+      "step": 9515
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5241803902690056e-05,
+      "loss": 0.6474,
+      "step": 9516
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5240804873646649e-05,
+      "loss": 0.6628,
+      "step": 9517
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5239805772485157e-05,
+      "loss": 0.6352,
+      "step": 9518
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5238806599219337e-05,
+      "loss": 0.6663,
+      "step": 9519
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5237807353862933e-05,
+      "loss": 0.66,
+      "step": 9520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5236808036429698e-05,
+      "loss": 0.7534,
+      "step": 9521
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5235808646933383e-05,
+      "loss": 0.6585,
+      "step": 9522
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5234809185387743e-05,
+      "loss": 0.6309,
+      "step": 9523
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5233809651806528e-05,
+      "loss": 0.7464,
+      "step": 9524
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523281004620349e-05,
+      "loss": 0.6604,
+      "step": 9525
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.523181036859239e-05,
+      "loss": 0.6689,
+      "step": 9526
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5230810618986982e-05,
+      "loss": 0.6146,
+      "step": 9527
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5229810797401026e-05,
+      "loss": 0.6223,
+      "step": 9528
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5228810903848276e-05,
+      "loss": 0.6598,
+      "step": 9529
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5227810938342493e-05,
+      "loss": 0.577,
+      "step": 9530
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522681090089744e-05,
+      "loss": 0.6752,
+      "step": 9531
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5225810791526873e-05,
+      "loss": 0.5726,
+      "step": 9532
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5224810610244563e-05,
+      "loss": 0.5788,
+      "step": 9533
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5223810357064261e-05,
+      "loss": 0.6174,
+      "step": 9534
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5222810031999746e-05,
+      "loss": 0.5957,
+      "step": 9535
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522180963506477e-05,
+      "loss": 0.6191,
+      "step": 9536
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5220809166273104e-05,
+      "loss": 0.5632,
+      "step": 9537
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5219808625638521e-05,
+      "loss": 0.6511,
+      "step": 9538
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5218808013174784e-05,
+      "loss": 0.6392,
+      "step": 9539
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5217807328895663e-05,
+      "loss": 0.6625,
+      "step": 9540
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5216806572814929e-05,
+      "loss": 0.5705,
+      "step": 9541
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.521580574494635e-05,
+      "loss": 0.5929,
+      "step": 9542
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.52148048453037e-05,
+      "loss": 0.7026,
+      "step": 9543
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.521380387390076e-05,
+      "loss": 0.6675,
+      "step": 9544
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5212802830751292e-05,
+      "loss": 0.6938,
+      "step": 9545
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5211801715869077e-05,
+      "loss": 0.6351,
+      "step": 9546
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5210800529267893e-05,
+      "loss": 0.6495,
+      "step": 9547
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5209799270961512e-05,
+      "loss": 0.6109,
+      "step": 9548
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5208797940963718e-05,
+      "loss": 0.5397,
+      "step": 9549
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5207796539288286e-05,
+      "loss": 0.6344,
+      "step": 9550
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5206795065948999e-05,
+      "loss": 0.6605,
+      "step": 9551
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5205793520959634e-05,
+      "loss": 0.6705,
+      "step": 9552
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.520479190433398e-05,
+      "loss": 0.671,
+      "step": 9553
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5203790216085814e-05,
+      "loss": 0.6929,
+      "step": 9554
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.520278845622892e-05,
+      "loss": 0.6202,
+      "step": 9555
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5201786624777088e-05,
+      "loss": 0.6423,
+      "step": 9556
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5200784721744103e-05,
+      "loss": 0.6333,
+      "step": 9557
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5199782747143745e-05,
+      "loss": 0.576,
+      "step": 9558
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5198780700989813e-05,
+      "loss": 0.6299,
+      "step": 9559
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5197778583296088e-05,
+      "loss": 0.6174,
+      "step": 9560
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519677639407636e-05,
+      "loss": 0.6564,
+      "step": 9561
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5195774133344428e-05,
+      "loss": 0.7024,
+      "step": 9562
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5194771801114074e-05,
+      "loss": 0.6291,
+      "step": 9563
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5193769397399099e-05,
+      "loss": 0.5927,
+      "step": 9564
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519276692221329e-05,
+      "loss": 0.6816,
+      "step": 9565
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5191764375570449e-05,
+      "loss": 0.6743,
+      "step": 9566
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5190761757484365e-05,
+      "loss": 0.5558,
+      "step": 9567
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5189759067968838e-05,
+      "loss": 0.6829,
+      "step": 9568
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.518875630703767e-05,
+      "loss": 0.6735,
+      "step": 9569
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5187753474704652e-05,
+      "loss": 0.5978,
+      "step": 9570
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.518675057098359e-05,
+      "loss": 0.6204,
+      "step": 9571
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5185747595888281e-05,
+      "loss": 0.6074,
+      "step": 9572
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5184744549432533e-05,
+      "loss": 0.6286,
+      "step": 9573
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5183741431630138e-05,
+      "loss": 0.7292,
+      "step": 9574
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5182738242494913e-05,
+      "loss": 0.625,
+      "step": 9575
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5181734982040648e-05,
+      "loss": 0.6459,
+      "step": 9576
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5180731650281161e-05,
+      "loss": 0.6466,
+      "step": 9577
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5179728247230255e-05,
+      "loss": 0.6394,
+      "step": 9578
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5178724772901733e-05,
+      "loss": 0.6369,
+      "step": 9579
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5177721227309412e-05,
+      "loss": 0.6294,
+      "step": 9580
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5176717610467096e-05,
+      "loss": 0.5882,
+      "step": 9581
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5175713922388598e-05,
+      "loss": 0.6777,
+      "step": 9582
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5174710163087727e-05,
+      "loss": 0.5945,
+      "step": 9583
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5173706332578298e-05,
+      "loss": 0.6443,
+      "step": 9584
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5172702430874122e-05,
+      "loss": 0.5849,
+      "step": 9585
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5171698457989017e-05,
+      "loss": 0.7117,
+      "step": 9586
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5170694413936798e-05,
+      "loss": 0.643,
+      "step": 9587
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5169690298731277e-05,
+      "loss": 0.6371,
+      "step": 9588
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5168686112386279e-05,
+      "loss": 0.6355,
+      "step": 9589
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5167681854915617e-05,
+      "loss": 0.6739,
+      "step": 9590
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5166677526333113e-05,
+      "loss": 0.6316,
+      "step": 9591
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5165673126652585e-05,
+      "loss": 0.594,
+      "step": 9592
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5164668655887855e-05,
+      "loss": 0.59,
+      "step": 9593
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5163664114052748e-05,
+      "loss": 0.6263,
+      "step": 9594
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5162659501161083e-05,
+      "loss": 0.5923,
+      "step": 9595
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.516165481722669e-05,
+      "loss": 0.6114,
+      "step": 9596
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5160650062263392e-05,
+      "loss": 0.6584,
+      "step": 9597
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.515964523628501e-05,
+      "loss": 0.6945,
+      "step": 9598
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5158640339305376e-05,
+      "loss": 0.6824,
+      "step": 9599
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5157635371338324e-05,
+      "loss": 0.7151,
+      "step": 9600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5156630332397675e-05,
+      "loss": 0.6768,
+      "step": 9601
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.515562522249726e-05,
+      "loss": 0.5995,
+      "step": 9602
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5154620041650914e-05,
+      "loss": 0.6365,
+      "step": 9603
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5153614789872464e-05,
+      "loss": 0.6779,
+      "step": 9604
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5152609467175748e-05,
+      "loss": 0.6619,
+      "step": 9605
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5151604073574598e-05,
+      "loss": 0.6564,
+      "step": 9606
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5150598609082849e-05,
+      "loss": 0.604,
+      "step": 9607
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5149593073714339e-05,
+      "loss": 0.6571,
+      "step": 9608
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5148587467482903e-05,
+      "loss": 0.5737,
+      "step": 9609
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5147581790402374e-05,
+      "loss": 0.6719,
+      "step": 9610
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5146576042486604e-05,
+      "loss": 0.7236,
+      "step": 9611
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5145570223749424e-05,
+      "loss": 0.7032,
+      "step": 9612
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5144564334204675e-05,
+      "loss": 0.6203,
+      "step": 9613
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5143558373866198e-05,
+      "loss": 0.6558,
+      "step": 9614
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5142552342747843e-05,
+      "loss": 0.6043,
+      "step": 9615
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5141546240863446e-05,
+      "loss": 0.6308,
+      "step": 9616
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5140540068226857e-05,
+      "loss": 0.6154,
+      "step": 9617
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5139533824851917e-05,
+      "loss": 0.6847,
+      "step": 9618
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5138527510752476e-05,
+      "loss": 0.5855,
+      "step": 9619
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5137521125942384e-05,
+      "loss": 0.6262,
+      "step": 9620
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5136514670435485e-05,
+      "loss": 0.5717,
+      "step": 9621
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5135508144245635e-05,
+      "loss": 0.6514,
+      "step": 9622
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5134501547386675e-05,
+      "loss": 0.5742,
+      "step": 9623
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5133494879872464e-05,
+      "loss": 0.6838,
+      "step": 9624
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5132488141716852e-05,
+      "loss": 0.5873,
+      "step": 9625
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5131481332933695e-05,
+      "loss": 0.6587,
+      "step": 9626
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5130474453536845e-05,
+      "loss": 0.6297,
+      "step": 9627
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5129467503540158e-05,
+      "loss": 0.6465,
+      "step": 9628
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512846048295749e-05,
+      "loss": 0.7188,
+      "step": 9629
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51274533918027e-05,
+      "loss": 0.6732,
+      "step": 9630
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5126446230089648e-05,
+      "loss": 0.6965,
+      "step": 9631
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5125438997832191e-05,
+      "loss": 0.6772,
+      "step": 9632
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5124431695044193e-05,
+      "loss": 0.6056,
+      "step": 9633
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5123424321739508e-05,
+      "loss": 0.7201,
+      "step": 9634
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5122416877932003e-05,
+      "loss": 0.6305,
+      "step": 9635
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5121409363635539e-05,
+      "loss": 0.6415,
+      "step": 9636
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5120401778863984e-05,
+      "loss": 0.6049,
+      "step": 9637
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5119394123631201e-05,
+      "loss": 0.6973,
+      "step": 9638
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5118386397951057e-05,
+      "loss": 0.6147,
+      "step": 9639
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5117378601837418e-05,
+      "loss": 0.5779,
+      "step": 9640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5116370735304153e-05,
+      "loss": 0.5127,
+      "step": 9641
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5115362798365131e-05,
+      "loss": 0.7516,
+      "step": 9642
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5114354791034225e-05,
+      "loss": 0.5915,
+      "step": 9643
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51133467133253e-05,
+      "loss": 0.6606,
+      "step": 9644
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5112338565252231e-05,
+      "loss": 0.6387,
+      "step": 9645
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5111330346828891e-05,
+      "loss": 0.6708,
+      "step": 9646
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5110322058069157e-05,
+      "loss": 0.5202,
+      "step": 9647
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5109313698986899e-05,
+      "loss": 0.6389,
+      "step": 9648
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5108305269595997e-05,
+      "loss": 0.6994,
+      "step": 9649
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5107296769910322e-05,
+      "loss": 0.6073,
+      "step": 9650
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5106288199943762e-05,
+      "loss": 0.6534,
+      "step": 9651
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5105279559710185e-05,
+      "loss": 0.6906,
+      "step": 9652
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5104270849223476e-05,
+      "loss": 0.531,
+      "step": 9653
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5103262068497516e-05,
+      "loss": 0.6773,
+      "step": 9654
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5102253217546185e-05,
+      "loss": 0.6579,
+      "step": 9655
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5101244296383368e-05,
+      "loss": 0.5719,
+      "step": 9656
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5100235305022945e-05,
+      "loss": 0.6387,
+      "step": 9657
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5099226243478805e-05,
+      "loss": 0.6388,
+      "step": 9658
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5098217111764827e-05,
+      "loss": 0.6897,
+      "step": 9659
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5097207909894907e-05,
+      "loss": 0.7409,
+      "step": 9660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5096198637882924e-05,
+      "loss": 0.6067,
+      "step": 9661
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5095189295742772e-05,
+      "loss": 0.6942,
+      "step": 9662
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5094179883488335e-05,
+      "loss": 0.6974,
+      "step": 9663
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5093170401133512e-05,
+      "loss": 0.6422,
+      "step": 9664
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5092160848692183e-05,
+      "loss": 0.5857,
+      "step": 9665
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5091151226178253e-05,
+      "loss": 0.6457,
+      "step": 9666
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5090141533605605e-05,
+      "loss": 0.5526,
+      "step": 9667
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5089131770988134e-05,
+      "loss": 0.6158,
+      "step": 9668
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5088121938339741e-05,
+      "loss": 0.6274,
+      "step": 9669
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5087112035674318e-05,
+      "loss": 0.7006,
+      "step": 9670
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5086102063005768e-05,
+      "loss": 0.5789,
+      "step": 9671
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508509202034798e-05,
+      "loss": 0.6048,
+      "step": 9672
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508408190771486e-05,
+      "loss": 0.6432,
+      "step": 9673
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5083071725120304e-05,
+      "loss": 0.6205,
+      "step": 9674
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5082061472578215e-05,
+      "loss": 0.5547,
+      "step": 9675
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5081051150102495e-05,
+      "loss": 0.5925,
+      "step": 9676
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5080040757707045e-05,
+      "loss": 0.713,
+      "step": 9677
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5079030295405774e-05,
+      "loss": 0.7478,
+      "step": 9678
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5078019763212582e-05,
+      "loss": 0.6271,
+      "step": 9679
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5077009161141375e-05,
+      "loss": 0.5916,
+      "step": 9680
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5075998489206063e-05,
+      "loss": 0.6642,
+      "step": 9681
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5074987747420551e-05,
+      "loss": 0.606,
+      "step": 9682
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5073976935798747e-05,
+      "loss": 0.6476,
+      "step": 9683
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5072966054354561e-05,
+      "loss": 0.6042,
+      "step": 9684
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5071955103101905e-05,
+      "loss": 0.6278,
+      "step": 9685
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5070944082054693e-05,
+      "loss": 0.6377,
+      "step": 9686
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5069932991226832e-05,
+      "loss": 0.5898,
+      "step": 9687
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5068921830632238e-05,
+      "loss": 0.5976,
+      "step": 9688
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5067910600284827e-05,
+      "loss": 0.6374,
+      "step": 9689
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5066899300198512e-05,
+      "loss": 0.6521,
+      "step": 9690
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5065887930387214e-05,
+      "loss": 0.7042,
+      "step": 9691
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5064876490864842e-05,
+      "loss": 0.6026,
+      "step": 9692
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5063864981645322e-05,
+      "loss": 0.6493,
+      "step": 9693
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506285340274257e-05,
+      "loss": 0.6063,
+      "step": 9694
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5061841754170505e-05,
+      "loss": 0.6822,
+      "step": 9695
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506083003594305e-05,
+      "loss": 0.6735,
+      "step": 9696
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.505981824807413e-05,
+      "loss": 0.5909,
+      "step": 9697
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.505880639057766e-05,
+      "loss": 0.6432,
+      "step": 9698
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5057794463467572e-05,
+      "loss": 0.6986,
+      "step": 9699
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5056782466757787e-05,
+      "loss": 0.5967,
+      "step": 9700
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5055770400462233e-05,
+      "loss": 0.6086,
+      "step": 9701
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5054758264594838e-05,
+      "loss": 0.6399,
+      "step": 9702
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5053746059169522e-05,
+      "loss": 0.5448,
+      "step": 9703
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5052733784200224e-05,
+      "loss": 0.617,
+      "step": 9704
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5051721439700869e-05,
+      "loss": 0.6325,
+      "step": 9705
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5050709025685387e-05,
+      "loss": 0.6129,
+      "step": 9706
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.504969654216771e-05,
+      "loss": 0.5811,
+      "step": 9707
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5048683989161771e-05,
+      "loss": 0.6997,
+      "step": 9708
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5047671366681509e-05,
+      "loss": 0.6605,
+      "step": 9709
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.504665867474085e-05,
+      "loss": 0.5644,
+      "step": 9710
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5045645913353735e-05,
+      "loss": 0.6857,
+      "step": 9711
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5044633082534098e-05,
+      "loss": 0.6172,
+      "step": 9712
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5043620182295877e-05,
+      "loss": 0.6252,
+      "step": 9713
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5042607212653008e-05,
+      "loss": 0.6881,
+      "step": 9714
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5041594173619439e-05,
+      "loss": 0.6468,
+      "step": 9715
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5040581065209098e-05,
+      "loss": 0.6646,
+      "step": 9716
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5039567887435938e-05,
+      "loss": 0.6442,
+      "step": 9717
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5038554640313892e-05,
+      "loss": 0.6114,
+      "step": 9718
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5037541323856906e-05,
+      "loss": 0.6034,
+      "step": 9719
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5036527938078926e-05,
+      "loss": 0.6195,
+      "step": 9720
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5035514482993898e-05,
+      "loss": 0.6091,
+      "step": 9721
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5034500958615765e-05,
+      "loss": 0.615,
+      "step": 9722
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5033487364958474e-05,
+      "loss": 0.5604,
+      "step": 9723
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5032473702035977e-05,
+      "loss": 0.5958,
+      "step": 9724
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5031459969862216e-05,
+      "loss": 0.626,
+      "step": 9725
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.503044616845115e-05,
+      "loss": 0.5803,
+      "step": 9726
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.502943229781672e-05,
+      "loss": 0.6807,
+      "step": 9727
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5028418357972882e-05,
+      "loss": 0.6418,
+      "step": 9728
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5027404348933593e-05,
+      "loss": 0.5945,
+      "step": 9729
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5026390270712799e-05,
+      "loss": 0.6839,
+      "step": 9730
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.502537612332446e-05,
+      "loss": 0.6824,
+      "step": 9731
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5024361906782529e-05,
+      "loss": 0.6995,
+      "step": 9732
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5023347621100969e-05,
+      "loss": 0.6817,
+      "step": 9733
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5022333266293725e-05,
+      "loss": 0.6507,
+      "step": 9734
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5021318842374763e-05,
+      "loss": 0.6645,
+      "step": 9735
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5020304349358043e-05,
+      "loss": 0.5358,
+      "step": 9736
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5019289787257526e-05,
+      "loss": 0.6579,
+      "step": 9737
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5018275156087167e-05,
+      "loss": 0.6758,
+      "step": 9738
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5017260455860933e-05,
+      "loss": 0.6106,
+      "step": 9739
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.501624568659279e-05,
+      "loss": 0.6406,
+      "step": 9740
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5015230848296695e-05,
+      "loss": 0.6397,
+      "step": 9741
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5014215940986621e-05,
+      "loss": 0.605,
+      "step": 9742
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5013200964676524e-05,
+      "loss": 0.594,
+      "step": 9743
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5012185919380382e-05,
+      "loss": 0.7049,
+      "step": 9744
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5011170805112154e-05,
+      "loss": 0.6821,
+      "step": 9745
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5010155621885814e-05,
+      "loss": 0.6627,
+      "step": 9746
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5009140369715332e-05,
+      "loss": 0.7009,
+      "step": 9747
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5008125048614675e-05,
+      "loss": 0.563,
+      "step": 9748
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5007109658597819e-05,
+      "loss": 0.6778,
+      "step": 9749
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.500609419967873e-05,
+      "loss": 0.6163,
+      "step": 9750
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5005078671871388e-05,
+      "loss": 0.6203,
+      "step": 9751
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5004063075189768e-05,
+      "loss": 0.6739,
+      "step": 9752
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5003047409647841e-05,
+      "loss": 0.733,
+      "step": 9753
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5002031675259585e-05,
+      "loss": 0.6559,
+      "step": 9754
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.500101587203898e-05,
+      "loss": 0.6872,
+      "step": 9755
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.6675,
+      "step": 9756
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499898405915663e-05,
+      "loss": 0.5367,
+      "step": 9757
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4997968049522844e-05,
+      "loss": 0.6534,
+      "step": 9758
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4996951971112628e-05,
+      "loss": 0.6347,
+      "step": 9759
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4995935823939962e-05,
+      "loss": 0.5488,
+      "step": 9760
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4994919608018828e-05,
+      "loss": 0.6315,
+      "step": 9761
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4993903323363212e-05,
+      "loss": 0.6293,
+      "step": 9762
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4992886969987096e-05,
+      "loss": 0.7281,
+      "step": 9763
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4991870547904472e-05,
+      "loss": 0.5722,
+      "step": 9764
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4990854057129321e-05,
+      "loss": 0.6881,
+      "step": 9765
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4989837497675632e-05,
+      "loss": 0.5894,
+      "step": 9766
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4988820869557397e-05,
+      "loss": 0.6554,
+      "step": 9767
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.49878041727886e-05,
+      "loss": 0.5688,
+      "step": 9768
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4986787407383237e-05,
+      "loss": 0.6927,
+      "step": 9769
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4985770573355297e-05,
+      "loss": 0.58,
+      "step": 9770
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4984753670718774e-05,
+      "loss": 0.7168,
+      "step": 9771
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4983736699487659e-05,
+      "loss": 0.6392,
+      "step": 9772
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4982719659675948e-05,
+      "loss": 0.634,
+      "step": 9773
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4981702551297636e-05,
+      "loss": 0.6907,
+      "step": 9774
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4980685374366722e-05,
+      "loss": 0.6992,
+      "step": 9775
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.49796681288972e-05,
+      "loss": 0.6362,
+      "step": 9776
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4978650814903066e-05,
+      "loss": 0.6188,
+      "step": 9777
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4977633432398326e-05,
+      "loss": 0.6537,
+      "step": 9778
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4976615981396973e-05,
+      "loss": 0.6219,
+      "step": 9779
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4975598461913015e-05,
+      "loss": 0.5928,
+      "step": 9780
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4974580873960447e-05,
+      "loss": 0.6218,
+      "step": 9781
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4973563217553278e-05,
+      "loss": 0.6522,
+      "step": 9782
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4972545492705505e-05,
+      "loss": 0.6823,
+      "step": 9783
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.497152769943114e-05,
+      "loss": 0.6458,
+      "step": 9784
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4970509837744185e-05,
+      "loss": 0.6558,
+      "step": 9785
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4969491907658644e-05,
+      "loss": 0.6676,
+      "step": 9786
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4968473909188532e-05,
+      "loss": 0.6007,
+      "step": 9787
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4967455842347851e-05,
+      "loss": 0.616,
+      "step": 9788
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4966437707150617e-05,
+      "loss": 0.6823,
+      "step": 9789
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.496541950361083e-05,
+      "loss": 0.6656,
+      "step": 9790
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4964401231742512e-05,
+      "loss": 0.6174,
+      "step": 9791
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4963382891559667e-05,
+      "loss": 0.6975,
+      "step": 9792
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4962364483076318e-05,
+      "loss": 0.6267,
+      "step": 9793
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4961346006306468e-05,
+      "loss": 0.6416,
+      "step": 9794
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4960327461264141e-05,
+      "loss": 0.6528,
+      "step": 9795
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4959308847963348e-05,
+      "loss": 0.6206,
+      "step": 9796
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4958290166418106e-05,
+      "loss": 0.5919,
+      "step": 9797
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4957271416642438e-05,
+      "loss": 0.6128,
+      "step": 9798
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4956252598650356e-05,
+      "loss": 0.5799,
+      "step": 9799
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4955233712455887e-05,
+      "loss": 0.6939,
+      "step": 9800
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4954214758073044e-05,
+      "loss": 0.6321,
+      "step": 9801
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4953195735515858e-05,
+      "loss": 0.6535,
+      "step": 9802
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4952176644798342e-05,
+      "loss": 0.6101,
+      "step": 9803
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4951157485934527e-05,
+      "loss": 0.6428,
+      "step": 9804
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4950138258938431e-05,
+      "loss": 0.6691,
+      "step": 9805
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4949118963824086e-05,
+      "loss": 0.6457,
+      "step": 9806
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4948099600605515e-05,
+      "loss": 0.673,
+      "step": 9807
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4947080169296744e-05,
+      "loss": 0.6044,
+      "step": 9808
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4946060669911803e-05,
+      "loss": 0.6237,
+      "step": 9809
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.494504110246472e-05,
+      "loss": 0.6175,
+      "step": 9810
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4944021466969532e-05,
+      "loss": 0.6646,
+      "step": 9811
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4943001763440262e-05,
+      "loss": 0.6168,
+      "step": 9812
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4941981991890941e-05,
+      "loss": 0.6052,
+      "step": 9813
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4940962152335607e-05,
+      "loss": 0.6131,
+      "step": 9814
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4939942244788294e-05,
+      "loss": 0.6741,
+      "step": 9815
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4938922269263035e-05,
+      "loss": 0.6686,
+      "step": 9816
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4937902225773864e-05,
+      "loss": 0.6783,
+      "step": 9817
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4936882114334821e-05,
+      "loss": 0.557,
+      "step": 9818
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4935861934959941e-05,
+      "loss": 0.6191,
+      "step": 9819
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4934841687663266e-05,
+      "loss": 0.596,
+      "step": 9820
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493382137245883e-05,
+      "loss": 0.6171,
+      "step": 9821
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493280098936068e-05,
+      "loss": 0.6137,
+      "step": 9822
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4931780538382852e-05,
+      "loss": 0.5919,
+      "step": 9823
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4930760019539392e-05,
+      "loss": 0.6447,
+      "step": 9824
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.492973943284434e-05,
+      "loss": 0.7706,
+      "step": 9825
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4928718778311742e-05,
+      "loss": 0.6208,
+      "step": 9826
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4927698055955643e-05,
+      "loss": 0.6395,
+      "step": 9827
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4926677265790088e-05,
+      "loss": 0.6186,
+      "step": 9828
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4925656407829125e-05,
+      "loss": 0.6227,
+      "step": 9829
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4924635482086803e-05,
+      "loss": 0.63,
+      "step": 9830
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.492361448857717e-05,
+      "loss": 0.5564,
+      "step": 9831
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4922593427314274e-05,
+      "loss": 0.5394,
+      "step": 9832
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4921572298312168e-05,
+      "loss": 0.6418,
+      "step": 9833
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4920551101584902e-05,
+      "loss": 0.6666,
+      "step": 9834
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4919529837146529e-05,
+      "loss": 0.6015,
+      "step": 9835
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4918508505011104e-05,
+      "loss": 0.6314,
+      "step": 9836
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4917487105192681e-05,
+      "loss": 0.644,
+      "step": 9837
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491646563770531e-05,
+      "loss": 0.6276,
+      "step": 9838
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4915444102563057e-05,
+      "loss": 0.7078,
+      "step": 9839
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4914422499779972e-05,
+      "loss": 0.6181,
+      "step": 9840
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4913400829370113e-05,
+      "loss": 0.6807,
+      "step": 9841
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4912379091347547e-05,
+      "loss": 0.6162,
+      "step": 9842
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4911357285726322e-05,
+      "loss": 0.6738,
+      "step": 9843
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491033541252051e-05,
+      "loss": 0.6232,
+      "step": 9844
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4909313471744167e-05,
+      "loss": 0.5803,
+      "step": 9845
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4908291463411356e-05,
+      "loss": 0.6088,
+      "step": 9846
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4907269387536141e-05,
+      "loss": 0.5648,
+      "step": 9847
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4906247244132588e-05,
+      "loss": 0.7049,
+      "step": 9848
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4905225033214763e-05,
+      "loss": 0.6477,
+      "step": 9849
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4904202754796729e-05,
+      "loss": 0.6466,
+      "step": 9850
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4903180408892558e-05,
+      "loss": 0.6805,
+      "step": 9851
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4902157995516315e-05,
+      "loss": 0.6685,
+      "step": 9852
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4901135514682072e-05,
+      "loss": 0.7432,
+      "step": 9853
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4900112966403897e-05,
+      "loss": 0.5746,
+      "step": 9854
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4899090350695859e-05,
+      "loss": 0.6571,
+      "step": 9855
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4898067667572038e-05,
+      "loss": 0.632,
+      "step": 9856
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4897044917046499e-05,
+      "loss": 0.5811,
+      "step": 9857
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4896022099133322e-05,
+      "loss": 0.6177,
+      "step": 9858
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4894999213846576e-05,
+      "loss": 0.7104,
+      "step": 9859
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4893976261200342e-05,
+      "loss": 0.5912,
+      "step": 9860
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4892953241208692e-05,
+      "loss": 0.6821,
+      "step": 9861
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4891930153885708e-05,
+      "loss": 0.6539,
+      "step": 9862
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4890906999245464e-05,
+      "loss": 0.6829,
+      "step": 9863
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4889883777302044e-05,
+      "loss": 0.6222,
+      "step": 9864
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4888860488069528e-05,
+      "loss": 0.614,
+      "step": 9865
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4887837131561994e-05,
+      "loss": 0.6581,
+      "step": 9866
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4886813707793528e-05,
+      "loss": 0.6521,
+      "step": 9867
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4885790216778208e-05,
+      "loss": 0.642,
+      "step": 9868
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4884766658530126e-05,
+      "loss": 0.624,
+      "step": 9869
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4883743033063357e-05,
+      "loss": 0.6162,
+      "step": 9870
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4882719340392e-05,
+      "loss": 0.6139,
+      "step": 9871
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4881695580530129e-05,
+      "loss": 0.6532,
+      "step": 9872
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.488067175349184e-05,
+      "loss": 0.647,
+      "step": 9873
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4879647859291217e-05,
+      "loss": 0.5813,
+      "step": 9874
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4878623897942353e-05,
+      "loss": 0.5938,
+      "step": 9875
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4877599869459337e-05,
+      "loss": 0.7085,
+      "step": 9876
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487657577385626e-05,
+      "loss": 0.6406,
+      "step": 9877
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4875551611147219e-05,
+      "loss": 0.6494,
+      "step": 9878
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48745273813463e-05,
+      "loss": 0.6278,
+      "step": 9879
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4873503084467603e-05,
+      "loss": 0.6448,
+      "step": 9880
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487247872052522e-05,
+      "loss": 0.6958,
+      "step": 9881
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4871454289533252e-05,
+      "loss": 0.6549,
+      "step": 9882
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4870429791505789e-05,
+      "loss": 0.5904,
+      "step": 9883
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4869405226456934e-05,
+      "loss": 0.633,
+      "step": 9884
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4868380594400785e-05,
+      "loss": 0.6815,
+      "step": 9885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4867355895351438e-05,
+      "loss": 0.6234,
+      "step": 9886
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4866331129323e-05,
+      "loss": 0.5732,
+      "step": 9887
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4865306296329567e-05,
+      "loss": 0.6518,
+      "step": 9888
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.486428139638525e-05,
+      "loss": 0.5687,
+      "step": 9889
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4863256429504141e-05,
+      "loss": 0.5945,
+      "step": 9890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4862231395700353e-05,
+      "loss": 0.6201,
+      "step": 9891
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4861206294987986e-05,
+      "loss": 0.6682,
+      "step": 9892
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4860181127381155e-05,
+      "loss": 0.6445,
+      "step": 9893
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4859155892893956e-05,
+      "loss": 0.6641,
+      "step": 9894
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4858130591540503e-05,
+      "loss": 0.6739,
+      "step": 9895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4857105223334905e-05,
+      "loss": 0.697,
+      "step": 9896
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485607978829127e-05,
+      "loss": 0.6599,
+      "step": 9897
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4855054286423714e-05,
+      "loss": 0.6517,
+      "step": 9898
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485402871774634e-05,
+      "loss": 0.5558,
+      "step": 9899
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4853003082273269e-05,
+      "loss": 0.6591,
+      "step": 9900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485197738001861e-05,
+      "loss": 0.6199,
+      "step": 9901
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485095161099648e-05,
+      "loss": 0.6662,
+      "step": 9902
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4849925775220993e-05,
+      "loss": 0.6485,
+      "step": 9903
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4848899872706267e-05,
+      "loss": 0.5628,
+      "step": 9904
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4847873903466414e-05,
+      "loss": 0.5869,
+      "step": 9905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484684786751556e-05,
+      "loss": 0.6839,
+      "step": 9906
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484582176486782e-05,
+      "loss": 0.6982,
+      "step": 9907
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4844795595537314e-05,
+      "loss": 0.6767,
+      "step": 9908
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4843769359538164e-05,
+      "loss": 0.6414,
+      "step": 9909
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484274305688449e-05,
+      "loss": 0.6562,
+      "step": 9910
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4841716687590419e-05,
+      "loss": 0.6602,
+      "step": 9911
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4840690251670071e-05,
+      "loss": 0.5657,
+      "step": 9912
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4839663749137572e-05,
+      "loss": 0.607,
+      "step": 9913
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4838637180007048e-05,
+      "loss": 0.5657,
+      "step": 9914
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4837610544292622e-05,
+      "loss": 0.6888,
+      "step": 9915
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4836583842008425e-05,
+      "loss": 0.6545,
+      "step": 9916
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4835557073168584e-05,
+      "loss": 0.5958,
+      "step": 9917
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4834530237787233e-05,
+      "loss": 0.6459,
+      "step": 9918
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4833503335878494e-05,
+      "loss": 0.6658,
+      "step": 9919
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4832476367456503e-05,
+      "loss": 0.6611,
+      "step": 9920
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.483144933253539e-05,
+      "loss": 0.6345,
+      "step": 9921
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.483042223112929e-05,
+      "loss": 0.7192,
+      "step": 9922
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4829395063252336e-05,
+      "loss": 0.6503,
+      "step": 9923
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4828367828918663e-05,
+      "loss": 0.6268,
+      "step": 9924
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4827340528142404e-05,
+      "loss": 0.5981,
+      "step": 9925
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48263131609377e-05,
+      "loss": 0.6067,
+      "step": 9926
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4825285727318685e-05,
+      "loss": 0.6421,
+      "step": 9927
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4824258227299497e-05,
+      "loss": 0.6673,
+      "step": 9928
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4823230660894282e-05,
+      "loss": 0.6336,
+      "step": 9929
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.482220302811717e-05,
+      "loss": 0.6388,
+      "step": 9930
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4821175328982309e-05,
+      "loss": 0.5913,
+      "step": 9931
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4820147563503838e-05,
+      "loss": 0.5769,
+      "step": 9932
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4819119731695903e-05,
+      "loss": 0.6496,
+      "step": 9933
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4818091833572643e-05,
+      "loss": 0.6617,
+      "step": 9934
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.481706386914821e-05,
+      "loss": 0.6893,
+      "step": 9935
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4816035838436743e-05,
+      "loss": 0.5813,
+      "step": 9936
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.481500774145239e-05,
+      "loss": 0.672,
+      "step": 9937
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4813979578209303e-05,
+      "loss": 0.6235,
+      "step": 9938
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4812951348721623e-05,
+      "loss": 0.7209,
+      "step": 9939
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4811923053003508e-05,
+      "loss": 0.6178,
+      "step": 9940
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48108946910691e-05,
+      "loss": 0.6546,
+      "step": 9941
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4809866262932558e-05,
+      "loss": 0.5832,
+      "step": 9942
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4808837768608024e-05,
+      "loss": 0.6497,
+      "step": 9943
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.480780920810966e-05,
+      "loss": 0.7058,
+      "step": 9944
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4806780581451619e-05,
+      "loss": 0.6734,
+      "step": 9945
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.480575188864805e-05,
+      "loss": 0.626,
+      "step": 9946
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4804723129713115e-05,
+      "loss": 0.6513,
+      "step": 9947
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4803694304660965e-05,
+      "loss": 0.6262,
+      "step": 9948
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4802665413505763e-05,
+      "loss": 0.6447,
+      "step": 9949
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4801636456261665e-05,
+      "loss": 0.616,
+      "step": 9950
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.480060743294283e-05,
+      "loss": 0.638,
+      "step": 9951
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4799578343563417e-05,
+      "loss": 0.6553,
+      "step": 9952
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4798549188137588e-05,
+      "loss": 0.6801,
+      "step": 9953
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4797519966679508e-05,
+      "loss": 0.6817,
+      "step": 9954
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4796490679203337e-05,
+      "loss": 0.6105,
+      "step": 9955
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.479546132572324e-05,
+      "loss": 0.5791,
+      "step": 9956
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4794431906253381e-05,
+      "loss": 0.6082,
+      "step": 9957
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4793402420807927e-05,
+      "loss": 0.6624,
+      "step": 9958
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4792372869401042e-05,
+      "loss": 0.6536,
+      "step": 9959
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.47913432520469e-05,
+      "loss": 0.686,
+      "step": 9960
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4790313568759658e-05,
+      "loss": 0.6479,
+      "step": 9961
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4789283819553497e-05,
+      "loss": 0.6146,
+      "step": 9962
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4788254004442582e-05,
+      "loss": 0.6785,
+      "step": 9963
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4787224123441081e-05,
+      "loss": 0.6016,
+      "step": 9964
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4786194176563174e-05,
+      "loss": 0.6733,
+      "step": 9965
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4785164163823026e-05,
+      "loss": 0.5983,
+      "step": 9966
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4784134085234817e-05,
+      "loss": 0.5813,
+      "step": 9967
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4783103940812719e-05,
+      "loss": 0.6572,
+      "step": 9968
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.478207373057091e-05,
+      "loss": 0.5935,
+      "step": 9969
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.478104345452356e-05,
+      "loss": 0.5613,
+      "step": 9970
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4780013112684855e-05,
+      "loss": 0.6106,
+      "step": 9971
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4778982705068968e-05,
+      "loss": 0.6412,
+      "step": 9972
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4777952231690078e-05,
+      "loss": 0.6954,
+      "step": 9973
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4776921692562369e-05,
+      "loss": 0.6423,
+      "step": 9974
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4775891087700018e-05,
+      "loss": 0.691,
+      "step": 9975
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4774860417117211e-05,
+      "loss": 0.6636,
+      "step": 9976
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4773829680828129e-05,
+      "loss": 0.658,
+      "step": 9977
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4772798878846957e-05,
+      "loss": 0.6602,
+      "step": 9978
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4771768011187878e-05,
+      "loss": 0.6121,
+      "step": 9979
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4770737077865078e-05,
+      "loss": 0.7028,
+      "step": 9980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4769706078892742e-05,
+      "loss": 0.5526,
+      "step": 9981
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4768675014285063e-05,
+      "loss": 0.6779,
+      "step": 9982
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4767643884056222e-05,
+      "loss": 0.6672,
+      "step": 9983
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4766612688220414e-05,
+      "loss": 0.6801,
+      "step": 9984
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4765581426791826e-05,
+      "loss": 0.7051,
+      "step": 9985
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4764550099784651e-05,
+      "loss": 0.7493,
+      "step": 9986
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4763518707213083e-05,
+      "loss": 0.7315,
+      "step": 9987
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4762487249091306e-05,
+      "loss": 0.6188,
+      "step": 9988
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4761455725433525e-05,
+      "loss": 0.656,
+      "step": 9989
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4760424136253926e-05,
+      "loss": 0.5973,
+      "step": 9990
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475939248156671e-05,
+      "loss": 0.6081,
+      "step": 9991
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475836076138607e-05,
+      "loss": 0.6214,
+      "step": 9992
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4757328975726207e-05,
+      "loss": 0.6164,
+      "step": 9993
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4756297124601316e-05,
+      "loss": 0.7006,
+      "step": 9994
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4755265208025595e-05,
+      "loss": 0.6462,
+      "step": 9995
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4754233226013251e-05,
+      "loss": 0.6439,
+      "step": 9996
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4753201178578476e-05,
+      "loss": 0.6081,
+      "step": 9997
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475216906573548e-05,
+      "loss": 0.6533,
+      "step": 9998
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475113688749846e-05,
+      "loss": 0.5568,
+      "step": 9999
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4750104643881623e-05,
+      "loss": 0.6139,
+      "step": 10000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4749072334899168e-05,
+      "loss": 0.5641,
+      "step": 10001
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474803996056531e-05,
+      "loss": 0.6565,
+      "step": 10002
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4747007520894245e-05,
+      "loss": 0.7393,
+      "step": 10003
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474597501590019e-05,
+      "loss": 0.5897,
+      "step": 10004
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4744942445597349e-05,
+      "loss": 0.5983,
+      "step": 10005
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4743909809999926e-05,
+      "loss": 0.6817,
+      "step": 10006
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474287710912214e-05,
+      "loss": 0.607,
+      "step": 10007
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4741844342978196e-05,
+      "loss": 0.6078,
+      "step": 10008
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4740811511582308e-05,
+      "loss": 0.6906,
+      "step": 10009
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4739778614948685e-05,
+      "loss": 0.6057,
+      "step": 10010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4738745653091547e-05,
+      "loss": 0.6087,
+      "step": 10011
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4737712626025103e-05,
+      "loss": 0.5716,
+      "step": 10012
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4736679533763573e-05,
+      "loss": 0.6696,
+      "step": 10013
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4735646376321166e-05,
+      "loss": 0.7025,
+      "step": 10014
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4734613153712107e-05,
+      "loss": 0.6362,
+      "step": 10015
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4733579865950611e-05,
+      "loss": 0.7003,
+      "step": 10016
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4732546513050894e-05,
+      "loss": 0.6339,
+      "step": 10017
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4731513095027183e-05,
+      "loss": 0.6404,
+      "step": 10018
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.473047961189369e-05,
+      "loss": 0.62,
+      "step": 10019
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4729446063664645e-05,
+      "loss": 0.6546,
+      "step": 10020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4728412450354264e-05,
+      "loss": 0.5833,
+      "step": 10021
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4727378771976772e-05,
+      "loss": 0.6046,
+      "step": 10022
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4726345028546396e-05,
+      "loss": 0.558,
+      "step": 10023
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472531122007736e-05,
+      "loss": 0.6807,
+      "step": 10024
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4724277346583889e-05,
+      "loss": 0.5848,
+      "step": 10025
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472324340808021e-05,
+      "loss": 0.6763,
+      "step": 10026
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4722209404580551e-05,
+      "loss": 0.5669,
+      "step": 10027
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4721175336099142e-05,
+      "loss": 0.7008,
+      "step": 10028
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4720141202650214e-05,
+      "loss": 0.6424,
+      "step": 10029
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4719107004247992e-05,
+      "loss": 0.6778,
+      "step": 10030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4718072740906713e-05,
+      "loss": 0.6508,
+      "step": 10031
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4717038412640607e-05,
+      "loss": 0.6089,
+      "step": 10032
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.471600401946391e-05,
+      "loss": 0.6472,
+      "step": 10033
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4714969561390851e-05,
+      "loss": 0.6335,
+      "step": 10034
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4713935038435666e-05,
+      "loss": 0.6676,
+      "step": 10035
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4712900450612598e-05,
+      "loss": 0.685,
+      "step": 10036
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4711865797935873e-05,
+      "loss": 0.5952,
+      "step": 10037
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4710831080419739e-05,
+      "loss": 0.5417,
+      "step": 10038
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4709796298078428e-05,
+      "loss": 0.5946,
+      "step": 10039
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4708761450926184e-05,
+      "loss": 0.6699,
+      "step": 10040
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.470772653897724e-05,
+      "loss": 0.6404,
+      "step": 10041
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4706691562245847e-05,
+      "loss": 0.6647,
+      "step": 10042
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.470565652074624e-05,
+      "loss": 0.6101,
+      "step": 10043
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4704621414492666e-05,
+      "loss": 0.6163,
+      "step": 10044
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4703586243499366e-05,
+      "loss": 0.6995,
+      "step": 10045
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4702551007780587e-05,
+      "loss": 0.5952,
+      "step": 10046
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4701515707350576e-05,
+      "loss": 0.6614,
+      "step": 10047
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4700480342223574e-05,
+      "loss": 0.6286,
+      "step": 10048
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4699444912413835e-05,
+      "loss": 0.6947,
+      "step": 10049
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4698409417935603e-05,
+      "loss": 0.5761,
+      "step": 10050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4697373858803133e-05,
+      "loss": 0.5914,
+      "step": 10051
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4696338235030668e-05,
+      "loss": 0.6821,
+      "step": 10052
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4695302546632462e-05,
+      "loss": 0.6342,
+      "step": 10053
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4694266793622766e-05,
+      "loss": 0.6783,
+      "step": 10054
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4693230976015837e-05,
+      "loss": 0.6281,
+      "step": 10055
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4692195093825927e-05,
+      "loss": 0.651,
+      "step": 10056
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4691159147067287e-05,
+      "loss": 0.66,
+      "step": 10057
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4690123135754175e-05,
+      "loss": 0.6733,
+      "step": 10058
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4689087059900848e-05,
+      "loss": 0.5628,
+      "step": 10059
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4688050919521563e-05,
+      "loss": 0.6302,
+      "step": 10060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4687014714630578e-05,
+      "loss": 0.59,
+      "step": 10061
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4685978445242153e-05,
+      "loss": 0.6625,
+      "step": 10062
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4684942111370545e-05,
+      "loss": 0.6581,
+      "step": 10063
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4683905713030018e-05,
+      "loss": 0.6739,
+      "step": 10064
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4682869250234835e-05,
+      "loss": 0.591,
+      "step": 10065
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4681832722999255e-05,
+      "loss": 0.6511,
+      "step": 10066
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4680796131337543e-05,
+      "loss": 0.6679,
+      "step": 10067
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4679759475263962e-05,
+      "loss": 0.7453,
+      "step": 10068
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.467872275479278e-05,
+      "loss": 0.7038,
+      "step": 10069
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.467768596993826e-05,
+      "loss": 0.712,
+      "step": 10070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4676649120714674e-05,
+      "loss": 0.6322,
+      "step": 10071
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4675612207136283e-05,
+      "loss": 0.6592,
+      "step": 10072
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4674575229217361e-05,
+      "loss": 0.5284,
+      "step": 10073
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4673538186972177e-05,
+      "loss": 0.6583,
+      "step": 10074
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4672501080415002e-05,
+      "loss": 0.726,
+      "step": 10075
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4671463909560107e-05,
+      "loss": 0.5914,
+      "step": 10076
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.467042667442176e-05,
+      "loss": 0.6703,
+      "step": 10077
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4669389375014243e-05,
+      "loss": 0.6319,
+      "step": 10078
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466835201135182e-05,
+      "loss": 0.6186,
+      "step": 10079
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4667314583448777e-05,
+      "loss": 0.6236,
+      "step": 10080
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466627709131938e-05,
+      "loss": 0.6247,
+      "step": 10081
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4665239534977912e-05,
+      "loss": 0.7149,
+      "step": 10082
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4664201914438647e-05,
+      "loss": 0.6425,
+      "step": 10083
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4663164229715868e-05,
+      "loss": 0.6186,
+      "step": 10084
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466212648082385e-05,
+      "loss": 0.6527,
+      "step": 10085
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4661088667776876e-05,
+      "loss": 0.636,
+      "step": 10086
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4660050790589227e-05,
+      "loss": 0.6466,
+      "step": 10087
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4659012849275183e-05,
+      "loss": 0.6844,
+      "step": 10088
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.465797484384903e-05,
+      "loss": 0.6189,
+      "step": 10089
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4656936774325047e-05,
+      "loss": 0.6411,
+      "step": 10090
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4655898640717525e-05,
+      "loss": 0.6991,
+      "step": 10091
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4654860443040746e-05,
+      "loss": 0.6191,
+      "step": 10092
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4653822181308997e-05,
+      "loss": 0.7146,
+      "step": 10093
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4652783855536565e-05,
+      "loss": 0.6424,
+      "step": 10094
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.465174546573774e-05,
+      "loss": 0.573,
+      "step": 10095
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4650707011926808e-05,
+      "loss": 0.6713,
+      "step": 10096
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4649668494118063e-05,
+      "loss": 0.6467,
+      "step": 10097
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4648629912325794e-05,
+      "loss": 0.6463,
+      "step": 10098
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4647591266564291e-05,
+      "loss": 0.6651,
+      "step": 10099
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464655255684785e-05,
+      "loss": 0.6269,
+      "step": 10100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4645513783190764e-05,
+      "loss": 0.6572,
+      "step": 10101
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4644474945607325e-05,
+      "loss": 0.6356,
+      "step": 10102
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4643436044111829e-05,
+      "loss": 0.6575,
+      "step": 10103
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4642397078718574e-05,
+      "loss": 0.6513,
+      "step": 10104
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4641358049441855e-05,
+      "loss": 0.6714,
+      "step": 10105
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4640318956295971e-05,
+      "loss": 0.6593,
+      "step": 10106
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4639279799295226e-05,
+      "loss": 0.5944,
+      "step": 10107
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4638240578453907e-05,
+      "loss": 0.5761,
+      "step": 10108
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4637201293786328e-05,
+      "loss": 0.6701,
+      "step": 10109
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.463616194530678e-05,
+      "loss": 0.6587,
+      "step": 10110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4635122533029573e-05,
+      "loss": 0.599,
+      "step": 10111
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4634083056969006e-05,
+      "loss": 0.5884,
+      "step": 10112
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4633043517139382e-05,
+      "loss": 0.6035,
+      "step": 10113
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4632003913555012e-05,
+      "loss": 0.6424,
+      "step": 10114
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4630964246230197e-05,
+      "loss": 0.6488,
+      "step": 10115
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4629924515179243e-05,
+      "loss": 0.6958,
+      "step": 10116
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4628884720416461e-05,
+      "loss": 0.6248,
+      "step": 10117
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.462784486195616e-05,
+      "loss": 0.6306,
+      "step": 10118
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4626804939812645e-05,
+      "loss": 0.6063,
+      "step": 10119
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4625764954000228e-05,
+      "loss": 0.6847,
+      "step": 10120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4624724904533221e-05,
+      "loss": 0.6846,
+      "step": 10121
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4623684791425939e-05,
+      "loss": 0.5964,
+      "step": 10122
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4622644614692687e-05,
+      "loss": 0.584,
+      "step": 10123
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4621604374347781e-05,
+      "loss": 0.6448,
+      "step": 10124
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4620564070405545e-05,
+      "loss": 0.6409,
+      "step": 10125
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4619523702880285e-05,
+      "loss": 0.7245,
+      "step": 10126
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461848327178632e-05,
+      "loss": 0.6389,
+      "step": 10127
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4617442777137965e-05,
+      "loss": 0.5848,
+      "step": 10128
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4616402218949541e-05,
+      "loss": 0.6415,
+      "step": 10129
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4615361597235366e-05,
+      "loss": 0.6961,
+      "step": 10130
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461432091200976e-05,
+      "loss": 0.5623,
+      "step": 10131
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4613280163287044e-05,
+      "loss": 0.6525,
+      "step": 10132
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461223935108154e-05,
+      "loss": 0.7035,
+      "step": 10133
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4611198475407572e-05,
+      "loss": 0.626,
+      "step": 10134
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4610157536279456e-05,
+      "loss": 0.6408,
+      "step": 10135
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4609116533711526e-05,
+      "loss": 0.6345,
+      "step": 10136
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4608075467718098e-05,
+      "loss": 0.5876,
+      "step": 10137
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4607034338313506e-05,
+      "loss": 0.6636,
+      "step": 10138
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4605993145512073e-05,
+      "loss": 0.6899,
+      "step": 10139
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4604951889328127e-05,
+      "loss": 0.685,
+      "step": 10140
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4603910569775997e-05,
+      "loss": 0.6159,
+      "step": 10141
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4602869186870012e-05,
+      "loss": 0.5625,
+      "step": 10142
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4601827740624501e-05,
+      "loss": 0.6257,
+      "step": 10143
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4600786231053796e-05,
+      "loss": 0.5875,
+      "step": 10144
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4599744658172234e-05,
+      "loss": 0.6968,
+      "step": 10145
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459870302199414e-05,
+      "loss": 0.6813,
+      "step": 10146
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4597661322533855e-05,
+      "loss": 0.6562,
+      "step": 10147
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4596619559805708e-05,
+      "loss": 0.6314,
+      "step": 10148
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4595577733824038e-05,
+      "loss": 0.6434,
+      "step": 10149
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4594535844603179e-05,
+      "loss": 0.6545,
+      "step": 10150
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4593493892157473e-05,
+      "loss": 0.6922,
+      "step": 10151
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4592451876501253e-05,
+      "loss": 0.5651,
+      "step": 10152
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4591409797648858e-05,
+      "loss": 0.6941,
+      "step": 10153
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459036765561463e-05,
+      "loss": 0.6246,
+      "step": 10154
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4589325450412912e-05,
+      "loss": 0.7076,
+      "step": 10155
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4588283182058042e-05,
+      "loss": 0.6342,
+      "step": 10156
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4587240850564364e-05,
+      "loss": 0.6738,
+      "step": 10157
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4586198455946222e-05,
+      "loss": 0.6176,
+      "step": 10158
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4585155998217956e-05,
+      "loss": 0.675,
+      "step": 10159
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4584113477393919e-05,
+      "loss": 0.6607,
+      "step": 10160
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4583070893488451e-05,
+      "loss": 0.727,
+      "step": 10161
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45820282465159e-05,
+      "loss": 0.6698,
+      "step": 10162
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4580985536490615e-05,
+      "loss": 0.6619,
+      "step": 10163
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4579942763426944e-05,
+      "loss": 0.6625,
+      "step": 10164
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4578899927339236e-05,
+      "loss": 0.6191,
+      "step": 10165
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4577857028241839e-05,
+      "loss": 0.6462,
+      "step": 10166
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4576814066149112e-05,
+      "loss": 0.6776,
+      "step": 10167
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4575771041075397e-05,
+      "loss": 0.6638,
+      "step": 10168
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4574727953035055e-05,
+      "loss": 0.6013,
+      "step": 10169
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4573684802042433e-05,
+      "loss": 0.571,
+      "step": 10170
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4572641588111891e-05,
+      "loss": 0.5673,
+      "step": 10171
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4571598311257783e-05,
+      "loss": 0.649,
+      "step": 10172
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4570554971494466e-05,
+      "loss": 0.582,
+      "step": 10173
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4569511568836299e-05,
+      "loss": 0.558,
+      "step": 10174
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4568468103297635e-05,
+      "loss": 0.6332,
+      "step": 10175
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4567424574892836e-05,
+      "loss": 0.6667,
+      "step": 10176
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4566380983636262e-05,
+      "loss": 0.5984,
+      "step": 10177
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456533732954227e-05,
+      "loss": 0.6208,
+      "step": 10178
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4564293612625229e-05,
+      "loss": 0.7096,
+      "step": 10179
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4563249832899495e-05,
+      "loss": 0.5682,
+      "step": 10180
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4562205990379434e-05,
+      "loss": 0.6497,
+      "step": 10181
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456116208507941e-05,
+      "loss": 0.6757,
+      "step": 10182
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456011811701379e-05,
+      "loss": 0.5683,
+      "step": 10183
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4559074086196935e-05,
+      "loss": 0.6782,
+      "step": 10184
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4558029992643216e-05,
+      "loss": 0.6903,
+      "step": 10185
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4556985836366999e-05,
+      "loss": 0.5653,
+      "step": 10186
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4555941617382653e-05,
+      "loss": 0.6388,
+      "step": 10187
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4554897335704548e-05,
+      "loss": 0.6479,
+      "step": 10188
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4553852991347053e-05,
+      "loss": 0.6541,
+      "step": 10189
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4552808584324538e-05,
+      "loss": 0.557,
+      "step": 10190
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.455176411465138e-05,
+      "loss": 0.7216,
+      "step": 10191
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4550719582341944e-05,
+      "loss": 0.5973,
+      "step": 10192
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4549674987410607e-05,
+      "loss": 0.619,
+      "step": 10193
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4548630329871749e-05,
+      "loss": 0.7053,
+      "step": 10194
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4547585609739737e-05,
+      "loss": 0.6616,
+      "step": 10195
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4546540827028955e-05,
+      "loss": 0.5869,
+      "step": 10196
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4545495981753773e-05,
+      "loss": 0.5784,
+      "step": 10197
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4544451073928574e-05,
+      "loss": 0.6372,
+      "step": 10198
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4543406103567735e-05,
+      "loss": 0.5757,
+      "step": 10199
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4542361070685636e-05,
+      "loss": 0.5788,
+      "step": 10200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4541315975296655e-05,
+      "loss": 0.5737,
+      "step": 10201
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4540270817415176e-05,
+      "loss": 0.7149,
+      "step": 10202
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4539225597055584e-05,
+      "loss": 0.6496,
+      "step": 10203
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4538180314232257e-05,
+      "loss": 0.6184,
+      "step": 10204
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4537134968959582e-05,
+      "loss": 0.6835,
+      "step": 10205
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4536089561251945e-05,
+      "loss": 0.5602,
+      "step": 10206
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453504409112373e-05,
+      "loss": 0.6978,
+      "step": 10207
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4533998558589319e-05,
+      "loss": 0.6563,
+      "step": 10208
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453295296366311e-05,
+      "loss": 0.6534,
+      "step": 10209
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453190730635948e-05,
+      "loss": 0.641,
+      "step": 10210
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4530861586692826e-05,
+      "loss": 0.6756,
+      "step": 10211
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4529815804677534e-05,
+      "loss": 0.677,
+      "step": 10212
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4528769960327997e-05,
+      "loss": 0.6471,
+      "step": 10213
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4527724053658608e-05,
+      "loss": 0.6767,
+      "step": 10214
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4526678084683755e-05,
+      "loss": 0.6537,
+      "step": 10215
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4525632053417836e-05,
+      "loss": 0.6455,
+      "step": 10216
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.452458595987524e-05,
+      "loss": 0.6658,
+      "step": 10217
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4523539804070368e-05,
+      "loss": 0.6787,
+      "step": 10218
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4522493586017611e-05,
+      "loss": 0.6515,
+      "step": 10219
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4521447305731374e-05,
+      "loss": 0.6074,
+      "step": 10220
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4520400963226044e-05,
+      "loss": 0.6526,
+      "step": 10221
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4519354558516029e-05,
+      "loss": 0.5508,
+      "step": 10222
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451830809161572e-05,
+      "loss": 0.6988,
+      "step": 10223
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4517261562539523e-05,
+      "loss": 0.7323,
+      "step": 10224
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4516214971301839e-05,
+      "loss": 0.6402,
+      "step": 10225
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4515168317917066e-05,
+      "loss": 0.649,
+      "step": 10226
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4514121602399613e-05,
+      "loss": 0.7149,
+      "step": 10227
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4513074824763877e-05,
+      "loss": 0.611,
+      "step": 10228
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4512027985024269e-05,
+      "loss": 0.6777,
+      "step": 10229
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4510981083195188e-05,
+      "loss": 0.6941,
+      "step": 10230
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4509934119291048e-05,
+      "loss": 0.6144,
+      "step": 10231
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4508887093326246e-05,
+      "loss": 0.6414,
+      "step": 10232
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45078400053152e-05,
+      "loss": 0.6999,
+      "step": 10233
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4506792855272314e-05,
+      "loss": 0.6853,
+      "step": 10234
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4505745643211997e-05,
+      "loss": 0.6532,
+      "step": 10235
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4504698369148664e-05,
+      "loss": 0.6768,
+      "step": 10236
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4503651033096719e-05,
+      "loss": 0.664,
+      "step": 10237
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4502603635070583e-05,
+      "loss": 0.6161,
+      "step": 10238
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450155617508466e-05,
+      "loss": 0.6777,
+      "step": 10239
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4500508653153372e-05,
+      "loss": 0.5961,
+      "step": 10240
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4499461069291128e-05,
+      "loss": 0.6693,
+      "step": 10241
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4498413423512347e-05,
+      "loss": 0.6259,
+      "step": 10242
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4497365715831446e-05,
+      "loss": 0.6396,
+      "step": 10243
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4496317946262839e-05,
+      "loss": 0.6179,
+      "step": 10244
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4495270114820947e-05,
+      "loss": 0.6245,
+      "step": 10245
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.449422222152019e-05,
+      "loss": 0.6012,
+      "step": 10246
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4493174266374985e-05,
+      "loss": 0.6393,
+      "step": 10247
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4492126249399752e-05,
+      "loss": 0.6198,
+      "step": 10248
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4491078170608916e-05,
+      "loss": 0.6541,
+      "step": 10249
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44900300300169e-05,
+      "loss": 0.6675,
+      "step": 10250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4488981827638125e-05,
+      "loss": 0.6058,
+      "step": 10251
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4487933563487014e-05,
+      "loss": 0.6421,
+      "step": 10252
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4486885237577994e-05,
+      "loss": 0.6559,
+      "step": 10253
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4485836849925492e-05,
+      "loss": 0.6484,
+      "step": 10254
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.448478840054393e-05,
+      "loss": 0.5996,
+      "step": 10255
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4483739889447743e-05,
+      "loss": 0.6474,
+      "step": 10256
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4482691316651354e-05,
+      "loss": 0.6547,
+      "step": 10257
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4481642682169194e-05,
+      "loss": 0.6564,
+      "step": 10258
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4480593986015691e-05,
+      "loss": 0.6023,
+      "step": 10259
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4479545228205281e-05,
+      "loss": 0.6298,
+      "step": 10260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4478496408752389e-05,
+      "loss": 0.6207,
+      "step": 10261
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4477447527671453e-05,
+      "loss": 0.6323,
+      "step": 10262
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4476398584976905e-05,
+      "loss": 0.5714,
+      "step": 10263
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4475349580683179e-05,
+      "loss": 0.5885,
+      "step": 10264
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4474300514804709e-05,
+      "loss": 0.6659,
+      "step": 10265
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4473251387355936e-05,
+      "loss": 0.7213,
+      "step": 10266
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.447220219835129e-05,
+      "loss": 0.6422,
+      "step": 10267
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4471152947805214e-05,
+      "loss": 0.5804,
+      "step": 10268
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4470103635732144e-05,
+      "loss": 0.6327,
+      "step": 10269
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.446905426214652e-05,
+      "loss": 0.6428,
+      "step": 10270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4468004827062782e-05,
+      "loss": 0.6458,
+      "step": 10271
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4466955330495374e-05,
+      "loss": 0.5854,
+      "step": 10272
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4465905772458732e-05,
+      "loss": 0.6841,
+      "step": 10273
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4464856152967305e-05,
+      "loss": 0.6664,
+      "step": 10274
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4463806472035532e-05,
+      "loss": 0.5772,
+      "step": 10275
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4462756729677863e-05,
+      "loss": 0.6608,
+      "step": 10276
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4461706925908735e-05,
+      "loss": 0.6259,
+      "step": 10277
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4460657060742607e-05,
+      "loss": 0.5959,
+      "step": 10278
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.445960713419391e-05,
+      "loss": 0.5838,
+      "step": 10279
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4458557146277106e-05,
+      "loss": 0.6651,
+      "step": 10280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4457507097006636e-05,
+      "loss": 0.7028,
+      "step": 10281
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4456456986396951e-05,
+      "loss": 0.5859,
+      "step": 10282
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4455406814462505e-05,
+      "loss": 0.6489,
+      "step": 10283
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4454356581217743e-05,
+      "loss": 0.6332,
+      "step": 10284
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4453306286677124e-05,
+      "loss": 0.6865,
+      "step": 10285
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4452255930855095e-05,
+      "loss": 0.5794,
+      "step": 10286
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4451205513766113e-05,
+      "loss": 0.6107,
+      "step": 10287
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4450155035424632e-05,
+      "loss": 0.7056,
+      "step": 10288
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4449104495845106e-05,
+      "loss": 0.6372,
+      "step": 10289
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4448053895041993e-05,
+      "loss": 0.5811,
+      "step": 10290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444700323302975e-05,
+      "loss": 0.6905,
+      "step": 10291
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4445952509822836e-05,
+      "loss": 0.5604,
+      "step": 10292
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444490172543571e-05,
+      "loss": 0.6542,
+      "step": 10293
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444385087988283e-05,
+      "loss": 0.6468,
+      "step": 10294
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4442799973178656e-05,
+      "loss": 0.5972,
+      "step": 10295
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4441749005337651e-05,
+      "loss": 0.6728,
+      "step": 10296
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4440697976374276e-05,
+      "loss": 0.6561,
+      "step": 10297
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4439646886302998e-05,
+      "loss": 0.6245,
+      "step": 10298
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4438595735138274e-05,
+      "loss": 0.5633,
+      "step": 10299
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4437544522894576e-05,
+      "loss": 0.6731,
+      "step": 10300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4436493249586363e-05,
+      "loss": 0.6416,
+      "step": 10301
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4435441915228106e-05,
+      "loss": 0.6295,
+      "step": 10302
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4434390519834272e-05,
+      "loss": 0.5542,
+      "step": 10303
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4433339063419326e-05,
+      "loss": 0.6397,
+      "step": 10304
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4432287545997741e-05,
+      "loss": 0.6328,
+      "step": 10305
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4431235967583983e-05,
+      "loss": 0.6414,
+      "step": 10306
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4430184328192527e-05,
+      "loss": 0.6574,
+      "step": 10307
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.442913262783784e-05,
+      "loss": 0.6965,
+      "step": 10308
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4428080866534397e-05,
+      "loss": 0.6705,
+      "step": 10309
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4427029044296669e-05,
+      "loss": 0.6541,
+      "step": 10310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4425977161139134e-05,
+      "loss": 0.6439,
+      "step": 10311
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4424925217076262e-05,
+      "loss": 0.6131,
+      "step": 10312
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.442387321212253e-05,
+      "loss": 0.5832,
+      "step": 10313
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4422821146292418e-05,
+      "loss": 0.7026,
+      "step": 10314
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44217690196004e-05,
+      "loss": 0.6527,
+      "step": 10315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4420716832060957e-05,
+      "loss": 0.6855,
+      "step": 10316
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4419664583688563e-05,
+      "loss": 0.6545,
+      "step": 10317
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44186122744977e-05,
+      "loss": 0.6151,
+      "step": 10318
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4417559904502852e-05,
+      "loss": 0.6008,
+      "step": 10319
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44165074737185e-05,
+      "loss": 0.5983,
+      "step": 10320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4415454982159121e-05,
+      "loss": 0.5937,
+      "step": 10321
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4414402429839201e-05,
+      "loss": 0.6464,
+      "step": 10322
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4413349816773228e-05,
+      "loss": 0.595,
+      "step": 10323
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.441229714297568e-05,
+      "loss": 0.6178,
+      "step": 10324
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.441124440846105e-05,
+      "loss": 0.6396,
+      "step": 10325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4410191613243817e-05,
+      "loss": 0.6393,
+      "step": 10326
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4409138757338477e-05,
+      "loss": 0.6616,
+      "step": 10327
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4408085840759509e-05,
+      "loss": 0.7025,
+      "step": 10328
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4407032863521411e-05,
+      "loss": 0.5716,
+      "step": 10329
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4405979825638663e-05,
+      "loss": 0.6528,
+      "step": 10330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4404926727125768e-05,
+      "loss": 0.6919,
+      "step": 10331
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4403873567997205e-05,
+      "loss": 0.6241,
+      "step": 10332
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4402820348267473e-05,
+      "loss": 0.6881,
+      "step": 10333
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4401767067951066e-05,
+      "loss": 0.6736,
+      "step": 10334
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4400713727062475e-05,
+      "loss": 0.6217,
+      "step": 10335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4399660325616198e-05,
+      "loss": 0.5705,
+      "step": 10336
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4398606863626727e-05,
+      "loss": 0.6296,
+      "step": 10337
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4397553341108563e-05,
+      "loss": 0.6641,
+      "step": 10338
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.43964997580762e-05,
+      "loss": 0.616,
+      "step": 10339
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4395446114544135e-05,
+      "loss": 0.6524,
+      "step": 10340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4394392410526871e-05,
+      "loss": 0.6247,
+      "step": 10341
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4393338646038906e-05,
+      "loss": 0.7086,
+      "step": 10342
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4392284821094742e-05,
+      "loss": 0.6226,
+      "step": 10343
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4391230935708878e-05,
+      "loss": 0.6416,
+      "step": 10344
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439017698989582e-05,
+      "loss": 0.606,
+      "step": 10345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4389122983670066e-05,
+      "loss": 0.6578,
+      "step": 10346
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4388068917046125e-05,
+      "loss": 0.6017,
+      "step": 10347
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.43870147900385e-05,
+      "loss": 0.6561,
+      "step": 10348
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4385960602661697e-05,
+      "loss": 0.6994,
+      "step": 10349
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4384906354930222e-05,
+      "loss": 0.6188,
+      "step": 10350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4383852046858584e-05,
+      "loss": 0.6717,
+      "step": 10351
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4382797678461287e-05,
+      "loss": 0.6834,
+      "step": 10352
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4381743249752842e-05,
+      "loss": 0.5394,
+      "step": 10353
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4380688760747762e-05,
+      "loss": 0.6383,
+      "step": 10354
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4379634211460556e-05,
+      "loss": 0.6334,
+      "step": 10355
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4378579601905735e-05,
+      "loss": 0.67,
+      "step": 10356
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4377524932097811e-05,
+      "loss": 0.6135,
+      "step": 10357
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4376470202051298e-05,
+      "loss": 0.6363,
+      "step": 10358
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4375415411780706e-05,
+      "loss": 0.6265,
+      "step": 10359
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4374360561300558e-05,
+      "loss": 0.5967,
+      "step": 10360
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4373305650625362e-05,
+      "loss": 0.6715,
+      "step": 10361
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4372250679769639e-05,
+      "loss": 0.5594,
+      "step": 10362
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4371195648747905e-05,
+      "loss": 0.5635,
+      "step": 10363
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4370140557574677e-05,
+      "loss": 0.6437,
+      "step": 10364
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4369085406264476e-05,
+      "loss": 0.7046,
+      "step": 10365
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4368030194831819e-05,
+      "loss": 0.6292,
+      "step": 10366
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.436697492329123e-05,
+      "loss": 0.6852,
+      "step": 10367
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.436591959165723e-05,
+      "loss": 0.6225,
+      "step": 10368
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4364864199944336e-05,
+      "loss": 0.5832,
+      "step": 10369
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4363808748167079e-05,
+      "loss": 0.5798,
+      "step": 10370
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4362753236339979e-05,
+      "loss": 0.6636,
+      "step": 10371
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.436169766447756e-05,
+      "loss": 0.6485,
+      "step": 10372
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4360642032594349e-05,
+      "loss": 0.6556,
+      "step": 10373
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4359586340704873e-05,
+      "loss": 0.6885,
+      "step": 10374
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4358530588823656e-05,
+      "loss": 0.6503,
+      "step": 10375
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435747477696523e-05,
+      "loss": 0.6122,
+      "step": 10376
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4356418905144123e-05,
+      "loss": 0.6616,
+      "step": 10377
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4355362973374863e-05,
+      "loss": 0.6042,
+      "step": 10378
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4354306981671982e-05,
+      "loss": 0.5955,
+      "step": 10379
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435325093005001e-05,
+      "loss": 0.6579,
+      "step": 10380
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435219481852348e-05,
+      "loss": 0.6679,
+      "step": 10381
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4351138647106923e-05,
+      "loss": 0.6687,
+      "step": 10382
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4350082415814878e-05,
+      "loss": 0.6426,
+      "step": 10383
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4349026124661878e-05,
+      "loss": 0.6418,
+      "step": 10384
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4347969773662453e-05,
+      "loss": 0.6521,
+      "step": 10385
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4346913362831147e-05,
+      "loss": 0.6456,
+      "step": 10386
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4345856892182492e-05,
+      "loss": 0.5753,
+      "step": 10387
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4344800361731028e-05,
+      "loss": 0.6461,
+      "step": 10388
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4343743771491292e-05,
+      "loss": 0.7298,
+      "step": 10389
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4342687121477827e-05,
+      "loss": 0.6602,
+      "step": 10390
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4341630411705169e-05,
+      "loss": 0.6582,
+      "step": 10391
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4340573642187865e-05,
+      "loss": 0.6508,
+      "step": 10392
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.433951681294045e-05,
+      "loss": 0.6029,
+      "step": 10393
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4338459923977474e-05,
+      "loss": 0.6561,
+      "step": 10394
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4337402975313475e-05,
+      "loss": 0.6681,
+      "step": 10395
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4336345966963e-05,
+      "loss": 0.6059,
+      "step": 10396
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4335288898940595e-05,
+      "loss": 0.6555,
+      "step": 10397
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4334231771260807e-05,
+      "loss": 0.6192,
+      "step": 10398
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4333174583938179e-05,
+      "loss": 0.6418,
+      "step": 10399
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4332117336987262e-05,
+      "loss": 0.5735,
+      "step": 10400
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4331060030422606e-05,
+      "loss": 0.6451,
+      "step": 10401
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4330002664258755e-05,
+      "loss": 0.6675,
+      "step": 10402
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4328945238510264e-05,
+      "loss": 0.5794,
+      "step": 10403
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4327887753191683e-05,
+      "loss": 0.6864,
+      "step": 10404
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4326830208317564e-05,
+      "loss": 0.6528,
+      "step": 10405
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4325772603902458e-05,
+      "loss": 0.6661,
+      "step": 10406
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4324714939960922e-05,
+      "loss": 0.6391,
+      "step": 10407
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4323657216507507e-05,
+      "loss": 0.6107,
+      "step": 10408
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.432259943355677e-05,
+      "loss": 0.6735,
+      "step": 10409
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4321541591123268e-05,
+      "loss": 0.5957,
+      "step": 10410
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.432048368922155e-05,
+      "loss": 0.5794,
+      "step": 10411
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431942572786619e-05,
+      "loss": 0.6896,
+      "step": 10412
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431836770707173e-05,
+      "loss": 0.5961,
+      "step": 10413
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4317309626852739e-05,
+      "loss": 0.5776,
+      "step": 10414
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4316251487223771e-05,
+      "loss": 0.6216,
+      "step": 10415
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4315193288199392e-05,
+      "loss": 0.6294,
+      "step": 10416
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431413502979416e-05,
+      "loss": 0.6329,
+      "step": 10417
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4313076712022642e-05,
+      "loss": 0.6342,
+      "step": 10418
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4312018334899398e-05,
+      "loss": 0.6473,
+      "step": 10419
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4310959898438989e-05,
+      "loss": 0.5811,
+      "step": 10420
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430990140265599e-05,
+      "loss": 0.6107,
+      "step": 10421
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4308842847564956e-05,
+      "loss": 0.6248,
+      "step": 10422
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4307784233180462e-05,
+      "loss": 0.6192,
+      "step": 10423
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430672555951707e-05,
+      "loss": 0.6152,
+      "step": 10424
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4305666826589348e-05,
+      "loss": 0.6669,
+      "step": 10425
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430460803441187e-05,
+      "loss": 0.6475,
+      "step": 10426
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4303549182999202e-05,
+      "loss": 0.6872,
+      "step": 10427
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4302490272365916e-05,
+      "loss": 0.5638,
+      "step": 10428
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4301431302526582e-05,
+      "loss": 0.6651,
+      "step": 10429
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4300372273495774e-05,
+      "loss": 0.6765,
+      "step": 10430
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4299313185288067e-05,
+      "loss": 0.6735,
+      "step": 10431
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4298254037918034e-05,
+      "loss": 0.5976,
+      "step": 10432
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4297194831400246e-05,
+      "loss": 0.6193,
+      "step": 10433
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.429613556574928e-05,
+      "loss": 0.617,
+      "step": 10434
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4295076240979716e-05,
+      "loss": 0.6402,
+      "step": 10435
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4294016857106128e-05,
+      "loss": 0.6338,
+      "step": 10436
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4292957414143096e-05,
+      "loss": 0.6541,
+      "step": 10437
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4291897912105198e-05,
+      "loss": 0.6227,
+      "step": 10438
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.429083835100701e-05,
+      "loss": 0.6596,
+      "step": 10439
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428977873086312e-05,
+      "loss": 0.6719,
+      "step": 10440
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4288719051688105e-05,
+      "loss": 0.663,
+      "step": 10441
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4287659313496545e-05,
+      "loss": 0.7156,
+      "step": 10442
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4286599516303028e-05,
+      "loss": 0.5902,
+      "step": 10443
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4285539660122135e-05,
+      "loss": 0.6016,
+      "step": 10444
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428447974496845e-05,
+      "loss": 0.602,
+      "step": 10445
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4283419770856557e-05,
+      "loss": 0.5999,
+      "step": 10446
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4282359737801046e-05,
+      "loss": 0.5983,
+      "step": 10447
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4281299645816502e-05,
+      "loss": 0.657,
+      "step": 10448
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4280239494917513e-05,
+      "loss": 0.6872,
+      "step": 10449
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4279179285118667e-05,
+      "loss": 0.5552,
+      "step": 10450
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4278119016434555e-05,
+      "loss": 0.6725,
+      "step": 10451
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4277058688879768e-05,
+      "loss": 0.5859,
+      "step": 10452
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.427599830246889e-05,
+      "loss": 0.6654,
+      "step": 10453
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4274937857216524e-05,
+      "loss": 0.5928,
+      "step": 10454
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4273877353137253e-05,
+      "loss": 0.5627,
+      "step": 10455
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4272816790245675e-05,
+      "loss": 0.5675,
+      "step": 10456
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4271756168556383e-05,
+      "loss": 0.6722,
+      "step": 10457
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4270695488083976e-05,
+      "loss": 0.7228,
+      "step": 10458
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4269634748843044e-05,
+      "loss": 0.6514,
+      "step": 10459
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4268573950848187e-05,
+      "loss": 0.6808,
+      "step": 10460
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4267513094114001e-05,
+      "loss": 0.5571,
+      "step": 10461
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4266452178655083e-05,
+      "loss": 0.6217,
+      "step": 10462
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4265391204486037e-05,
+      "loss": 0.7216,
+      "step": 10463
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4264330171621459e-05,
+      "loss": 0.5885,
+      "step": 10464
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4263269080075952e-05,
+      "loss": 0.6652,
+      "step": 10465
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4262207929864116e-05,
+      "loss": 0.6851,
+      "step": 10466
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4261146721000554e-05,
+      "loss": 0.6132,
+      "step": 10467
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4260085453499868e-05,
+      "loss": 0.6978,
+      "step": 10468
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4259024127376665e-05,
+      "loss": 0.6899,
+      "step": 10469
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4257962742645544e-05,
+      "loss": 0.6719,
+      "step": 10470
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4256901299321118e-05,
+      "loss": 0.5949,
+      "step": 10471
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4255839797417987e-05,
+      "loss": 0.6141,
+      "step": 10472
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4254778236950761e-05,
+      "loss": 0.6763,
+      "step": 10473
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.425371661793405e-05,
+      "loss": 0.6083,
+      "step": 10474
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.425265494038246e-05,
+      "loss": 0.6122,
+      "step": 10475
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4251593204310601e-05,
+      "loss": 0.6668,
+      "step": 10476
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4250531409733084e-05,
+      "loss": 0.6268,
+      "step": 10477
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4249469556664518e-05,
+      "loss": 0.7,
+      "step": 10478
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4248407645119522e-05,
+      "loss": 0.6936,
+      "step": 10479
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4247345675112697e-05,
+      "loss": 0.6823,
+      "step": 10480
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.424628364665867e-05,
+      "loss": 0.5671,
+      "step": 10481
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4245221559772045e-05,
+      "loss": 0.7074,
+      "step": 10482
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4244159414467442e-05,
+      "loss": 0.6568,
+      "step": 10483
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4243097210759476e-05,
+      "loss": 0.6529,
+      "step": 10484
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4242034948662764e-05,
+      "loss": 0.6341,
+      "step": 10485
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4240972628191924e-05,
+      "loss": 0.7033,
+      "step": 10486
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4239910249361575e-05,
+      "loss": 0.5978,
+      "step": 10487
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4238847812186334e-05,
+      "loss": 0.6755,
+      "step": 10488
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4237785316680821e-05,
+      "loss": 0.6448,
+      "step": 10489
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4236722762859662e-05,
+      "loss": 0.6702,
+      "step": 10490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4235660150737474e-05,
+      "loss": 0.6894,
+      "step": 10491
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.423459748032888e-05,
+      "loss": 0.706,
+      "step": 10492
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4233534751648504e-05,
+      "loss": 0.6189,
+      "step": 10493
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.423247196471097e-05,
+      "loss": 0.6352,
+      "step": 10494
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4231409119530901e-05,
+      "loss": 0.7026,
+      "step": 10495
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4230346216122927e-05,
+      "loss": 0.5723,
+      "step": 10496
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4229283254501672e-05,
+      "loss": 0.614,
+      "step": 10497
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.422822023468176e-05,
+      "loss": 0.6649,
+      "step": 10498
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4227157156677826e-05,
+      "loss": 0.6802,
+      "step": 10499
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4226094020504495e-05,
+      "loss": 0.5958,
+      "step": 10500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4225030826176396e-05,
+      "loss": 0.5788,
+      "step": 10501
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4223967573708159e-05,
+      "loss": 0.6809,
+      "step": 10502
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4222904263114419e-05,
+      "loss": 0.641,
+      "step": 10503
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4221840894409803e-05,
+      "loss": 0.6882,
+      "step": 10504
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.422077746760895e-05,
+      "loss": 0.6883,
+      "step": 10505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4219713982726488e-05,
+      "loss": 0.5961,
+      "step": 10506
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4218650439777055e-05,
+      "loss": 0.7017,
+      "step": 10507
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4217586838775283e-05,
+      "loss": 0.5876,
+      "step": 10508
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4216523179735812e-05,
+      "loss": 0.6171,
+      "step": 10509
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4215459462673278e-05,
+      "loss": 0.6246,
+      "step": 10510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4214395687602316e-05,
+      "loss": 0.5699,
+      "step": 10511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4213331854537566e-05,
+      "loss": 0.6169,
+      "step": 10512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421226796349367e-05,
+      "loss": 0.7231,
+      "step": 10513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4211204014485264e-05,
+      "loss": 0.5722,
+      "step": 10514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421014000752699e-05,
+      "loss": 0.627,
+      "step": 10515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.420907594263349e-05,
+      "loss": 0.7103,
+      "step": 10516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4208011819819407e-05,
+      "loss": 0.5575,
+      "step": 10517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4206947639099386e-05,
+      "loss": 0.6311,
+      "step": 10518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4205883400488066e-05,
+      "loss": 0.5596,
+      "step": 10519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4204819104000093e-05,
+      "loss": 0.7078,
+      "step": 10520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.420375474965012e-05,
+      "loss": 0.6255,
+      "step": 10521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4202690337452784e-05,
+      "loss": 0.6129,
+      "step": 10522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.420162586742274e-05,
+      "loss": 0.6722,
+      "step": 10523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4200561339574629e-05,
+      "loss": 0.7156,
+      "step": 10524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4199496753923104e-05,
+      "loss": 0.7107,
+      "step": 10525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4198432110482812e-05,
+      "loss": 0.6093,
+      "step": 10526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4197367409268407e-05,
+      "loss": 0.6384,
+      "step": 10527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4196302650294538e-05,
+      "loss": 0.6828,
+      "step": 10528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4195237833575854e-05,
+      "loss": 0.6577,
+      "step": 10529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4194172959127018e-05,
+      "loss": 0.6808,
+      "step": 10530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.419310802696267e-05,
+      "loss": 0.6007,
+      "step": 10531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4192043037097473e-05,
+      "loss": 0.6816,
+      "step": 10532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4190977989546082e-05,
+      "loss": 0.6483,
+      "step": 10533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4189912884323152e-05,
+      "loss": 0.6526,
+      "step": 10534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4188847721443337e-05,
+      "loss": 0.6323,
+      "step": 10535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.41877825009213e-05,
+      "loss": 0.6071,
+      "step": 10536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.418671722277169e-05,
+      "loss": 0.6911,
+      "step": 10537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.418565188700918e-05,
+      "loss": 0.6013,
+      "step": 10538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4184586493648416e-05,
+      "loss": 0.6614,
+      "step": 10539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4183521042704067e-05,
+      "loss": 0.6869,
+      "step": 10540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4182455534190793e-05,
+      "loss": 0.651,
+      "step": 10541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4181389968123254e-05,
+      "loss": 0.6248,
+      "step": 10542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4180324344516119e-05,
+      "loss": 0.5433,
+      "step": 10543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4179258663384043e-05,
+      "loss": 0.6598,
+      "step": 10544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4178192924741699e-05,
+      "loss": 0.6121,
+      "step": 10545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4177127128603748e-05,
+      "loss": 0.5964,
+      "step": 10546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4176061274984858e-05,
+      "loss": 0.6462,
+      "step": 10547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4174995363899693e-05,
+      "loss": 0.6344,
+      "step": 10548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4173929395362926e-05,
+      "loss": 0.707,
+      "step": 10549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.417286336938922e-05,
+      "loss": 0.6749,
+      "step": 10550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.417179728599325e-05,
+      "loss": 0.6492,
+      "step": 10551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4170731145189685e-05,
+      "loss": 0.6917,
+      "step": 10552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4169664946993192e-05,
+      "loss": 0.6255,
+      "step": 10553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4168598691418444e-05,
+      "loss": 0.6574,
+      "step": 10554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.416753237848012e-05,
+      "loss": 0.5649,
+      "step": 10555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4166466008192885e-05,
+      "loss": 0.6233,
+      "step": 10556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4165399580571417e-05,
+      "loss": 0.7138,
+      "step": 10557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4164333095630394e-05,
+      "loss": 0.5695,
+      "step": 10558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4163266553384484e-05,
+      "loss": 0.6577,
+      "step": 10559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.416219995384837e-05,
+      "loss": 0.6783,
+      "step": 10560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4161133297036729e-05,
+      "loss": 0.5866,
+      "step": 10561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4160066582964233e-05,
+      "loss": 0.6613,
+      "step": 10562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4158999811645571e-05,
+      "loss": 0.5992,
+      "step": 10563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4157932983095415e-05,
+      "loss": 0.6082,
+      "step": 10564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4156866097328448e-05,
+      "loss": 0.6661,
+      "step": 10565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4155799154359349e-05,
+      "loss": 0.6687,
+      "step": 10566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4154732154202803e-05,
+      "loss": 0.6801,
+      "step": 10567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4153665096873492e-05,
+      "loss": 0.6077,
+      "step": 10568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.41525979823861e-05,
+      "loss": 0.6806,
+      "step": 10569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4151530810755309e-05,
+      "loss": 0.6869,
+      "step": 10570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4150463581995811e-05,
+      "loss": 0.7088,
+      "step": 10571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4149396296122282e-05,
+      "loss": 0.5526,
+      "step": 10572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4148328953149416e-05,
+      "loss": 0.6319,
+      "step": 10573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.41472615530919e-05,
+      "loss": 0.6442,
+      "step": 10574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4146194095964419e-05,
+      "loss": 0.6885,
+      "step": 10575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4145126581781665e-05,
+      "loss": 0.736,
+      "step": 10576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4144059010558327e-05,
+      "loss": 0.6458,
+      "step": 10577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4142991382309097e-05,
+      "loss": 0.6958,
+      "step": 10578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4141923697048663e-05,
+      "loss": 0.6136,
+      "step": 10579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4140855954791719e-05,
+      "loss": 0.664,
+      "step": 10580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4139788155552961e-05,
+      "loss": 0.7135,
+      "step": 10581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.413872029934708e-05,
+      "loss": 0.6067,
+      "step": 10582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4137652386188769e-05,
+      "loss": 0.6372,
+      "step": 10583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4136584416092727e-05,
+      "loss": 0.6193,
+      "step": 10584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4135516389073648e-05,
+      "loss": 0.6492,
+      "step": 10585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4134448305146231e-05,
+      "loss": 0.6407,
+      "step": 10586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.413338016432517e-05,
+      "loss": 0.5311,
+      "step": 10587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.413231196662517e-05,
+      "loss": 0.5733,
+      "step": 10588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4131243712060922e-05,
+      "loss": 0.6934,
+      "step": 10589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4130175400647135e-05,
+      "loss": 0.6693,
+      "step": 10590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4129107032398501e-05,
+      "loss": 0.669,
+      "step": 10591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4128038607329728e-05,
+      "loss": 0.6356,
+      "step": 10592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4126970125455517e-05,
+      "loss": 0.7192,
+      "step": 10593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.412590158679057e-05,
+      "loss": 0.7441,
+      "step": 10594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4124832991349592e-05,
+      "loss": 0.6338,
+      "step": 10595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.412376433914729e-05,
+      "loss": 0.6471,
+      "step": 10596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4122695630198365e-05,
+      "loss": 0.6459,
+      "step": 10597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4121626864517523e-05,
+      "loss": 0.6464,
+      "step": 10598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4120558042119479e-05,
+      "loss": 0.7026,
+      "step": 10599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4119489163018932e-05,
+      "loss": 0.574,
+      "step": 10600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4118420227230596e-05,
+      "loss": 0.679,
+      "step": 10601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4117351234769178e-05,
+      "loss": 0.5617,
+      "step": 10602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.411628218564939e-05,
+      "loss": 0.6316,
+      "step": 10603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4115213079885941e-05,
+      "loss": 0.6761,
+      "step": 10604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4114143917493546e-05,
+      "loss": 0.6474,
+      "step": 10605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4113074698486915e-05,
+      "loss": 0.6649,
+      "step": 10606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4112005422880762e-05,
+      "loss": 0.6288,
+      "step": 10607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.41109360906898e-05,
+      "loss": 0.5385,
+      "step": 10608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4109866701928745e-05,
+      "loss": 0.5901,
+      "step": 10609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4108797256612317e-05,
+      "loss": 0.6615,
+      "step": 10610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4107727754755222e-05,
+      "loss": 0.5415,
+      "step": 10611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410665819637219e-05,
+      "loss": 0.5413,
+      "step": 10612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410558858147793e-05,
+      "loss": 0.6461,
+      "step": 10613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4104518910087164e-05,
+      "loss": 0.5499,
+      "step": 10614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410344918221461e-05,
+      "loss": 0.6391,
+      "step": 10615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4102379397874992e-05,
+      "loss": 0.6963,
+      "step": 10616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4101309557083027e-05,
+      "loss": 0.6628,
+      "step": 10617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4100239659853439e-05,
+      "loss": 0.5731,
+      "step": 10618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.409916970620095e-05,
+      "loss": 0.6903,
+      "step": 10619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4098099696140284e-05,
+      "loss": 0.7411,
+      "step": 10620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4097029629686167e-05,
+      "loss": 0.5568,
+      "step": 10621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4095959506853323e-05,
+      "loss": 0.6248,
+      "step": 10622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4094889327656475e-05,
+      "loss": 0.6111,
+      "step": 10623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4093819092110353e-05,
+      "loss": 0.6321,
+      "step": 10624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4092748800229684e-05,
+      "loss": 0.6264,
+      "step": 10625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4091678452029192e-05,
+      "loss": 0.6298,
+      "step": 10626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.409060804752361e-05,
+      "loss": 0.6575,
+      "step": 10627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4089537586727668e-05,
+      "loss": 0.6211,
+      "step": 10628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4088467069656095e-05,
+      "loss": 0.6326,
+      "step": 10629
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4087396496323625e-05,
+      "loss": 0.6027,
+      "step": 10630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4086325866744984e-05,
+      "loss": 0.5871,
+      "step": 10631
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4085255180934912e-05,
+      "loss": 0.6866,
+      "step": 10632
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4084184438908136e-05,
+      "loss": 0.5686,
+      "step": 10633
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4083113640679395e-05,
+      "loss": 0.6454,
+      "step": 10634
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.408204278626342e-05,
+      "loss": 0.6658,
+      "step": 10635
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4080971875674957e-05,
+      "loss": 0.5582,
+      "step": 10636
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4079900908928727e-05,
+      "loss": 0.6556,
+      "step": 10637
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4078829886039478e-05,
+      "loss": 0.6593,
+      "step": 10638
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4077758807021946e-05,
+      "loss": 0.62,
+      "step": 10639
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.407668767189087e-05,
+      "loss": 0.6429,
+      "step": 10640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4075616480660989e-05,
+      "loss": 0.6338,
+      "step": 10641
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4074545233347045e-05,
+      "loss": 0.6342,
+      "step": 10642
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.407347392996378e-05,
+      "loss": 0.6648,
+      "step": 10643
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4072402570525931e-05,
+      "loss": 0.5436,
+      "step": 10644
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4071331155048246e-05,
+      "loss": 0.6111,
+      "step": 10645
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4070259683545469e-05,
+      "loss": 0.5671,
+      "step": 10646
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.406918815603234e-05,
+      "loss": 0.6848,
+      "step": 10647
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4068116572523605e-05,
+      "loss": 0.6189,
+      "step": 10648
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4067044933034013e-05,
+      "loss": 0.6484,
+      "step": 10649
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.406597323757831e-05,
+      "loss": 0.7082,
+      "step": 10650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.406490148617124e-05,
+      "loss": 0.619,
+      "step": 10651
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4063829678827556e-05,
+      "loss": 0.5759,
+      "step": 10652
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4062757815562005e-05,
+      "loss": 0.6335,
+      "step": 10653
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4061685896389337e-05,
+      "loss": 0.6398,
+      "step": 10654
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.40606139213243e-05,
+      "loss": 0.6329,
+      "step": 10655
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405954189038165e-05,
+      "loss": 0.6366,
+      "step": 10656
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4058469803576133e-05,
+      "loss": 0.7135,
+      "step": 10657
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405739766092251e-05,
+      "loss": 0.676,
+      "step": 10658
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4056325462435525e-05,
+      "loss": 0.6396,
+      "step": 10659
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405525320812994e-05,
+      "loss": 0.6326,
+      "step": 10660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405418089802051e-05,
+      "loss": 0.6229,
+      "step": 10661
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4053108532121984e-05,
+      "loss": 0.5917,
+      "step": 10662
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4052036110449126e-05,
+      "loss": 0.6681,
+      "step": 10663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4050963633016687e-05,
+      "loss": 0.703,
+      "step": 10664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4049891099839434e-05,
+      "loss": 0.5364,
+      "step": 10665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4048818510932118e-05,
+      "loss": 0.5978,
+      "step": 10666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4047745866309504e-05,
+      "loss": 0.6739,
+      "step": 10667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4046673165986348e-05,
+      "loss": 0.6355,
+      "step": 10668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4045600409977414e-05,
+      "loss": 0.7081,
+      "step": 10669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4044527598297463e-05,
+      "loss": 0.6876,
+      "step": 10670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.404345473096126e-05,
+      "loss": 0.6128,
+      "step": 10671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4042381807983567e-05,
+      "loss": 0.5921,
+      "step": 10672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4041308829379148e-05,
+      "loss": 0.6382,
+      "step": 10673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.404023579516277e-05,
+      "loss": 0.578,
+      "step": 10674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4039162705349193e-05,
+      "loss": 0.6168,
+      "step": 10675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4038089559953192e-05,
+      "loss": 0.6874,
+      "step": 10676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4037016358989527e-05,
+      "loss": 0.6587,
+      "step": 10677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4035943102472973e-05,
+      "loss": 0.5938,
+      "step": 10678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4034869790418295e-05,
+      "loss": 0.6486,
+      "step": 10679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4033796422840262e-05,
+      "loss": 0.5213,
+      "step": 10680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4032722999753647e-05,
+      "loss": 0.6366,
+      "step": 10681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4031649521173218e-05,
+      "loss": 0.6288,
+      "step": 10682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4030575987113753e-05,
+      "loss": 0.644,
+      "step": 10683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4029502397590019e-05,
+      "loss": 0.693,
+      "step": 10684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402842875261679e-05,
+      "loss": 0.6006,
+      "step": 10685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4027355052208842e-05,
+      "loss": 0.6123,
+      "step": 10686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402628129638095e-05,
+      "loss": 0.5805,
+      "step": 10687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4025207485147892e-05,
+      "loss": 0.6733,
+      "step": 10688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4024133618524438e-05,
+      "loss": 0.6237,
+      "step": 10689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4023059696525375e-05,
+      "loss": 0.6466,
+      "step": 10690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402198571916547e-05,
+      "loss": 0.6544,
+      "step": 10691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402091168645951e-05,
+      "loss": 0.6207,
+      "step": 10692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401983759842227e-05,
+      "loss": 0.5962,
+      "step": 10693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4018763455068536e-05,
+      "loss": 0.6463,
+      "step": 10694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4017689256413083e-05,
+      "loss": 0.6543,
+      "step": 10695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4016615002470697e-05,
+      "loss": 0.6696,
+      "step": 10696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401554069325616e-05,
+      "loss": 0.65,
+      "step": 10697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4014466328784251e-05,
+      "loss": 0.6888,
+      "step": 10698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4013391909069763e-05,
+      "loss": 0.6743,
+      "step": 10699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4012317434127472e-05,
+      "loss": 0.5512,
+      "step": 10700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401124290397217e-05,
+      "loss": 0.6787,
+      "step": 10701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4010168318618637e-05,
+      "loss": 0.6081,
+      "step": 10702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4009093678081669e-05,
+      "loss": 0.6495,
+      "step": 10703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4008018982376045e-05,
+      "loss": 0.6578,
+      "step": 10704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4006944231516564e-05,
+      "loss": 0.6685,
+      "step": 10705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4005869425518003e-05,
+      "loss": 0.6814,
+      "step": 10706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4004794564395163e-05,
+      "loss": 0.614,
+      "step": 10707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.400371964816283e-05,
+      "loss": 0.6019,
+      "step": 10708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4002644676835796e-05,
+      "loss": 0.6909,
+      "step": 10709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4001569650428855e-05,
+      "loss": 0.6127,
+      "step": 10710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4000494568956797e-05,
+      "loss": 0.6331,
+      "step": 10711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3999419432434422e-05,
+      "loss": 0.6877,
+      "step": 10712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399834424087652e-05,
+      "loss": 0.6473,
+      "step": 10713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399726899429789e-05,
+      "loss": 0.6185,
+      "step": 10714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3996193692713323e-05,
+      "loss": 0.5774,
+      "step": 10715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3995118336137624e-05,
+      "loss": 0.6182,
+      "step": 10716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399404292458558e-05,
+      "loss": 0.6695,
+      "step": 10717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3992967458071997e-05,
+      "loss": 0.6032,
+      "step": 10718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3991891936611678e-05,
+      "loss": 0.6235,
+      "step": 10719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3990816360219415e-05,
+      "loss": 0.569,
+      "step": 10720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3989740728910014e-05,
+      "loss": 0.6391,
+      "step": 10721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3988665042698273e-05,
+      "loss": 0.6236,
+      "step": 10722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3987589301598998e-05,
+      "loss": 0.7273,
+      "step": 10723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3986513505626988e-05,
+      "loss": 0.6522,
+      "step": 10724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3985437654797052e-05,
+      "loss": 0.6224,
+      "step": 10725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.398436174912399e-05,
+      "loss": 0.6488,
+      "step": 10726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.398328578862261e-05,
+      "loss": 0.5289,
+      "step": 10727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3982209773307717e-05,
+      "loss": 0.5354,
+      "step": 10728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3981133703194118e-05,
+      "loss": 0.5769,
+      "step": 10729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3980057578296624e-05,
+      "loss": 0.5744,
+      "step": 10730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3978981398630038e-05,
+      "loss": 0.605,
+      "step": 10731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3977905164209174e-05,
+      "loss": 0.6347,
+      "step": 10732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3976828875048836e-05,
+      "loss": 0.6398,
+      "step": 10733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3975752531163842e-05,
+      "loss": 0.6541,
+      "step": 10734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3974676132568999e-05,
+      "loss": 0.6566,
+      "step": 10735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3973599679279119e-05,
+      "loss": 0.5747,
+      "step": 10736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3972523171309015e-05,
+      "loss": 0.6664,
+      "step": 10737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3971446608673502e-05,
+      "loss": 0.6242,
+      "step": 10738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3970369991387397e-05,
+      "loss": 0.6372,
+      "step": 10739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3969293319465508e-05,
+      "loss": 0.6256,
+      "step": 10740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3968216592922656e-05,
+      "loss": 0.7156,
+      "step": 10741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3967139811773658e-05,
+      "loss": 0.618,
+      "step": 10742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3966062976033332e-05,
+      "loss": 0.5708,
+      "step": 10743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3964986085716493e-05,
+      "loss": 0.6221,
+      "step": 10744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3963909140837962e-05,
+      "loss": 0.6509,
+      "step": 10745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3962832141412558e-05,
+      "loss": 0.5961,
+      "step": 10746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.39617550874551e-05,
+      "loss": 0.5392,
+      "step": 10747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3960677978980413e-05,
+      "loss": 0.6217,
+      "step": 10748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3959600816003314e-05,
+      "loss": 0.6503,
+      "step": 10749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3958523598538632e-05,
+      "loss": 0.5577,
+      "step": 10750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3957446326601185e-05,
+      "loss": 0.6532,
+      "step": 10751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3956369000205801e-05,
+      "loss": 0.6764,
+      "step": 10752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3955291619367302e-05,
+      "loss": 0.5997,
+      "step": 10753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3954214184100515e-05,
+      "loss": 0.6657,
+      "step": 10754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3953136694420267e-05,
+      "loss": 0.6227,
+      "step": 10755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3952059150341383e-05,
+      "loss": 0.575,
+      "step": 10756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3950981551878693e-05,
+      "loss": 0.6598,
+      "step": 10757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3949903899047023e-05,
+      "loss": 0.6204,
+      "step": 10758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3948826191861211e-05,
+      "loss": 0.6288,
+      "step": 10759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3947748430336075e-05,
+      "loss": 0.7043,
+      "step": 10760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3946670614486455e-05,
+      "loss": 0.7376,
+      "step": 10761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3945592744327177e-05,
+      "loss": 0.6516,
+      "step": 10762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3944514819873078e-05,
+      "loss": 0.58,
+      "step": 10763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3943436841138988e-05,
+      "loss": 0.5982,
+      "step": 10764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3942358808139745e-05,
+      "loss": 0.592,
+      "step": 10765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3941280720890178e-05,
+      "loss": 0.5706,
+      "step": 10766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3940202579405125e-05,
+      "loss": 0.7359,
+      "step": 10767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3939124383699422e-05,
+      "loss": 0.7071,
+      "step": 10768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3938046133787906e-05,
+      "loss": 0.6706,
+      "step": 10769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3936967829685418e-05,
+      "loss": 0.6276,
+      "step": 10770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.393588947140679e-05,
+      "loss": 0.6003,
+      "step": 10771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3934811058966865e-05,
+      "loss": 0.5704,
+      "step": 10772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3933732592380485e-05,
+      "loss": 0.5693,
+      "step": 10773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3932654071662486e-05,
+      "loss": 0.5833,
+      "step": 10774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3931575496827711e-05,
+      "loss": 0.6002,
+      "step": 10775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3930496867891004e-05,
+      "loss": 0.5807,
+      "step": 10776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3929418184867206e-05,
+      "loss": 0.6394,
+      "step": 10777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3928339447771162e-05,
+      "loss": 0.6017,
+      "step": 10778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3927260656617713e-05,
+      "loss": 0.6595,
+      "step": 10779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3926181811421705e-05,
+      "loss": 0.66,
+      "step": 10780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3925102912197993e-05,
+      "loss": 0.7013,
+      "step": 10781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3924023958961408e-05,
+      "loss": 0.5401,
+      "step": 10782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3922944951726811e-05,
+      "loss": 0.6958,
+      "step": 10783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3921865890509042e-05,
+      "loss": 0.5555,
+      "step": 10784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3920786775322952e-05,
+      "loss": 0.6456,
+      "step": 10785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.391970760618339e-05,
+      "loss": 0.615,
+      "step": 10786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.391862838310521e-05,
+      "loss": 0.6919,
+      "step": 10787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.391754910610326e-05,
+      "loss": 0.6779,
+      "step": 10788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3916469775192386e-05,
+      "loss": 0.6449,
+      "step": 10789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3915390390387455e-05,
+      "loss": 0.5815,
+      "step": 10790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3914310951703306e-05,
+      "loss": 0.6792,
+      "step": 10791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3913231459154803e-05,
+      "loss": 0.6274,
+      "step": 10792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3912151912756796e-05,
+      "loss": 0.5395,
+      "step": 10793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.391107231252414e-05,
+      "loss": 0.6412,
+      "step": 10794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3909992658471692e-05,
+      "loss": 0.6495,
+      "step": 10795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3908912950614309e-05,
+      "loss": 0.6849,
+      "step": 10796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.390783318896685e-05,
+      "loss": 0.6549,
+      "step": 10797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3906753373544173e-05,
+      "loss": 0.5908,
+      "step": 10798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3905673504361136e-05,
+      "loss": 0.6684,
+      "step": 10799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.39045935814326e-05,
+      "loss": 0.6165,
+      "step": 10800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3903513604773427e-05,
+      "loss": 0.6934,
+      "step": 10801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3902433574398476e-05,
+      "loss": 0.6789,
+      "step": 10802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.390135349032261e-05,
+      "loss": 0.6113,
+      "step": 10803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3900273352560689e-05,
+      "loss": 0.6586,
+      "step": 10804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3899193161127583e-05,
+      "loss": 0.6199,
+      "step": 10805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3898112916038151e-05,
+      "loss": 0.6708,
+      "step": 10806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.389703261730726e-05,
+      "loss": 0.6027,
+      "step": 10807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3895952264949779e-05,
+      "loss": 0.5855,
+      "step": 10808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3894871858980568e-05,
+      "loss": 0.6637,
+      "step": 10809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.38937913994145e-05,
+      "loss": 0.5373,
+      "step": 10810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.389271088626644e-05,
+      "loss": 0.6054,
+      "step": 10811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.389163031955126e-05,
+      "loss": 0.6343,
+      "step": 10812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3890549699283824e-05,
+      "loss": 0.6765,
+      "step": 10813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3889469025479009e-05,
+      "loss": 0.5982,
+      "step": 10814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3888388298151679e-05,
+      "loss": 0.628,
+      "step": 10815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388730751731671e-05,
+      "loss": 0.5903,
+      "step": 10816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3886226682988975e-05,
+      "loss": 0.583,
+      "step": 10817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3885145795183345e-05,
+      "loss": 0.648,
+      "step": 10818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3884064853914698e-05,
+      "loss": 0.65,
+      "step": 10819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3882983859197903e-05,
+      "loss": 0.6573,
+      "step": 10820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388190281104784e-05,
+      "loss": 0.7035,
+      "step": 10821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3880821709479383e-05,
+      "loss": 0.6183,
+      "step": 10822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3879740554507408e-05,
+      "loss": 0.6516,
+      "step": 10823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3878659346146795e-05,
+      "loss": 0.6287,
+      "step": 10824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3877578084412422e-05,
+      "loss": 0.6413,
+      "step": 10825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3876496769319166e-05,
+      "loss": 0.6613,
+      "step": 10826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3875415400881907e-05,
+      "loss": 0.649,
+      "step": 10827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3874333979115529e-05,
+      "loss": 0.6417,
+      "step": 10828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387325250403491e-05,
+      "loss": 0.5951,
+      "step": 10829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3872170975654935e-05,
+      "loss": 0.674,
+      "step": 10830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3871089393990482e-05,
+      "loss": 0.605,
+      "step": 10831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3870007759056441e-05,
+      "loss": 0.6339,
+      "step": 10832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386892607086769e-05,
+      "loss": 0.6234,
+      "step": 10833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386784432943912e-05,
+      "loss": 0.6653,
+      "step": 10834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386676253478561e-05,
+      "loss": 0.6302,
+      "step": 10835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386568068692205e-05,
+      "loss": 0.6505,
+      "step": 10836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386459878586333e-05,
+      "loss": 0.6058,
+      "step": 10837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3863516831624333e-05,
+      "loss": 0.6842,
+      "step": 10838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386243482421995e-05,
+      "loss": 0.607,
+      "step": 10839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3861352763665067e-05,
+      "loss": 0.5351,
+      "step": 10840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3860270649974582e-05,
+      "loss": 0.6085,
+      "step": 10841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.385918848316338e-05,
+      "loss": 0.5455,
+      "step": 10842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3858106263246351e-05,
+      "loss": 0.5981,
+      "step": 10843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3857023990238391e-05,
+      "loss": 0.5642,
+      "step": 10844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3855941664154393e-05,
+      "loss": 0.596,
+      "step": 10845
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3854859285009248e-05,
+      "loss": 0.6215,
+      "step": 10846
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.385377685281785e-05,
+      "loss": 0.6642,
+      "step": 10847
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3852694367595102e-05,
+      "loss": 0.6604,
+      "step": 10848
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.385161182935589e-05,
+      "loss": 0.661,
+      "step": 10849
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3850529238115115e-05,
+      "loss": 0.6893,
+      "step": 10850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3849446593887677e-05,
+      "loss": 0.591,
+      "step": 10851
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384836389668847e-05,
+      "loss": 0.602,
+      "step": 10852
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3847281146532392e-05,
+      "loss": 0.6924,
+      "step": 10853
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384619834343435e-05,
+      "loss": 0.6473,
+      "step": 10854
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3845115487409233e-05,
+      "loss": 0.5872,
+      "step": 10855
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3844032578471953e-05,
+      "loss": 0.6331,
+      "step": 10856
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3842949616637404e-05,
+      "loss": 0.5471,
+      "step": 10857
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3841866601920493e-05,
+      "loss": 0.6257,
+      "step": 10858
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3840783534336123e-05,
+      "loss": 0.667,
+      "step": 10859
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3839700413899194e-05,
+      "loss": 0.6227,
+      "step": 10860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3838617240624616e-05,
+      "loss": 0.674,
+      "step": 10861
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3837534014527292e-05,
+      "loss": 0.6169,
+      "step": 10862
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3836450735622126e-05,
+      "loss": 0.6894,
+      "step": 10863
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.383536740392403e-05,
+      "loss": 0.6414,
+      "step": 10864
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3834284019447906e-05,
+      "loss": 0.556,
+      "step": 10865
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3833200582208664e-05,
+      "loss": 0.6461,
+      "step": 10866
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3832117092221213e-05,
+      "loss": 0.6402,
+      "step": 10867
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3831033549500468e-05,
+      "loss": 0.5762,
+      "step": 10868
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3829949954061332e-05,
+      "loss": 0.6953,
+      "step": 10869
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382886630591872e-05,
+      "loss": 0.5418,
+      "step": 10870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3827782605087543e-05,
+      "loss": 0.6033,
+      "step": 10871
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3826698851582716e-05,
+      "loss": 0.6145,
+      "step": 10872
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3825615045419148e-05,
+      "loss": 0.6024,
+      "step": 10873
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3824531186611756e-05,
+      "loss": 0.6657,
+      "step": 10874
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3823447275175454e-05,
+      "loss": 0.6956,
+      "step": 10875
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382236331112516e-05,
+      "loss": 0.6194,
+      "step": 10876
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3821279294475785e-05,
+      "loss": 0.6888,
+      "step": 10877
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382019522524225e-05,
+      "loss": 0.6455,
+      "step": 10878
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3819111103439473e-05,
+      "loss": 0.6646,
+      "step": 10879
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.381802692908237e-05,
+      "loss": 0.649,
+      "step": 10880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3816942702185864e-05,
+      "loss": 0.6493,
+      "step": 10881
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3815858422764872e-05,
+      "loss": 0.6142,
+      "step": 10882
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3814774090834315e-05,
+      "loss": 0.5886,
+      "step": 10883
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3813689706409114e-05,
+      "loss": 0.6519,
+      "step": 10884
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3812605269504194e-05,
+      "loss": 0.5915,
+      "step": 10885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3811520780134471e-05,
+      "loss": 0.5964,
+      "step": 10886
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3810436238314877e-05,
+      "loss": 0.6296,
+      "step": 10887
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3809351644060329e-05,
+      "loss": 0.6462,
+      "step": 10888
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3808266997385756e-05,
+      "loss": 0.5842,
+      "step": 10889
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3807182298306086e-05,
+      "loss": 0.6499,
+      "step": 10890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.380609754683624e-05,
+      "loss": 0.6861,
+      "step": 10891
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3805012742991147e-05,
+      "loss": 0.5839,
+      "step": 10892
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3803927886785732e-05,
+      "loss": 0.5951,
+      "step": 10893
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3802842978234933e-05,
+      "loss": 0.6507,
+      "step": 10894
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.380175801735367e-05,
+      "loss": 0.676,
+      "step": 10895
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.380067300415688e-05,
+      "loss": 0.6562,
+      "step": 10896
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3799587938659485e-05,
+      "loss": 0.6712,
+      "step": 10897
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.379850282087642e-05,
+      "loss": 0.6376,
+      "step": 10898
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3797417650822626e-05,
+      "loss": 0.633,
+      "step": 10899
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3796332428513024e-05,
+      "loss": 0.6829,
+      "step": 10900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3795247153962553e-05,
+      "loss": 0.669,
+      "step": 10901
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3794161827186143e-05,
+      "loss": 0.668,
+      "step": 10902
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3793076448198737e-05,
+      "loss": 0.656,
+      "step": 10903
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3791991017015262e-05,
+      "loss": 0.6813,
+      "step": 10904
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3790905533650661e-05,
+      "loss": 0.6095,
+      "step": 10905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3789819998119869e-05,
+      "loss": 0.607,
+      "step": 10906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3788734410437821e-05,
+      "loss": 0.5942,
+      "step": 10907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3787648770619462e-05,
+      "loss": 0.6924,
+      "step": 10908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3786563078679727e-05,
+      "loss": 0.6165,
+      "step": 10909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3785477334633557e-05,
+      "loss": 0.6374,
+      "step": 10910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3784391538495892e-05,
+      "loss": 0.6227,
+      "step": 10911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3783305690281675e-05,
+      "loss": 0.598,
+      "step": 10912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3782219790005846e-05,
+      "loss": 0.5879,
+      "step": 10913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.378113383768335e-05,
+      "loss": 0.5717,
+      "step": 10914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3780047833329129e-05,
+      "loss": 0.5325,
+      "step": 10915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3778961776958128e-05,
+      "loss": 0.6162,
+      "step": 10916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3777875668585296e-05,
+      "loss": 0.6338,
+      "step": 10917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3776789508225572e-05,
+      "loss": 0.6087,
+      "step": 10918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3775703295893907e-05,
+      "loss": 0.651,
+      "step": 10919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3774617031605246e-05,
+      "loss": 0.6041,
+      "step": 10920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3773530715374542e-05,
+      "loss": 0.6465,
+      "step": 10921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3772444347216734e-05,
+      "loss": 0.6128,
+      "step": 10922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.377135792714678e-05,
+      "loss": 0.6192,
+      "step": 10923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3770271455179627e-05,
+      "loss": 0.5547,
+      "step": 10924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3769184931330223e-05,
+      "loss": 0.6022,
+      "step": 10925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3768098355613528e-05,
+      "loss": 0.6011,
+      "step": 10926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3767011728044483e-05,
+      "loss": 0.5604,
+      "step": 10927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3765925048638051e-05,
+      "loss": 0.6594,
+      "step": 10928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376483831740918e-05,
+      "loss": 0.6457,
+      "step": 10929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3763751534372825e-05,
+      "loss": 0.6369,
+      "step": 10930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3762664699543941e-05,
+      "loss": 0.5677,
+      "step": 10931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3761577812937489e-05,
+      "loss": 0.6187,
+      "step": 10932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3760490874568415e-05,
+      "loss": 0.6611,
+      "step": 10933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3759403884451687e-05,
+      "loss": 0.6197,
+      "step": 10934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3758316842602256e-05,
+      "loss": 0.5738,
+      "step": 10935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3757229749035081e-05,
+      "loss": 0.6539,
+      "step": 10936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3756142603765126e-05,
+      "loss": 0.7034,
+      "step": 10937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3755055406807348e-05,
+      "loss": 0.6379,
+      "step": 10938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375396815817671e-05,
+      "loss": 0.6745,
+      "step": 10939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3752880857888166e-05,
+      "loss": 0.5939,
+      "step": 10940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375179350595669e-05,
+      "loss": 0.608,
+      "step": 10941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3750706102397235e-05,
+      "loss": 0.764,
+      "step": 10942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3749618647224772e-05,
+      "loss": 0.604,
+      "step": 10943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3748531140454258e-05,
+      "loss": 0.6721,
+      "step": 10944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3747443582100661e-05,
+      "loss": 0.6129,
+      "step": 10945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3746355972178951e-05,
+      "loss": 0.5899,
+      "step": 10946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3745268310704087e-05,
+      "loss": 0.6647,
+      "step": 10947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3744180597691044e-05,
+      "loss": 0.6523,
+      "step": 10948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3743092833154785e-05,
+      "loss": 0.6747,
+      "step": 10949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3742005017110281e-05,
+      "loss": 0.5949,
+      "step": 10950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3740917149572497e-05,
+      "loss": 0.6175,
+      "step": 10951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3739829230556408e-05,
+      "loss": 0.6528,
+      "step": 10952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373874126007698e-05,
+      "loss": 0.6793,
+      "step": 10953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3737653238149192e-05,
+      "loss": 0.7125,
+      "step": 10954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3736565164788008e-05,
+      "loss": 0.6023,
+      "step": 10955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3735477040008403e-05,
+      "loss": 0.6877,
+      "step": 10956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3734388863825352e-05,
+      "loss": 0.6027,
+      "step": 10957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373330063625383e-05,
+      "loss": 0.5848,
+      "step": 10958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3732212357308812e-05,
+      "loss": 0.6267,
+      "step": 10959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373112402700527e-05,
+      "loss": 0.6144,
+      "step": 10960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3730035645358187e-05,
+      "loss": 0.6026,
+      "step": 10961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.372894721238253e-05,
+      "loss": 0.7091,
+      "step": 10962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3727858728093288e-05,
+      "loss": 0.6339,
+      "step": 10963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3726770192505432e-05,
+      "loss": 0.638,
+      "step": 10964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3725681605633946e-05,
+      "loss": 0.6567,
+      "step": 10965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3724592967493804e-05,
+      "loss": 0.6916,
+      "step": 10966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3723504278099991e-05,
+      "loss": 0.5563,
+      "step": 10967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3722415537467488e-05,
+      "loss": 0.5777,
+      "step": 10968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3721326745611277e-05,
+      "loss": 0.7059,
+      "step": 10969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3720237902546339e-05,
+      "loss": 0.6223,
+      "step": 10970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3719149008287658e-05,
+      "loss": 0.696,
+      "step": 10971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.371806006285022e-05,
+      "loss": 0.6262,
+      "step": 10972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3716971066249008e-05,
+      "loss": 0.5628,
+      "step": 10973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.371588201849901e-05,
+      "loss": 0.6829,
+      "step": 10974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3714792919615207e-05,
+      "loss": 0.6572,
+      "step": 10975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.371370376961259e-05,
+      "loss": 0.6403,
+      "step": 10976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3712614568506149e-05,
+      "loss": 0.5843,
+      "step": 10977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3711525316310867e-05,
+      "loss": 0.5873,
+      "step": 10978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3710436013041738e-05,
+      "loss": 0.5869,
+      "step": 10979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3709346658713745e-05,
+      "loss": 0.672,
+      "step": 10980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3708257253341886e-05,
+      "loss": 0.5868,
+      "step": 10981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3707167796941146e-05,
+      "loss": 0.6189,
+      "step": 10982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3706078289526521e-05,
+      "loss": 0.6577,
+      "step": 10983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3704988731113003e-05,
+      "loss": 0.6918,
+      "step": 10984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3703899121715585e-05,
+      "loss": 0.6592,
+      "step": 10985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3702809461349258e-05,
+      "loss": 0.661,
+      "step": 10986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3701719750029021e-05,
+      "loss": 0.5887,
+      "step": 10987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3700629987769868e-05,
+      "loss": 0.7001,
+      "step": 10988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3699540174586793e-05,
+      "loss": 0.6031,
+      "step": 10989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3698450310494798e-05,
+      "loss": 0.6315,
+      "step": 10990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3697360395508871e-05,
+      "loss": 0.632,
+      "step": 10991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3696270429644023e-05,
+      "loss": 0.613,
+      "step": 10992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369518041291524e-05,
+      "loss": 0.6613,
+      "step": 10993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3694090345337534e-05,
+      "loss": 0.6557,
+      "step": 10994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3693000226925894e-05,
+      "loss": 0.6939,
+      "step": 10995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3691910057695328e-05,
+      "loss": 0.6956,
+      "step": 10996
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3690819837660834e-05,
+      "loss": 0.5842,
+      "step": 10997
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3689729566837416e-05,
+      "loss": 0.6272,
+      "step": 10998
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3688639245240078e-05,
+      "loss": 0.5491,
+      "step": 10999
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3687548872883824e-05,
+      "loss": 0.58,
+      "step": 11000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3686458449783657e-05,
+      "loss": 0.6664,
+      "step": 11001
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368536797595458e-05,
+      "loss": 0.5962,
+      "step": 11002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3684277451411606e-05,
+      "loss": 0.6191,
+      "step": 11003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3683186876169734e-05,
+      "loss": 0.6525,
+      "step": 11004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3682096250243975e-05,
+      "loss": 0.6266,
+      "step": 11005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3681005573649335e-05,
+      "loss": 0.6485,
+      "step": 11006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3679914846400824e-05,
+      "loss": 0.647,
+      "step": 11007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3678824068513455e-05,
+      "loss": 0.6279,
+      "step": 11008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3677733240002228e-05,
+      "loss": 0.5131,
+      "step": 11009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3676642360882167e-05,
+      "loss": 0.6457,
+      "step": 11010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.367555143116827e-05,
+      "loss": 0.5818,
+      "step": 11011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.367446045087556e-05,
+      "loss": 0.7108,
+      "step": 11012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3673369420019043e-05,
+      "loss": 0.6051,
+      "step": 11013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3672278338613736e-05,
+      "loss": 0.6405,
+      "step": 11014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3671187206674652e-05,
+      "loss": 0.5663,
+      "step": 11015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3670096024216806e-05,
+      "loss": 0.6063,
+      "step": 11016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3669004791255216e-05,
+      "loss": 0.6082,
+      "step": 11017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3667913507804892e-05,
+      "loss": 0.6152,
+      "step": 11018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3666822173880861e-05,
+      "loss": 0.6602,
+      "step": 11019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.366573078949813e-05,
+      "loss": 0.6465,
+      "step": 11020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3664639354671724e-05,
+      "loss": 0.6292,
+      "step": 11021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.366354786941666e-05,
+      "loss": 0.6492,
+      "step": 11022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.366245633374796e-05,
+      "loss": 0.6201,
+      "step": 11023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.366136474768064e-05,
+      "loss": 0.5383,
+      "step": 11024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3660273111229727e-05,
+      "loss": 0.5847,
+      "step": 11025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365918142441024e-05,
+      "loss": 0.6439,
+      "step": 11026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3658089687237197e-05,
+      "loss": 0.6222,
+      "step": 11027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365699789972563e-05,
+      "loss": 0.6675,
+      "step": 11028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3655906061890557e-05,
+      "loss": 0.6489,
+      "step": 11029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3654814173747005e-05,
+      "loss": 0.6087,
+      "step": 11030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3653722235309996e-05,
+      "loss": 0.674,
+      "step": 11031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365263024659456e-05,
+      "loss": 0.5766,
+      "step": 11032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3651538207615724e-05,
+      "loss": 0.6017,
+      "step": 11033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3650446118388509e-05,
+      "loss": 0.6325,
+      "step": 11034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3649353978927952e-05,
+      "loss": 0.6503,
+      "step": 11035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3648261789249077e-05,
+      "loss": 0.6789,
+      "step": 11036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3647169549366917e-05,
+      "loss": 0.5788,
+      "step": 11037
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3646077259296492e-05,
+      "loss": 0.5936,
+      "step": 11038
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3644984919052849e-05,
+      "loss": 0.6487,
+      "step": 11039
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3643892528651004e-05,
+      "loss": 0.6692,
+      "step": 11040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3642800088105998e-05,
+      "loss": 0.7272,
+      "step": 11041
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3641707597432862e-05,
+      "loss": 0.568,
+      "step": 11042
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3640615056646632e-05,
+      "loss": 0.6991,
+      "step": 11043
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3639522465762336e-05,
+      "loss": 0.6954,
+      "step": 11044
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3638429824795014e-05,
+      "loss": 0.6732,
+      "step": 11045
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3637337133759703e-05,
+      "loss": 0.6032,
+      "step": 11046
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3636244392671434e-05,
+      "loss": 0.6058,
+      "step": 11047
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3635151601545249e-05,
+      "loss": 0.6366,
+      "step": 11048
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.363405876039618e-05,
+      "loss": 0.6872,
+      "step": 11049
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3632965869239274e-05,
+      "loss": 0.6575,
+      "step": 11050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3631872928089564e-05,
+      "loss": 0.6323,
+      "step": 11051
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3630779936962092e-05,
+      "loss": 0.6764,
+      "step": 11052
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3629686895871896e-05,
+      "loss": 0.6245,
+      "step": 11053
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3628593804834021e-05,
+      "loss": 0.6252,
+      "step": 11054
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3627500663863505e-05,
+      "loss": 0.6438,
+      "step": 11055
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3626407472975393e-05,
+      "loss": 0.7428,
+      "step": 11056
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3625314232184727e-05,
+      "loss": 0.6691,
+      "step": 11057
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3624220941506554e-05,
+      "loss": 0.6417,
+      "step": 11058
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3623127600955914e-05,
+      "loss": 0.6621,
+      "step": 11059
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3622034210547856e-05,
+      "loss": 0.679,
+      "step": 11060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3620940770297427e-05,
+      "loss": 0.5859,
+      "step": 11061
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3619847280219664e-05,
+      "loss": 0.545,
+      "step": 11062
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3618753740329629e-05,
+      "loss": 0.6345,
+      "step": 11063
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.361766015064236e-05,
+      "loss": 0.6562,
+      "step": 11064
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3616566511172908e-05,
+      "loss": 0.5534,
+      "step": 11065
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3615472821936324e-05,
+      "loss": 0.6244,
+      "step": 11066
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3614379082947657e-05,
+      "loss": 0.5945,
+      "step": 11067
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3613285294221959e-05,
+      "loss": 0.6815,
+      "step": 11068
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3612191455774277e-05,
+      "loss": 0.6022,
+      "step": 11069
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3611097567619671e-05,
+      "loss": 0.7118,
+      "step": 11070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3610003629773186e-05,
+      "loss": 0.7003,
+      "step": 11071
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3608909642249882e-05,
+      "loss": 0.6828,
+      "step": 11072
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3607815605064807e-05,
+      "loss": 0.5984,
+      "step": 11073
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3606721518233021e-05,
+      "loss": 0.6736,
+      "step": 11074
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3605627381769577e-05,
+      "loss": 0.6846,
+      "step": 11075
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3604533195689532e-05,
+      "loss": 0.6331,
+      "step": 11076
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3603438960007943e-05,
+      "loss": 0.6171,
+      "step": 11077
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3602344674739868e-05,
+      "loss": 0.6555,
+      "step": 11078
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3601250339900367e-05,
+      "loss": 0.6136,
+      "step": 11079
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3600155955504496e-05,
+      "loss": 0.6762,
+      "step": 11080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3599061521567317e-05,
+      "loss": 0.6513,
+      "step": 11081
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3597967038103885e-05,
+      "loss": 0.6486,
+      "step": 11082
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3596872505129269e-05,
+      "loss": 0.5954,
+      "step": 11083
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3595777922658524e-05,
+      "loss": 0.6411,
+      "step": 11084
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3594683290706714e-05,
+      "loss": 0.65,
+      "step": 11085
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3593588609288909e-05,
+      "loss": 0.59,
+      "step": 11086
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3592493878420161e-05,
+      "loss": 0.5861,
+      "step": 11087
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3591399098115545e-05,
+      "loss": 0.6405,
+      "step": 11088
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3590304268390118e-05,
+      "loss": 0.6151,
+      "step": 11089
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3589209389258953e-05,
+      "loss": 0.6778,
+      "step": 11090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3588114460737112e-05,
+      "loss": 0.6538,
+      "step": 11091
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3587019482839664e-05,
+      "loss": 0.5945,
+      "step": 11092
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3585924455581672e-05,
+      "loss": 0.6723,
+      "step": 11093
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3584829378978213e-05,
+      "loss": 0.5951,
+      "step": 11094
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3583734253044349e-05,
+      "loss": 0.6561,
+      "step": 11095
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3582639077795154e-05,
+      "loss": 0.5717,
+      "step": 11096
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3581543853245698e-05,
+      "loss": 0.544,
+      "step": 11097
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.358044857941105e-05,
+      "loss": 0.5741,
+      "step": 11098
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3579353256306287e-05,
+      "loss": 0.6163,
+      "step": 11099
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3578257883946473e-05,
+      "loss": 0.6685,
+      "step": 11100
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3577162462346692e-05,
+      "loss": 0.6624,
+      "step": 11101
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3576066991522008e-05,
+      "loss": 0.5836,
+      "step": 11102
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3574971471487502e-05,
+      "loss": 0.6234,
+      "step": 11103
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3573875902258244e-05,
+      "loss": 0.571,
+      "step": 11104
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3572780283849318e-05,
+      "loss": 0.5608,
+      "step": 11105
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3571684616275792e-05,
+      "loss": 0.6399,
+      "step": 11106
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357058889955275e-05,
+      "loss": 0.6565,
+      "step": 11107
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3569493133695266e-05,
+      "loss": 0.6306,
+      "step": 11108
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356839731871842e-05,
+      "loss": 0.6545,
+      "step": 11109
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3567301454637297e-05,
+      "loss": 0.5969,
+      "step": 11110
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3566205541466966e-05,
+      "loss": 0.6429,
+      "step": 11111
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3565109579222511e-05,
+      "loss": 0.6327,
+      "step": 11112
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356401356791902e-05,
+      "loss": 0.6337,
+      "step": 11113
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3562917507571566e-05,
+      "loss": 0.6851,
+      "step": 11114
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3561821398195242e-05,
+      "loss": 0.6169,
+      "step": 11115
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3560725239805121e-05,
+      "loss": 0.6095,
+      "step": 11116
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3559629032416297e-05,
+      "loss": 0.5748,
+      "step": 11117
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3558532776043846e-05,
+      "loss": 0.6325,
+      "step": 11118
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3557436470702858e-05,
+      "loss": 0.6875,
+      "step": 11119
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.355634011640842e-05,
+      "loss": 0.6496,
+      "step": 11120
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3555243713175617e-05,
+      "loss": 0.5883,
+      "step": 11121
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3554147261019535e-05,
+      "loss": 0.6552,
+      "step": 11122
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3553050759955262e-05,
+      "loss": 0.5828,
+      "step": 11123
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3551954209997891e-05,
+      "loss": 0.6449,
+      "step": 11124
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3550857611162508e-05,
+      "loss": 0.6891,
+      "step": 11125
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3549760963464205e-05,
+      "loss": 0.7289,
+      "step": 11126
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3548664266918072e-05,
+      "loss": 0.5403,
+      "step": 11127
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3547567521539202e-05,
+      "loss": 0.6158,
+      "step": 11128
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3546470727342682e-05,
+      "loss": 0.6166,
+      "step": 11129
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3545373884343614e-05,
+      "loss": 0.6203,
+      "step": 11130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.354427699255708e-05,
+      "loss": 0.7019,
+      "step": 11131
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3543180051998182e-05,
+      "loss": 0.5797,
+      "step": 11132
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3542083062682016e-05,
+      "loss": 0.6183,
+      "step": 11133
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3540986024623673e-05,
+      "loss": 0.6373,
+      "step": 11134
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3539888937838253e-05,
+      "loss": 0.6359,
+      "step": 11135
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3538791802340846e-05,
+      "loss": 0.6588,
+      "step": 11136
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353769461814656e-05,
+      "loss": 0.6436,
+      "step": 11137
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3536597385270482e-05,
+      "loss": 0.7012,
+      "step": 11138
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3535500103727722e-05,
+      "loss": 0.6087,
+      "step": 11139
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3534402773533368e-05,
+      "loss": 0.6423,
+      "step": 11140
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3533305394702532e-05,
+      "loss": 0.5498,
+      "step": 11141
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3532207967250305e-05,
+      "loss": 0.7108,
+      "step": 11142
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3531110491191793e-05,
+      "loss": 0.6052,
+      "step": 11143
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.35300129665421e-05,
+      "loss": 0.5259,
+      "step": 11144
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3528915393316327e-05,
+      "loss": 0.5899,
+      "step": 11145
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3527817771529577e-05,
+      "loss": 0.5694,
+      "step": 11146
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3526720101196953e-05,
+      "loss": 0.6797,
+      "step": 11147
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3525622382333562e-05,
+      "loss": 0.5863,
+      "step": 11148
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352452461495451e-05,
+      "loss": 0.7167,
+      "step": 11149
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3523426799074903e-05,
+      "loss": 0.568,
+      "step": 11150
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3522328934709847e-05,
+      "loss": 0.7459,
+      "step": 11151
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352123102187445e-05,
+      "loss": 0.5668,
+      "step": 11152
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3520133060583818e-05,
+      "loss": 0.6834,
+      "step": 11153
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3519035050853063e-05,
+      "loss": 0.7013,
+      "step": 11154
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3517936992697296e-05,
+      "loss": 0.5509,
+      "step": 11155
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.351683888613162e-05,
+      "loss": 0.6704,
+      "step": 11156
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3515740731171157e-05,
+      "loss": 0.5873,
+      "step": 11157
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3514642527831008e-05,
+      "loss": 0.6451,
+      "step": 11158
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3513544276126292e-05,
+      "loss": 0.7089,
+      "step": 11159
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3512445976072114e-05,
+      "loss": 0.6974,
+      "step": 11160
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.35113476276836e-05,
+      "loss": 0.6243,
+      "step": 11161
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3510249230975853e-05,
+      "loss": 0.55,
+      "step": 11162
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3509150785963994e-05,
+      "loss": 0.6121,
+      "step": 11163
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3508052292663137e-05,
+      "loss": 0.6225,
+      "step": 11164
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3506953751088396e-05,
+      "loss": 0.5747,
+      "step": 11165
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3505855161254893e-05,
+      "loss": 0.6428,
+      "step": 11166
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.350475652317774e-05,
+      "loss": 0.5831,
+      "step": 11167
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.350365783687206e-05,
+      "loss": 0.682,
+      "step": 11168
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3502559102352967e-05,
+      "loss": 0.6592,
+      "step": 11169
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3501460319635584e-05,
+      "loss": 0.6347,
+      "step": 11170
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.350036148873503e-05,
+      "loss": 0.6726,
+      "step": 11171
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3499262609666428e-05,
+      "loss": 0.6256,
+      "step": 11172
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3498163682444896e-05,
+      "loss": 0.5811,
+      "step": 11173
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3497064707085557e-05,
+      "loss": 0.6707,
+      "step": 11174
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3495965683603537e-05,
+      "loss": 0.717,
+      "step": 11175
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3494866612013956e-05,
+      "loss": 0.5954,
+      "step": 11176
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3493767492331942e-05,
+      "loss": 0.6229,
+      "step": 11177
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3492668324572615e-05,
+      "loss": 0.5345,
+      "step": 11178
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3491569108751106e-05,
+      "loss": 0.6043,
+      "step": 11179
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3490469844882535e-05,
+      "loss": 0.6475,
+      "step": 11180
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3489370532982034e-05,
+      "loss": 0.6336,
+      "step": 11181
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3488271173064725e-05,
+      "loss": 0.6251,
+      "step": 11182
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3487171765145743e-05,
+      "loss": 0.6021,
+      "step": 11183
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3486072309240212e-05,
+      "loss": 0.6975,
+      "step": 11184
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3484972805363262e-05,
+      "loss": 0.6985,
+      "step": 11185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3483873253530027e-05,
+      "loss": 0.6372,
+      "step": 11186
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.348277365375563e-05,
+      "loss": 0.687,
+      "step": 11187
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3481674006055212e-05,
+      "loss": 0.6921,
+      "step": 11188
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3480574310443897e-05,
+      "loss": 0.6327,
+      "step": 11189
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3479474566936825e-05,
+      "loss": 0.6422,
+      "step": 11190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3478374775549121e-05,
+      "loss": 0.6066,
+      "step": 11191
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3477274936295928e-05,
+      "loss": 0.6083,
+      "step": 11192
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3476175049192373e-05,
+      "loss": 0.6373,
+      "step": 11193
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3475075114253597e-05,
+      "loss": 0.7302,
+      "step": 11194
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3473975131494735e-05,
+      "loss": 0.6744,
+      "step": 11195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.347287510093092e-05,
+      "loss": 0.5888,
+      "step": 11196
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3471775022577294e-05,
+      "loss": 0.5473,
+      "step": 11197
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.347067489644899e-05,
+      "loss": 0.6414,
+      "step": 11198
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3469574722561154e-05,
+      "loss": 0.7048,
+      "step": 11199
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3468474500928919e-05,
+      "loss": 0.728,
+      "step": 11200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3467374231567427e-05,
+      "loss": 0.6328,
+      "step": 11201
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3466273914491818e-05,
+      "loss": 0.6085,
+      "step": 11202
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.346517354971724e-05,
+      "loss": 0.638,
+      "step": 11203
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3464073137258823e-05,
+      "loss": 0.6751,
+      "step": 11204
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3462972677131716e-05,
+      "loss": 0.6029,
+      "step": 11205
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3461872169351064e-05,
+      "loss": 0.6417,
+      "step": 11206
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3460771613932008e-05,
+      "loss": 0.6289,
+      "step": 11207
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3459671010889696e-05,
+      "loss": 0.7068,
+      "step": 11208
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3458570360239268e-05,
+      "loss": 0.6454,
+      "step": 11209
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3457469661995877e-05,
+      "loss": 0.6359,
+      "step": 11210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345636891617466e-05,
+      "loss": 0.6281,
+      "step": 11211
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3455268122790774e-05,
+      "loss": 0.5878,
+      "step": 11212
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3454167281859363e-05,
+      "loss": 0.6649,
+      "step": 11213
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3453066393395572e-05,
+      "loss": 0.5739,
+      "step": 11214
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3451965457414558e-05,
+      "loss": 0.6849,
+      "step": 11215
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345086447393146e-05,
+      "loss": 0.5723,
+      "step": 11216
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3449763442961438e-05,
+      "loss": 0.6545,
+      "step": 11217
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3448662364519638e-05,
+      "loss": 0.6781,
+      "step": 11218
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3447561238621216e-05,
+      "loss": 0.5742,
+      "step": 11219
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3446460065281318e-05,
+      "loss": 0.668,
+      "step": 11220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3445358844515105e-05,
+      "loss": 0.6683,
+      "step": 11221
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3444257576337728e-05,
+      "loss": 0.6968,
+      "step": 11222
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3443156260764336e-05,
+      "loss": 0.6382,
+      "step": 11223
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3442054897810093e-05,
+      "loss": 0.5791,
+      "step": 11224
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3440953487490145e-05,
+      "loss": 0.6662,
+      "step": 11225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3439852029819658e-05,
+      "loss": 0.7101,
+      "step": 11226
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3438750524813781e-05,
+      "loss": 0.5907,
+      "step": 11227
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3437648972487681e-05,
+      "loss": 0.6256,
+      "step": 11228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3436547372856503e-05,
+      "loss": 0.5971,
+      "step": 11229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3435445725935422e-05,
+      "loss": 0.6682,
+      "step": 11230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3434344031739581e-05,
+      "loss": 0.7021,
+      "step": 11231
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.343324229028415e-05,
+      "loss": 0.6437,
+      "step": 11232
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3432140501584293e-05,
+      "loss": 0.6072,
+      "step": 11233
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3431038665655163e-05,
+      "loss": 0.5624,
+      "step": 11234
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3429936782511927e-05,
+      "loss": 0.5492,
+      "step": 11235
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3428834852169747e-05,
+      "loss": 0.6111,
+      "step": 11236
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3427732874643788e-05,
+      "loss": 0.7046,
+      "step": 11237
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3426630849949212e-05,
+      "loss": 0.6261,
+      "step": 11238
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3425528778101184e-05,
+      "loss": 0.6682,
+      "step": 11239
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342442665911487e-05,
+      "loss": 0.6711,
+      "step": 11240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3423324493005438e-05,
+      "loss": 0.6111,
+      "step": 11241
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342222227978805e-05,
+      "loss": 0.6684,
+      "step": 11242
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3421120019477877e-05,
+      "loss": 0.6531,
+      "step": 11243
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342001771209009e-05,
+      "loss": 0.6351,
+      "step": 11244
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.341891535763985e-05,
+      "loss": 0.6834,
+      "step": 11245
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3417812956142333e-05,
+      "loss": 0.671,
+      "step": 11246
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3416710507612705e-05,
+      "loss": 0.6339,
+      "step": 11247
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.341560801206614e-05,
+      "loss": 0.6617,
+      "step": 11248
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3414505469517807e-05,
+      "loss": 0.607,
+      "step": 11249
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3413402879982878e-05,
+      "loss": 0.6412,
+      "step": 11250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3412300243476526e-05,
+      "loss": 0.6037,
+      "step": 11251
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3411197560013922e-05,
+      "loss": 0.6438,
+      "step": 11252
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3410094829610248e-05,
+      "loss": 0.5668,
+      "step": 11253
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3408992052280667e-05,
+      "loss": 0.5742,
+      "step": 11254
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3407889228040366e-05,
+      "loss": 0.5673,
+      "step": 11255
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3406786356904509e-05,
+      "loss": 0.6715,
+      "step": 11256
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3405683438888281e-05,
+      "loss": 0.612,
+      "step": 11257
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3404580474006856e-05,
+      "loss": 0.6715,
+      "step": 11258
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3403477462275412e-05,
+      "loss": 0.6276,
+      "step": 11259
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3402374403709126e-05,
+      "loss": 0.6372,
+      "step": 11260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3401271298323182e-05,
+      "loss": 0.6783,
+      "step": 11261
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3400168146132755e-05,
+      "loss": 0.5687,
+      "step": 11262
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3399064947153024e-05,
+      "loss": 0.6015,
+      "step": 11263
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3397961701399175e-05,
+      "loss": 0.601,
+      "step": 11264
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3396858408886386e-05,
+      "loss": 0.6058,
+      "step": 11265
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3395755069629845e-05,
+      "loss": 0.6294,
+      "step": 11266
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3394651683644727e-05,
+      "loss": 0.6582,
+      "step": 11267
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3393548250946217e-05,
+      "loss": 0.5786,
+      "step": 11268
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3392444771549504e-05,
+      "loss": 0.6578,
+      "step": 11269
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.339134124546977e-05,
+      "loss": 0.7229,
+      "step": 11270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33902376727222e-05,
+      "loss": 0.5667,
+      "step": 11271
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3389134053321981e-05,
+      "loss": 0.671,
+      "step": 11272
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3388030387284299e-05,
+      "loss": 0.6649,
+      "step": 11273
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3386926674624341e-05,
+      "loss": 0.6597,
+      "step": 11274
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3385822915357298e-05,
+      "loss": 0.6533,
+      "step": 11275
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3384719109498354e-05,
+      "loss": 0.6694,
+      "step": 11276
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3383615257062705e-05,
+      "loss": 0.6383,
+      "step": 11277
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3382511358065533e-05,
+      "loss": 0.5997,
+      "step": 11278
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3381407412522036e-05,
+      "loss": 0.653,
+      "step": 11279
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33803034204474e-05,
+      "loss": 0.6724,
+      "step": 11280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.337919938185682e-05,
+      "loss": 0.6803,
+      "step": 11281
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3378095296765485e-05,
+      "loss": 0.6619,
+      "step": 11282
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.337699116518859e-05,
+      "loss": 0.6573,
+      "step": 11283
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3375886987141331e-05,
+      "loss": 0.6543,
+      "step": 11284
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3374782762638901e-05,
+      "loss": 0.6957,
+      "step": 11285
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3373678491696495e-05,
+      "loss": 0.5903,
+      "step": 11286
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3372574174329306e-05,
+      "loss": 0.6781,
+      "step": 11287
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3371469810552538e-05,
+      "loss": 0.6394,
+      "step": 11288
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3370365400381377e-05,
+      "loss": 0.665,
+      "step": 11289
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3369260943831032e-05,
+      "loss": 0.6036,
+      "step": 11290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3368156440916693e-05,
+      "loss": 0.6597,
+      "step": 11291
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3367051891653564e-05,
+      "loss": 0.6074,
+      "step": 11292
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3365947296056838e-05,
+      "loss": 0.6052,
+      "step": 11293
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3364842654141723e-05,
+      "loss": 0.62,
+      "step": 11294
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3363737965923415e-05,
+      "loss": 0.6484,
+      "step": 11295
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336263323141712e-05,
+      "loss": 0.6405,
+      "step": 11296
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3361528450638036e-05,
+      "loss": 0.6002,
+      "step": 11297
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3360423623601361e-05,
+      "loss": 0.6322,
+      "step": 11298
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3359318750322312e-05,
+      "loss": 0.5772,
+      "step": 11299
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335821383081608e-05,
+      "loss": 0.5538,
+      "step": 11300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3357108865097881e-05,
+      "loss": 0.6981,
+      "step": 11301
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3356003853182908e-05,
+      "loss": 0.6217,
+      "step": 11302
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3354898795086374e-05,
+      "loss": 0.6137,
+      "step": 11303
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3353793690823489e-05,
+      "loss": 0.6704,
+      "step": 11304
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335268854040945e-05,
+      "loss": 0.6433,
+      "step": 11305
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3351583343859475e-05,
+      "loss": 0.5962,
+      "step": 11306
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3350478101188767e-05,
+      "loss": 0.6088,
+      "step": 11307
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3349372812412537e-05,
+      "loss": 0.6385,
+      "step": 11308
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3348267477545993e-05,
+      "loss": 0.6488,
+      "step": 11309
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3347162096604347e-05,
+      "loss": 0.5827,
+      "step": 11310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334605666960281e-05,
+      "loss": 0.6151,
+      "step": 11311
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3344951196556592e-05,
+      "loss": 0.6179,
+      "step": 11312
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3343845677480906e-05,
+      "loss": 0.6477,
+      "step": 11313
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3342740112390967e-05,
+      "loss": 0.657,
+      "step": 11314
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3341634501301985e-05,
+      "loss": 0.5928,
+      "step": 11315
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3340528844229177e-05,
+      "loss": 0.6739,
+      "step": 11316
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3339423141187759e-05,
+      "loss": 0.5824,
+      "step": 11317
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3338317392192942e-05,
+      "loss": 0.5888,
+      "step": 11318
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3337211597259945e-05,
+      "loss": 0.6894,
+      "step": 11319
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3336105756403982e-05,
+      "loss": 0.5698,
+      "step": 11320
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3334999869640274e-05,
+      "loss": 0.6843,
+      "step": 11321
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3333893936984042e-05,
+      "loss": 0.543,
+      "step": 11322
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3332787958450492e-05,
+      "loss": 0.6489,
+      "step": 11323
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3331681934054857e-05,
+      "loss": 0.6739,
+      "step": 11324
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3330575863812347e-05,
+      "loss": 0.6415,
+      "step": 11325
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.332946974773819e-05,
+      "loss": 0.6389,
+      "step": 11326
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3328363585847602e-05,
+      "loss": 0.6231,
+      "step": 11327
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3327257378155807e-05,
+      "loss": 0.7237,
+      "step": 11328
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3326151124678026e-05,
+      "loss": 0.6452,
+      "step": 11329
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3325044825429483e-05,
+      "loss": 0.5958,
+      "step": 11330
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3323938480425404e-05,
+      "loss": 0.6091,
+      "step": 11331
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3322832089681009e-05,
+      "loss": 0.6353,
+      "step": 11332
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3321725653211526e-05,
+      "loss": 0.6277,
+      "step": 11333
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3320619171032176e-05,
+      "loss": 0.6357,
+      "step": 11334
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3319512643158193e-05,
+      "loss": 0.6304,
+      "step": 11335
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3318406069604794e-05,
+      "loss": 0.6232,
+      "step": 11336
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3317299450387216e-05,
+      "loss": 0.6294,
+      "step": 11337
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.331619278552068e-05,
+      "loss": 0.6342,
+      "step": 11338
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.331508607502042e-05,
+      "loss": 0.6149,
+      "step": 11339
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.331397931890166e-05,
+      "loss": 0.6546,
+      "step": 11340
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3312872517179633e-05,
+      "loss": 0.5961,
+      "step": 11341
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3311765669869572e-05,
+      "loss": 0.6185,
+      "step": 11342
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3310658776986702e-05,
+      "loss": 0.5524,
+      "step": 11343
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3309551838546261e-05,
+      "loss": 0.672,
+      "step": 11344
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3308444854563477e-05,
+      "loss": 0.6213,
+      "step": 11345
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3307337825053586e-05,
+      "loss": 0.6139,
+      "step": 11346
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3306230750031818e-05,
+      "loss": 0.7288,
+      "step": 11347
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3305123629513413e-05,
+      "loss": 0.5868,
+      "step": 11348
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33040164635136e-05,
+      "loss": 0.5938,
+      "step": 11349
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.330290925204762e-05,
+      "loss": 0.6655,
+      "step": 11350
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3301801995130703e-05,
+      "loss": 0.6058,
+      "step": 11351
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3300694692778091e-05,
+      "loss": 0.6381,
+      "step": 11352
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.329958734500502e-05,
+      "loss": 0.6607,
+      "step": 11353
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3298479951826727e-05,
+      "loss": 0.7484,
+      "step": 11354
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3297372513258455e-05,
+      "loss": 0.6375,
+      "step": 11355
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3296265029315435e-05,
+      "loss": 0.6287,
+      "step": 11356
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3295157500012917e-05,
+      "loss": 0.5964,
+      "step": 11357
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3294049925366133e-05,
+      "loss": 0.6476,
+      "step": 11358
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3292942305390329e-05,
+      "loss": 0.6586,
+      "step": 11359
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3291834640100745e-05,
+      "loss": 0.5405,
+      "step": 11360
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.329072692951262e-05,
+      "loss": 0.6327,
+      "step": 11361
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3289619173641208e-05,
+      "loss": 0.58,
+      "step": 11362
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3288511372501741e-05,
+      "loss": 0.5596,
+      "step": 11363
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3287403526109471e-05,
+      "loss": 0.655,
+      "step": 11364
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3286295634479636e-05,
+      "loss": 0.6916,
+      "step": 11365
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3285187697627488e-05,
+      "loss": 0.6867,
+      "step": 11366
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.328407971556827e-05,
+      "loss": 0.6471,
+      "step": 11367
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.328297168831723e-05,
+      "loss": 0.6504,
+      "step": 11368
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3281863615889614e-05,
+      "loss": 0.6281,
+      "step": 11369
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3280755498300676e-05,
+      "loss": 0.6811,
+      "step": 11370
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327964733556565e-05,
+      "loss": 0.6707,
+      "step": 11371
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3278539127699802e-05,
+      "loss": 0.5602,
+      "step": 11372
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3277430874718373e-05,
+      "loss": 0.5474,
+      "step": 11373
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3276322576636616e-05,
+      "loss": 0.5732,
+      "step": 11374
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327521423346978e-05,
+      "loss": 0.6993,
+      "step": 11375
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3274105845233118e-05,
+      "loss": 0.6334,
+      "step": 11376
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3272997411941885e-05,
+      "loss": 0.6889,
+      "step": 11377
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327188893361133e-05,
+      "loss": 0.6841,
+      "step": 11378
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327078041025671e-05,
+      "loss": 0.6772,
+      "step": 11379
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3269671841893275e-05,
+      "loss": 0.6183,
+      "step": 11380
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3268563228536287e-05,
+      "loss": 0.5629,
+      "step": 11381
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.326745457020099e-05,
+      "loss": 0.6304,
+      "step": 11382
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3266345866902653e-05,
+      "loss": 0.6575,
+      "step": 11383
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3265237118656525e-05,
+      "loss": 0.6303,
+      "step": 11384
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3264128325477864e-05,
+      "loss": 0.5878,
+      "step": 11385
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.326301948738193e-05,
+      "loss": 0.6753,
+      "step": 11386
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3261910604383981e-05,
+      "loss": 0.6029,
+      "step": 11387
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3260801676499275e-05,
+      "loss": 0.6608,
+      "step": 11388
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3259692703743073e-05,
+      "loss": 0.6136,
+      "step": 11389
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3258583686130637e-05,
+      "loss": 0.603,
+      "step": 11390
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3257474623677224e-05,
+      "loss": 0.749,
+      "step": 11391
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3256365516398098e-05,
+      "loss": 0.6874,
+      "step": 11392
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3255256364308525e-05,
+      "loss": 0.6423,
+      "step": 11393
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3254147167423759e-05,
+      "loss": 0.6464,
+      "step": 11394
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325303792575907e-05,
+      "loss": 0.6003,
+      "step": 11395
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325192863932972e-05,
+      "loss": 0.6603,
+      "step": 11396
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325081930815098e-05,
+      "loss": 0.6534,
+      "step": 11397
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3249709932238105e-05,
+      "loss": 0.6904,
+      "step": 11398
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3248600511606366e-05,
+      "loss": 0.5965,
+      "step": 11399
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.324749104627103e-05,
+      "loss": 0.6507,
+      "step": 11400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3246381536247364e-05,
+      "loss": 0.6465,
+      "step": 11401
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3245271981550637e-05,
+      "loss": 0.5593,
+      "step": 11402
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3244162382196111e-05,
+      "loss": 0.5613,
+      "step": 11403
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3243052738199067e-05,
+      "loss": 0.5993,
+      "step": 11404
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3241943049574762e-05,
+      "loss": 0.5742,
+      "step": 11405
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3240833316338475e-05,
+      "loss": 0.6714,
+      "step": 11406
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3239723538505473e-05,
+      "loss": 0.6553,
+      "step": 11407
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3238613716091028e-05,
+      "loss": 0.62,
+      "step": 11408
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3237503849110414e-05,
+      "loss": 0.6367,
+      "step": 11409
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3236393937578902e-05,
+      "loss": 0.7296,
+      "step": 11410
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3235283981511764e-05,
+      "loss": 0.5691,
+      "step": 11411
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.323417398092428e-05,
+      "loss": 0.5465,
+      "step": 11412
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3233063935831717e-05,
+      "loss": 0.6406,
+      "step": 11413
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.323195384624935e-05,
+      "loss": 0.666,
+      "step": 11414
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3230843712192463e-05,
+      "loss": 0.6358,
+      "step": 11415
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3229733533676325e-05,
+      "loss": 0.6564,
+      "step": 11416
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.322862331071622e-05,
+      "loss": 0.6139,
+      "step": 11417
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3227513043327416e-05,
+      "loss": 0.6818,
+      "step": 11418
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.32264027315252e-05,
+      "loss": 0.5861,
+      "step": 11419
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3225292375324848e-05,
+      "loss": 0.6907,
+      "step": 11420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3224181974741637e-05,
+      "loss": 0.5661,
+      "step": 11421
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3223071529790848e-05,
+      "loss": 0.5855,
+      "step": 11422
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3221961040487763e-05,
+      "loss": 0.5962,
+      "step": 11423
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3220850506847667e-05,
+      "loss": 0.5957,
+      "step": 11424
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3219739928885834e-05,
+      "loss": 0.5972,
+      "step": 11425
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3218629306617552e-05,
+      "loss": 0.5585,
+      "step": 11426
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3217518640058103e-05,
+      "loss": 0.6252,
+      "step": 11427
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321640792922277e-05,
+      "loss": 0.6898,
+      "step": 11428
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3215297174126835e-05,
+      "loss": 0.6104,
+      "step": 11429
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3214186374785589e-05,
+      "loss": 0.6922,
+      "step": 11430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3213075531214315e-05,
+      "loss": 0.6,
+      "step": 11431
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3211964643428295e-05,
+      "loss": 0.6117,
+      "step": 11432
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321085371144282e-05,
+      "loss": 0.6281,
+      "step": 11433
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3209742735273179e-05,
+      "loss": 0.6146,
+      "step": 11434
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3208631714934655e-05,
+      "loss": 0.6099,
+      "step": 11435
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.320752065044254e-05,
+      "loss": 0.6651,
+      "step": 11436
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3206409541812123e-05,
+      "loss": 0.704,
+      "step": 11437
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3205298389058694e-05,
+      "loss": 0.5917,
+      "step": 11438
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.320418719219754e-05,
+      "loss": 0.6746,
+      "step": 11439
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3203075951243957e-05,
+      "loss": 0.6231,
+      "step": 11440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3201964666213234e-05,
+      "loss": 0.6258,
+      "step": 11441
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3200853337120665e-05,
+      "loss": 0.6533,
+      "step": 11442
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3199741963981538e-05,
+      "loss": 0.6373,
+      "step": 11443
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3198630546811155e-05,
+      "loss": 0.5558,
+      "step": 11444
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3197519085624801e-05,
+      "loss": 0.6515,
+      "step": 11445
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3196407580437777e-05,
+      "loss": 0.5488,
+      "step": 11446
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3195296031265375e-05,
+      "loss": 0.5644,
+      "step": 11447
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3194184438122894e-05,
+      "loss": 0.6836,
+      "step": 11448
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3193072801025625e-05,
+      "loss": 0.6429,
+      "step": 11449
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.319196111998887e-05,
+      "loss": 0.648,
+      "step": 11450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3190849395027926e-05,
+      "loss": 0.7103,
+      "step": 11451
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.318973762615809e-05,
+      "loss": 0.6763,
+      "step": 11452
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3188625813394663e-05,
+      "loss": 0.6587,
+      "step": 11453
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.318751395675294e-05,
+      "loss": 0.5941,
+      "step": 11454
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.318640205624823e-05,
+      "loss": 0.6077,
+      "step": 11455
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.318529011189582e-05,
+      "loss": 0.6033,
+      "step": 11456
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3184178123711025e-05,
+      "loss": 0.6352,
+      "step": 11457
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3183066091709138e-05,
+      "loss": 0.6006,
+      "step": 11458
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3181954015905467e-05,
+      "loss": 0.6401,
+      "step": 11459
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3180841896315311e-05,
+      "loss": 0.6263,
+      "step": 11460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3179729732953973e-05,
+      "loss": 0.5592,
+      "step": 11461
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3178617525836765e-05,
+      "loss": 0.6274,
+      "step": 11462
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3177505274978986e-05,
+      "loss": 0.6326,
+      "step": 11463
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3176392980395941e-05,
+      "loss": 0.6722,
+      "step": 11464
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3175280642102939e-05,
+      "loss": 0.6058,
+      "step": 11465
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3174168260115286e-05,
+      "loss": 0.6287,
+      "step": 11466
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3173055834448285e-05,
+      "loss": 0.5316,
+      "step": 11467
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.317194336511725e-05,
+      "loss": 0.6271,
+      "step": 11468
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3170830852137488e-05,
+      "loss": 0.601,
+      "step": 11469
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3169718295524306e-05,
+      "loss": 0.6889,
+      "step": 11470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3168605695293015e-05,
+      "loss": 0.6087,
+      "step": 11471
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3167493051458923e-05,
+      "loss": 0.6387,
+      "step": 11472
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.316638036403735e-05,
+      "loss": 0.6829,
+      "step": 11473
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3165267633043594e-05,
+      "loss": 0.6357,
+      "step": 11474
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3164154858492979e-05,
+      "loss": 0.6434,
+      "step": 11475
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3163042040400811e-05,
+      "loss": 0.6282,
+      "step": 11476
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3161929178782407e-05,
+      "loss": 0.6239,
+      "step": 11477
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3160816273653077e-05,
+      "loss": 0.5866,
+      "step": 11478
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315970332502814e-05,
+      "loss": 0.604,
+      "step": 11479
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3158590332922904e-05,
+      "loss": 0.6111,
+      "step": 11480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3157477297352693e-05,
+      "loss": 0.6831,
+      "step": 11481
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315636421833282e-05,
+      "loss": 0.5782,
+      "step": 11482
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.31552510958786e-05,
+      "loss": 0.6693,
+      "step": 11483
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3154137930005357e-05,
+      "loss": 0.5899,
+      "step": 11484
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.31530247207284e-05,
+      "loss": 0.6237,
+      "step": 11485
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3151911468063056e-05,
+      "loss": 0.686,
+      "step": 11486
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3150798172024639e-05,
+      "loss": 0.6356,
+      "step": 11487
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3149684832628471e-05,
+      "loss": 0.6016,
+      "step": 11488
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3148571449889872e-05,
+      "loss": 0.5923,
+      "step": 11489
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3147458023824163e-05,
+      "loss": 0.6001,
+      "step": 11490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3146344554446664e-05,
+      "loss": 0.5921,
+      "step": 11491
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3145231041772701e-05,
+      "loss": 0.6633,
+      "step": 11492
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3144117485817595e-05,
+      "loss": 0.5897,
+      "step": 11493
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.314300388659667e-05,
+      "loss": 0.6122,
+      "step": 11494
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3141890244125253e-05,
+      "loss": 0.6289,
+      "step": 11495
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.314077655841866e-05,
+      "loss": 0.5684,
+      "step": 11496
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3139662829492226e-05,
+      "loss": 0.6774,
+      "step": 11497
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.313854905736127e-05,
+      "loss": 0.756,
+      "step": 11498
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3137435242041123e-05,
+      "loss": 0.6795,
+      "step": 11499
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3136321383547109e-05,
+      "loss": 0.5396,
+      "step": 11500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3135207481894557e-05,
+      "loss": 0.6289,
+      "step": 11501
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3134093537098797e-05,
+      "loss": 0.578,
+      "step": 11502
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3132979549175153e-05,
+      "loss": 0.6164,
+      "step": 11503
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.313186551813896e-05,
+      "loss": 0.6868,
+      "step": 11504
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3130751444005546e-05,
+      "loss": 0.5685,
+      "step": 11505
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.312963732679024e-05,
+      "loss": 0.6575,
+      "step": 11506
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3128523166508374e-05,
+      "loss": 0.5755,
+      "step": 11507
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.312740896317528e-05,
+      "loss": 0.6088,
+      "step": 11508
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3126294716806292e-05,
+      "loss": 0.594,
+      "step": 11509
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3125180427416741e-05,
+      "loss": 0.6111,
+      "step": 11510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.312406609502196e-05,
+      "loss": 0.587,
+      "step": 11511
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3122951719637287e-05,
+      "loss": 0.5582,
+      "step": 11512
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3121837301278055e-05,
+      "loss": 0.6158,
+      "step": 11513
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3120722839959595e-05,
+      "loss": 0.5718,
+      "step": 11514
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3119608335697249e-05,
+      "loss": 0.6466,
+      "step": 11515
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3118493788506351e-05,
+      "loss": 0.6097,
+      "step": 11516
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311737919840224e-05,
+      "loss": 0.6178,
+      "step": 11517
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3116264565400247e-05,
+      "loss": 0.6682,
+      "step": 11518
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3115149889515722e-05,
+      "loss": 0.6233,
+      "step": 11519
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311403517076399e-05,
+      "loss": 0.5963,
+      "step": 11520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3112920409160403e-05,
+      "loss": 0.6869,
+      "step": 11521
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3111805604720296e-05,
+      "loss": 0.5713,
+      "step": 11522
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3110690757459005e-05,
+      "loss": 0.6042,
+      "step": 11523
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3109575867391881e-05,
+      "loss": 0.6421,
+      "step": 11524
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3108460934534257e-05,
+      "loss": 0.6249,
+      "step": 11525
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3107345958901484e-05,
+      "loss": 0.66,
+      "step": 11526
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3106230940508895e-05,
+      "loss": 0.6184,
+      "step": 11527
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3105115879371844e-05,
+      "loss": 0.7085,
+      "step": 11528
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3104000775505665e-05,
+      "loss": 0.5893,
+      "step": 11529
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3102885628925712e-05,
+      "loss": 0.6167,
+      "step": 11530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3101770439647325e-05,
+      "loss": 0.7237,
+      "step": 11531
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.310065520768585e-05,
+      "loss": 0.7003,
+      "step": 11532
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3099539933056638e-05,
+      "loss": 0.6232,
+      "step": 11533
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3098424615775027e-05,
+      "loss": 0.5832,
+      "step": 11534
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3097309255856377e-05,
+      "loss": 0.6458,
+      "step": 11535
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3096193853316027e-05,
+      "loss": 0.6474,
+      "step": 11536
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.309507840816933e-05,
+      "loss": 0.6052,
+      "step": 11537
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3093962920431633e-05,
+      "loss": 0.6225,
+      "step": 11538
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3092847390118288e-05,
+      "loss": 0.5946,
+      "step": 11539
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3091731817244647e-05,
+      "loss": 0.6408,
+      "step": 11540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3090616201826056e-05,
+      "loss": 0.5878,
+      "step": 11541
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3089500543877876e-05,
+      "loss": 0.5699,
+      "step": 11542
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3088384843415449e-05,
+      "loss": 0.5665,
+      "step": 11543
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3087269100454136e-05,
+      "loss": 0.571,
+      "step": 11544
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3086153315009285e-05,
+      "loss": 0.6711,
+      "step": 11545
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3085037487096255e-05,
+      "loss": 0.6143,
+      "step": 11546
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3083921616730396e-05,
+      "loss": 0.6484,
+      "step": 11547
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3082805703927066e-05,
+      "loss": 0.6418,
+      "step": 11548
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3081689748701623e-05,
+      "loss": 0.5847,
+      "step": 11549
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.308057375106942e-05,
+      "loss": 0.605,
+      "step": 11550
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3079457711045815e-05,
+      "loss": 0.6661,
+      "step": 11551
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3078341628646168e-05,
+      "loss": 0.6409,
+      "step": 11552
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3077225503885834e-05,
+      "loss": 0.6182,
+      "step": 11553
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3076109336780172e-05,
+      "loss": 0.6444,
+      "step": 11554
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3074993127344545e-05,
+      "loss": 0.5664,
+      "step": 11555
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3073876875594307e-05,
+      "loss": 0.6293,
+      "step": 11556
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3072760581544826e-05,
+      "loss": 0.6048,
+      "step": 11557
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3071644245211456e-05,
+      "loss": 0.6308,
+      "step": 11558
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3070527866609564e-05,
+      "loss": 0.5823,
+      "step": 11559
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306941144575451e-05,
+      "loss": 0.6478,
+      "step": 11560
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3068294982661659e-05,
+      "loss": 0.6142,
+      "step": 11561
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3067178477346372e-05,
+      "loss": 0.7263,
+      "step": 11562
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3066061929824013e-05,
+      "loss": 0.6446,
+      "step": 11563
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306494534010995e-05,
+      "loss": 0.6679,
+      "step": 11564
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3063828708219543e-05,
+      "loss": 0.5857,
+      "step": 11565
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3062712034168166e-05,
+      "loss": 0.5626,
+      "step": 11566
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3061595317971176e-05,
+      "loss": 0.5805,
+      "step": 11567
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3060478559643946e-05,
+      "loss": 0.5665,
+      "step": 11568
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305936175920184e-05,
+      "loss": 0.6136,
+      "step": 11569
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3058244916660228e-05,
+      "loss": 0.6936,
+      "step": 11570
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305712803203448e-05,
+      "loss": 0.6465,
+      "step": 11571
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3056011105339965e-05,
+      "loss": 0.67,
+      "step": 11572
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3054894136592052e-05,
+      "loss": 0.5802,
+      "step": 11573
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3053777125806108e-05,
+      "loss": 0.6317,
+      "step": 11574
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3052660072997511e-05,
+      "loss": 0.628,
+      "step": 11575
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3051542978181629e-05,
+      "loss": 0.5859,
+      "step": 11576
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3050425841373833e-05,
+      "loss": 0.6238,
+      "step": 11577
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3049308662589498e-05,
+      "loss": 0.6759,
+      "step": 11578
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3048191441843993e-05,
+      "loss": 0.6527,
+      "step": 11579
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30470741791527e-05,
+      "loss": 0.6592,
+      "step": 11580
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3045956874530986e-05,
+      "loss": 0.6265,
+      "step": 11581
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3044839527994231e-05,
+      "loss": 0.5504,
+      "step": 11582
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3043722139557806e-05,
+      "loss": 0.6461,
+      "step": 11583
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3042604709237092e-05,
+      "loss": 0.6217,
+      "step": 11584
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3041487237047462e-05,
+      "loss": 0.6155,
+      "step": 11585
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30403697230043e-05,
+      "loss": 0.7082,
+      "step": 11586
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3039252167122973e-05,
+      "loss": 0.6824,
+      "step": 11587
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303813456941887e-05,
+      "loss": 0.6499,
+      "step": 11588
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3037016929907362e-05,
+      "loss": 0.6465,
+      "step": 11589
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3035899248603837e-05,
+      "loss": 0.6682,
+      "step": 11590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3034781525523668e-05,
+      "loss": 0.6092,
+      "step": 11591
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3033663760682239e-05,
+      "loss": 0.6599,
+      "step": 11592
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3032545954094931e-05,
+      "loss": 0.6569,
+      "step": 11593
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3031428105777124e-05,
+      "loss": 0.6591,
+      "step": 11594
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3030310215744206e-05,
+      "loss": 0.6926,
+      "step": 11595
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3029192284011554e-05,
+      "loss": 0.577,
+      "step": 11596
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3028074310594558e-05,
+      "loss": 0.5759,
+      "step": 11597
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3026956295508594e-05,
+      "loss": 0.6389,
+      "step": 11598
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3025838238769058e-05,
+      "loss": 0.6562,
+      "step": 11599
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3024720140391322e-05,
+      "loss": 0.7065,
+      "step": 11600
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3023602000390783e-05,
+      "loss": 0.5725,
+      "step": 11601
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3022483818782823e-05,
+      "loss": 0.6271,
+      "step": 11602
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.302136559558283e-05,
+      "loss": 0.6257,
+      "step": 11603
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3020247330806193e-05,
+      "loss": 0.6807,
+      "step": 11604
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3019129024468295e-05,
+      "loss": 0.6152,
+      "step": 11605
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3018010676584533e-05,
+      "loss": 0.6379,
+      "step": 11606
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.301689228717029e-05,
+      "loss": 0.6571,
+      "step": 11607
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.301577385624096e-05,
+      "loss": 0.629,
+      "step": 11608
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3014655383811929e-05,
+      "loss": 0.6027,
+      "step": 11609
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3013536869898593e-05,
+      "loss": 0.6394,
+      "step": 11610
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3012418314516342e-05,
+      "loss": 0.6166,
+      "step": 11611
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3011299717680568e-05,
+      "loss": 0.6131,
+      "step": 11612
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3010181079406663e-05,
+      "loss": 0.6068,
+      "step": 11613
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3009062399710021e-05,
+      "loss": 0.6639,
+      "step": 11614
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3007943678606038e-05,
+      "loss": 0.718,
+      "step": 11615
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3006824916110105e-05,
+      "loss": 0.7466,
+      "step": 11616
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3005706112237622e-05,
+      "loss": 0.6914,
+      "step": 11617
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3004587267003979e-05,
+      "loss": 0.6596,
+      "step": 11618
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3003468380424578e-05,
+      "loss": 0.5822,
+      "step": 11619
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.300234945251481e-05,
+      "loss": 0.6393,
+      "step": 11620
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3001230483290076e-05,
+      "loss": 0.6276,
+      "step": 11621
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3000111472765776e-05,
+      "loss": 0.6158,
+      "step": 11622
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2998992420957301e-05,
+      "loss": 0.588,
+      "step": 11623
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299787332788006e-05,
+      "loss": 0.5594,
+      "step": 11624
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2996754193549446e-05,
+      "loss": 0.5807,
+      "step": 11625
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2995635017980863e-05,
+      "loss": 0.6369,
+      "step": 11626
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2994515801189706e-05,
+      "loss": 0.6669,
+      "step": 11627
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2993396543191383e-05,
+      "loss": 0.5996,
+      "step": 11628
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2992277244001294e-05,
+      "loss": 0.7088,
+      "step": 11629
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299115790363484e-05,
+      "loss": 0.7241,
+      "step": 11630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2990038522107424e-05,
+      "loss": 0.598,
+      "step": 11631
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.298891909943445e-05,
+      "loss": 0.6664,
+      "step": 11632
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2987799635631327e-05,
+      "loss": 0.6552,
+      "step": 11633
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2986680130713451e-05,
+      "loss": 0.6441,
+      "step": 11634
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2985560584696236e-05,
+      "loss": 0.6155,
+      "step": 11635
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.298444099759508e-05,
+      "loss": 0.6697,
+      "step": 11636
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2983321369425397e-05,
+      "loss": 0.6151,
+      "step": 11637
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.298220170020259e-05,
+      "loss": 0.6482,
+      "step": 11638
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2981081989942066e-05,
+      "loss": 0.6724,
+      "step": 11639
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2979962238659236e-05,
+      "loss": 0.6463,
+      "step": 11640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2978842446369506e-05,
+      "loss": 0.6524,
+      "step": 11641
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2977722613088286e-05,
+      "loss": 0.6171,
+      "step": 11642
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2976602738830986e-05,
+      "loss": 0.6279,
+      "step": 11643
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2975482823613019e-05,
+      "loss": 0.6059,
+      "step": 11644
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2974362867449792e-05,
+      "loss": 0.6227,
+      "step": 11645
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.297324287035672e-05,
+      "loss": 0.6018,
+      "step": 11646
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2972122832349212e-05,
+      "loss": 0.7098,
+      "step": 11647
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2971002753442683e-05,
+      "loss": 0.5564,
+      "step": 11648
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2969882633652546e-05,
+      "loss": 0.642,
+      "step": 11649
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2968762472994212e-05,
+      "loss": 0.6298,
+      "step": 11650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29676422714831e-05,
+      "loss": 0.5525,
+      "step": 11651
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2966522029134622e-05,
+      "loss": 0.6858,
+      "step": 11652
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2965401745964197e-05,
+      "loss": 0.6412,
+      "step": 11653
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2964281421987235e-05,
+      "loss": 0.6776,
+      "step": 11654
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2963161057219157e-05,
+      "loss": 0.5994,
+      "step": 11655
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.296204065167538e-05,
+      "loss": 0.6344,
+      "step": 11656
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2960920205371322e-05,
+      "loss": 0.6068,
+      "step": 11657
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2959799718322396e-05,
+      "loss": 0.6139,
+      "step": 11658
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2958679190544028e-05,
+      "loss": 0.5965,
+      "step": 11659
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2957558622051633e-05,
+      "loss": 0.5777,
+      "step": 11660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2956438012860634e-05,
+      "loss": 0.6189,
+      "step": 11661
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2955317362986453e-05,
+      "loss": 0.5775,
+      "step": 11662
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2954196672444503e-05,
+      "loss": 0.5841,
+      "step": 11663
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2953075941250216e-05,
+      "loss": 0.522,
+      "step": 11664
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2951955169419004e-05,
+      "loss": 0.5747,
+      "step": 11665
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29508343569663e-05,
+      "loss": 0.6742,
+      "step": 11666
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2949713503907516e-05,
+      "loss": 0.6021,
+      "step": 11667
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2948592610258088e-05,
+      "loss": 0.6176,
+      "step": 11668
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2947471676033434e-05,
+      "loss": 0.685,
+      "step": 11669
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2946350701248977e-05,
+      "loss": 0.6759,
+      "step": 11670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2945229685920149e-05,
+      "loss": 0.5978,
+      "step": 11671
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.294410863006237e-05,
+      "loss": 0.6099,
+      "step": 11672
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2942987533691071e-05,
+      "loss": 0.658,
+      "step": 11673
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2941866396821676e-05,
+      "loss": 0.6855,
+      "step": 11674
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2940745219469619e-05,
+      "loss": 0.7059,
+      "step": 11675
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.293962400165032e-05,
+      "loss": 0.5989,
+      "step": 11676
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2938502743379212e-05,
+      "loss": 0.6678,
+      "step": 11677
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2937381444671723e-05,
+      "loss": 0.6201,
+      "step": 11678
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2936260105543286e-05,
+      "loss": 0.623,
+      "step": 11679
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2935138726009328e-05,
+      "loss": 0.5641,
+      "step": 11680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2934017306085284e-05,
+      "loss": 0.6043,
+      "step": 11681
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2932895845786587e-05,
+      "loss": 0.5961,
+      "step": 11682
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2931774345128661e-05,
+      "loss": 0.6559,
+      "step": 11683
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2930652804126944e-05,
+      "loss": 0.6603,
+      "step": 11684
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2929531222796872e-05,
+      "loss": 0.6747,
+      "step": 11685
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2928409601153875e-05,
+      "loss": 0.5914,
+      "step": 11686
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2927287939213388e-05,
+      "loss": 0.5704,
+      "step": 11687
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2926166236990847e-05,
+      "loss": 0.6407,
+      "step": 11688
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2925044494501688e-05,
+      "loss": 0.6026,
+      "step": 11689
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2923922711761347e-05,
+      "loss": 0.5836,
+      "step": 11690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292280088878526e-05,
+      "loss": 0.6664,
+      "step": 11691
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2921679025588865e-05,
+      "loss": 0.6161,
+      "step": 11692
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29205571221876e-05,
+      "loss": 0.6022,
+      "step": 11693
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2919435178596901e-05,
+      "loss": 0.551,
+      "step": 11694
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.291831319483221e-05,
+      "loss": 0.5916,
+      "step": 11695
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2917191170908964e-05,
+      "loss": 0.6456,
+      "step": 11696
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2916069106842607e-05,
+      "loss": 0.5796,
+      "step": 11697
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2914947002648573e-05,
+      "loss": 0.5761,
+      "step": 11698
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2913824858342308e-05,
+      "loss": 0.6253,
+      "step": 11699
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2912702673939252e-05,
+      "loss": 0.6354,
+      "step": 11700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.291158044945485e-05,
+      "loss": 0.6003,
+      "step": 11701
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2910458184904542e-05,
+      "loss": 0.6247,
+      "step": 11702
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2909335880303772e-05,
+      "loss": 0.6306,
+      "step": 11703
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2908213535667985e-05,
+      "loss": 0.6091,
+      "step": 11704
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2907091151012621e-05,
+      "loss": 0.6785,
+      "step": 11705
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2905968726353133e-05,
+      "loss": 0.6706,
+      "step": 11706
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2904846261704957e-05,
+      "loss": 0.6222,
+      "step": 11707
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2903723757083548e-05,
+      "loss": 0.6585,
+      "step": 11708
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2902601212504346e-05,
+      "loss": 0.6777,
+      "step": 11709
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2901478627982802e-05,
+      "loss": 0.6054,
+      "step": 11710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2900356003534364e-05,
+      "loss": 0.6484,
+      "step": 11711
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2899233339174474e-05,
+      "loss": 0.6864,
+      "step": 11712
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2898110634918592e-05,
+      "loss": 0.664,
+      "step": 11713
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2896987890782158e-05,
+      "loss": 0.5807,
+      "step": 11714
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2895865106780625e-05,
+      "loss": 0.5728,
+      "step": 11715
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2894742282929444e-05,
+      "loss": 0.5484,
+      "step": 11716
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2893619419244068e-05,
+      "loss": 0.596,
+      "step": 11717
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2892496515739943e-05,
+      "loss": 0.6681,
+      "step": 11718
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2891373572432524e-05,
+      "loss": 0.6045,
+      "step": 11719
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2890250589337265e-05,
+      "loss": 0.6476,
+      "step": 11720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288912756646962e-05,
+      "loss": 0.6324,
+      "step": 11721
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288800450384504e-05,
+      "loss": 0.6178,
+      "step": 11722
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2886881401478978e-05,
+      "loss": 0.5914,
+      "step": 11723
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2885758259386896e-05,
+      "loss": 0.6346,
+      "step": 11724
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288463507758424e-05,
+      "loss": 0.7245,
+      "step": 11725
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2883511856086475e-05,
+      "loss": 0.644,
+      "step": 11726
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288238859490905e-05,
+      "loss": 0.6422,
+      "step": 11727
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2881265294067426e-05,
+      "loss": 0.5944,
+      "step": 11728
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288014195357706e-05,
+      "loss": 0.6039,
+      "step": 11729
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.287901857345341e-05,
+      "loss": 0.6968,
+      "step": 11730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2877895153711935e-05,
+      "loss": 0.659,
+      "step": 11731
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2876771694368094e-05,
+      "loss": 0.6705,
+      "step": 11732
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2875648195437347e-05,
+      "loss": 0.6248,
+      "step": 11733
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2874524656935153e-05,
+      "loss": 0.6275,
+      "step": 11734
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2873401078876974e-05,
+      "loss": 0.5869,
+      "step": 11735
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2872277461278272e-05,
+      "loss": 0.6058,
+      "step": 11736
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2871153804154511e-05,
+      "loss": 0.6459,
+      "step": 11737
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2870030107521145e-05,
+      "loss": 0.6736,
+      "step": 11738
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2868906371393648e-05,
+      "loss": 0.752,
+      "step": 11739
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2867782595787474e-05,
+      "loss": 0.5744,
+      "step": 11740
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2866658780718096e-05,
+      "loss": 0.6162,
+      "step": 11741
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2865534926200974e-05,
+      "loss": 0.5887,
+      "step": 11742
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.286441103225157e-05,
+      "loss": 0.6709,
+      "step": 11743
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2863287098885358e-05,
+      "loss": 0.6463,
+      "step": 11744
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2862163126117794e-05,
+      "loss": 0.5502,
+      "step": 11745
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2861039113964354e-05,
+      "loss": 0.6621,
+      "step": 11746
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2859915062440502e-05,
+      "loss": 0.6084,
+      "step": 11747
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2858790971561704e-05,
+      "loss": 0.5914,
+      "step": 11748
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2857666841343433e-05,
+      "loss": 0.6474,
+      "step": 11749
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.285654267180115e-05,
+      "loss": 0.6624,
+      "step": 11750
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2855418462950335e-05,
+      "loss": 0.5809,
+      "step": 11751
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2854294214806453e-05,
+      "loss": 0.6835,
+      "step": 11752
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2853169927384972e-05,
+      "loss": 0.5843,
+      "step": 11753
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2852045600701366e-05,
+      "loss": 0.6451,
+      "step": 11754
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.285092123477111e-05,
+      "loss": 0.6925,
+      "step": 11755
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2849796829609669e-05,
+      "loss": 0.6002,
+      "step": 11756
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2848672385232519e-05,
+      "loss": 0.6274,
+      "step": 11757
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2847547901655134e-05,
+      "loss": 0.6002,
+      "step": 11758
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2846423378892992e-05,
+      "loss": 0.635,
+      "step": 11759
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284529881696156e-05,
+      "loss": 0.6993,
+      "step": 11760
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2844174215876315e-05,
+      "loss": 0.6635,
+      "step": 11761
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2843049575652737e-05,
+      "loss": 0.6663,
+      "step": 11762
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2841924896306298e-05,
+      "loss": 0.6275,
+      "step": 11763
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2840800177852474e-05,
+      "loss": 0.6607,
+      "step": 11764
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2839675420306742e-05,
+      "loss": 0.6728,
+      "step": 11765
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2838550623684585e-05,
+      "loss": 0.6128,
+      "step": 11766
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2837425788001475e-05,
+      "loss": 0.562,
+      "step": 11767
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2836300913272892e-05,
+      "loss": 0.6336,
+      "step": 11768
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2835175999514318e-05,
+      "loss": 0.6693,
+      "step": 11769
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2834051046741229e-05,
+      "loss": 0.5617,
+      "step": 11770
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2832926054969111e-05,
+      "loss": 0.6151,
+      "step": 11771
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2831801024213437e-05,
+      "loss": 0.5974,
+      "step": 11772
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2830675954489697e-05,
+      "loss": 0.6653,
+      "step": 11773
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2829550845813365e-05,
+      "loss": 0.623,
+      "step": 11774
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2828425698199928e-05,
+      "loss": 0.5561,
+      "step": 11775
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2827300511664868e-05,
+      "loss": 0.6531,
+      "step": 11776
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.282617528622367e-05,
+      "loss": 0.5976,
+      "step": 11777
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2825050021891816e-05,
+      "loss": 0.6528,
+      "step": 11778
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.282392471868479e-05,
+      "loss": 0.6922,
+      "step": 11779
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2822799376618083e-05,
+      "loss": 0.6608,
+      "step": 11780
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2821673995707171e-05,
+      "loss": 0.7092,
+      "step": 11781
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2820548575967549e-05,
+      "loss": 0.6621,
+      "step": 11782
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2819423117414698e-05,
+      "loss": 0.604,
+      "step": 11783
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2818297620064109e-05,
+      "loss": 0.6113,
+      "step": 11784
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2817172083931266e-05,
+      "loss": 0.5884,
+      "step": 11785
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2816046509031663e-05,
+      "loss": 0.6028,
+      "step": 11786
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2814920895380785e-05,
+      "loss": 0.6681,
+      "step": 11787
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2813795242994119e-05,
+      "loss": 0.603,
+      "step": 11788
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2812669551887163e-05,
+      "loss": 0.624,
+      "step": 11789
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2811543822075396e-05,
+      "loss": 0.6492,
+      "step": 11790
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.281041805357432e-05,
+      "loss": 0.6788,
+      "step": 11791
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2809292246399422e-05,
+      "loss": 0.523,
+      "step": 11792
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2808166400566195e-05,
+      "loss": 0.6004,
+      "step": 11793
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2807040516090128e-05,
+      "loss": 0.6546,
+      "step": 11794
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.280591459298672e-05,
+      "loss": 0.6518,
+      "step": 11795
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2804788631271459e-05,
+      "loss": 0.7016,
+      "step": 11796
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2803662630959845e-05,
+      "loss": 0.5732,
+      "step": 11797
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2802536592067367e-05,
+      "loss": 0.5374,
+      "step": 11798
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2801410514609522e-05,
+      "loss": 0.6516,
+      "step": 11799
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.280028439860181e-05,
+      "loss": 0.6795,
+      "step": 11800
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2799158244059724e-05,
+      "loss": 0.6623,
+      "step": 11801
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279803205099876e-05,
+      "loss": 0.5966,
+      "step": 11802
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2796905819434415e-05,
+      "loss": 0.617,
+      "step": 11803
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279577954938219e-05,
+      "loss": 0.6448,
+      "step": 11804
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2794653240857582e-05,
+      "loss": 0.6582,
+      "step": 11805
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279352689387609e-05,
+      "loss": 0.528,
+      "step": 11806
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279240050845321e-05,
+      "loss": 0.5777,
+      "step": 11807
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2791274084604449e-05,
+      "loss": 0.7054,
+      "step": 11808
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2790147622345303e-05,
+      "loss": 0.738,
+      "step": 11809
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2789021121691273e-05,
+      "loss": 0.6389,
+      "step": 11810
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2787894582657864e-05,
+      "loss": 0.6404,
+      "step": 11811
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2786768005260574e-05,
+      "loss": 0.6617,
+      "step": 11812
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.278564138951491e-05,
+      "loss": 0.6005,
+      "step": 11813
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.278451473543637e-05,
+      "loss": 0.6126,
+      "step": 11814
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.278338804304046e-05,
+      "loss": 0.6789,
+      "step": 11815
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2782261312342687e-05,
+      "loss": 0.628,
+      "step": 11816
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2781134543358556e-05,
+      "loss": 0.6519,
+      "step": 11817
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2780007736103563e-05,
+      "loss": 0.6056,
+      "step": 11818
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2778880890593227e-05,
+      "loss": 0.6333,
+      "step": 11819
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2777754006843046e-05,
+      "loss": 0.5761,
+      "step": 11820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2776627084868527e-05,
+      "loss": 0.633,
+      "step": 11821
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2775500124685183e-05,
+      "loss": 0.5661,
+      "step": 11822
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2774373126308517e-05,
+      "loss": 0.5951,
+      "step": 11823
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277324608975404e-05,
+      "loss": 0.652,
+      "step": 11824
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2772119015037259e-05,
+      "loss": 0.7272,
+      "step": 11825
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2770991902173683e-05,
+      "loss": 0.6101,
+      "step": 11826
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2769864751178827e-05,
+      "loss": 0.571,
+      "step": 11827
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2768737562068197e-05,
+      "loss": 0.5815,
+      "step": 11828
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2767610334857303e-05,
+      "loss": 0.6423,
+      "step": 11829
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2766483069561662e-05,
+      "loss": 0.5934,
+      "step": 11830
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2765355766196781e-05,
+      "loss": 0.6243,
+      "step": 11831
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2764228424778175e-05,
+      "loss": 0.6009,
+      "step": 11832
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.276310104532136e-05,
+      "loss": 0.609,
+      "step": 11833
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2761973627841844e-05,
+      "loss": 0.6734,
+      "step": 11834
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2760846172355144e-05,
+      "loss": 0.6586,
+      "step": 11835
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2759718678876775e-05,
+      "loss": 0.6179,
+      "step": 11836
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2758591147422253e-05,
+      "loss": 0.5921,
+      "step": 11837
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2757463578007092e-05,
+      "loss": 0.5893,
+      "step": 11838
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2756335970646809e-05,
+      "loss": 0.6518,
+      "step": 11839
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2755208325356922e-05,
+      "loss": 0.666,
+      "step": 11840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2754080642152946e-05,
+      "loss": 0.6821,
+      "step": 11841
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2752952921050403e-05,
+      "loss": 0.6453,
+      "step": 11842
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2751825162064807e-05,
+      "loss": 0.577,
+      "step": 11843
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2750697365211679e-05,
+      "loss": 0.6493,
+      "step": 11844
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2749569530506538e-05,
+      "loss": 0.7112,
+      "step": 11845
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2748441657964907e-05,
+      "loss": 0.6896,
+      "step": 11846
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2747313747602299e-05,
+      "loss": 0.7099,
+      "step": 11847
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2746185799434243e-05,
+      "loss": 0.5922,
+      "step": 11848
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2745057813476258e-05,
+      "loss": 0.6205,
+      "step": 11849
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2743929789743863e-05,
+      "loss": 0.6411,
+      "step": 11850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2742801728252587e-05,
+      "loss": 0.6264,
+      "step": 11851
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2741673629017946e-05,
+      "loss": 0.6442,
+      "step": 11852
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2740545492055467e-05,
+      "loss": 0.6422,
+      "step": 11853
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2739417317380674e-05,
+      "loss": 0.6915,
+      "step": 11854
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2738289105009091e-05,
+      "loss": 0.6966,
+      "step": 11855
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2737160854956245e-05,
+      "loss": 0.6363,
+      "step": 11856
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273603256723766e-05,
+      "loss": 0.6267,
+      "step": 11857
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2734904241868864e-05,
+      "loss": 0.6391,
+      "step": 11858
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2733775878865382e-05,
+      "loss": 0.642,
+      "step": 11859
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2732647478242741e-05,
+      "loss": 0.6919,
+      "step": 11860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273151904001647e-05,
+      "loss": 0.5811,
+      "step": 11861
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2730390564202097e-05,
+      "loss": 0.6496,
+      "step": 11862
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272926205081515e-05,
+      "loss": 0.7376,
+      "step": 11863
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272813349987116e-05,
+      "loss": 0.6064,
+      "step": 11864
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2727004911385657e-05,
+      "loss": 0.6583,
+      "step": 11865
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2725876285374166e-05,
+      "loss": 0.5891,
+      "step": 11866
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2724747621852227e-05,
+      "loss": 0.5718,
+      "step": 11867
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2723618920835364e-05,
+      "loss": 0.6407,
+      "step": 11868
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2722490182339114e-05,
+      "loss": 0.5691,
+      "step": 11869
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2721361406379003e-05,
+      "loss": 0.6686,
+      "step": 11870
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2720232592970571e-05,
+      "loss": 0.6776,
+      "step": 11871
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2719103742129346e-05,
+      "loss": 0.6762,
+      "step": 11872
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2717974853870866e-05,
+      "loss": 0.717,
+      "step": 11873
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2716845928210662e-05,
+      "loss": 0.6708,
+      "step": 11874
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2715716965164273e-05,
+      "loss": 0.6556,
+      "step": 11875
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2714587964747231e-05,
+      "loss": 0.6232,
+      "step": 11876
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2713458926975073e-05,
+      "loss": 0.6527,
+      "step": 11877
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2712329851863337e-05,
+      "loss": 0.6429,
+      "step": 11878
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2711200739427558e-05,
+      "loss": 0.5521,
+      "step": 11879
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2710071589683275e-05,
+      "loss": 0.5704,
+      "step": 11880
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2708942402646025e-05,
+      "loss": 0.601,
+      "step": 11881
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270781317833135e-05,
+      "loss": 0.5924,
+      "step": 11882
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2706683916754782e-05,
+      "loss": 0.6521,
+      "step": 11883
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270555461793187e-05,
+      "loss": 0.6061,
+      "step": 11884
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2704425281878145e-05,
+      "loss": 0.5421,
+      "step": 11885
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2703295908609153e-05,
+      "loss": 0.678,
+      "step": 11886
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2702166498140436e-05,
+      "loss": 0.645,
+      "step": 11887
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270103705048753e-05,
+      "loss": 0.6869,
+      "step": 11888
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2699907565665983e-05,
+      "loss": 0.6219,
+      "step": 11889
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2698778043691336e-05,
+      "loss": 0.6335,
+      "step": 11890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2697648484579133e-05,
+      "loss": 0.6029,
+      "step": 11891
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2696518888344911e-05,
+      "loss": 0.5553,
+      "step": 11892
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2695389255004224e-05,
+      "loss": 0.667,
+      "step": 11893
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2694259584572612e-05,
+      "loss": 0.6264,
+      "step": 11894
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2693129877065622e-05,
+      "loss": 0.6337,
+      "step": 11895
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2692000132498796e-05,
+      "loss": 0.64,
+      "step": 11896
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2690870350887683e-05,
+      "loss": 0.6299,
+      "step": 11897
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2689740532247832e-05,
+      "loss": 0.6871,
+      "step": 11898
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2688610676594783e-05,
+      "loss": 0.6803,
+      "step": 11899
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2687480783944096e-05,
+      "loss": 0.621,
+      "step": 11900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2686350854311304e-05,
+      "loss": 0.6432,
+      "step": 11901
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.268522088771197e-05,
+      "loss": 0.6674,
+      "step": 11902
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2684090884161636e-05,
+      "loss": 0.6588,
+      "step": 11903
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2682960843675852e-05,
+      "loss": 0.635,
+      "step": 11904
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2681830766270168e-05,
+      "loss": 0.6949,
+      "step": 11905
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2680700651960138e-05,
+      "loss": 0.6526,
+      "step": 11906
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2679570500761309e-05,
+      "loss": 0.6111,
+      "step": 11907
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2678440312689238e-05,
+      "loss": 0.6139,
+      "step": 11908
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2677310087759476e-05,
+      "loss": 0.5637,
+      "step": 11909
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2676179825987572e-05,
+      "loss": 0.5727,
+      "step": 11910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2675049527389083e-05,
+      "loss": 0.6413,
+      "step": 11911
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.267391919197956e-05,
+      "loss": 0.656,
+      "step": 11912
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2672788819774562e-05,
+      "loss": 0.6942,
+      "step": 11913
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2671658410789642e-05,
+      "loss": 0.6459,
+      "step": 11914
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2670527965040353e-05,
+      "loss": 0.5476,
+      "step": 11915
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2669397482542252e-05,
+      "loss": 0.6195,
+      "step": 11916
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2668266963310898e-05,
+      "loss": 0.5667,
+      "step": 11917
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2667136407361844e-05,
+      "loss": 0.6945,
+      "step": 11918
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2666005814710652e-05,
+      "loss": 0.6541,
+      "step": 11919
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2664875185372875e-05,
+      "loss": 0.5815,
+      "step": 11920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2663744519364073e-05,
+      "loss": 0.6025,
+      "step": 11921
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2662613816699812e-05,
+      "loss": 0.5986,
+      "step": 11922
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2661483077395639e-05,
+      "loss": 0.6184,
+      "step": 11923
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2660352301467126e-05,
+      "loss": 0.684,
+      "step": 11924
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2659221488929823e-05,
+      "loss": 0.602,
+      "step": 11925
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2658090639799297e-05,
+      "loss": 0.6398,
+      "step": 11926
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2656959754091107e-05,
+      "loss": 0.7219,
+      "step": 11927
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2655828831820815e-05,
+      "loss": 0.6127,
+      "step": 11928
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.265469787300399e-05,
+      "loss": 0.5995,
+      "step": 11929
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2653566877656184e-05,
+      "loss": 0.7231,
+      "step": 11930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2652435845792969e-05,
+      "loss": 0.5535,
+      "step": 11931
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2651304777429907e-05,
+      "loss": 0.606,
+      "step": 11932
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2650173672582562e-05,
+      "loss": 0.6025,
+      "step": 11933
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2649042531266495e-05,
+      "loss": 0.6268,
+      "step": 11934
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2647911353497277e-05,
+      "loss": 0.6717,
+      "step": 11935
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2646780139290473e-05,
+      "loss": 0.588,
+      "step": 11936
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2645648888661648e-05,
+      "loss": 0.6133,
+      "step": 11937
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.264451760162637e-05,
+      "loss": 0.5892,
+      "step": 11938
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2643386278200204e-05,
+      "loss": 0.5896,
+      "step": 11939
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2642254918398722e-05,
+      "loss": 0.6299,
+      "step": 11940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.264112352223749e-05,
+      "loss": 0.6694,
+      "step": 11941
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2639992089732081e-05,
+      "loss": 0.6044,
+      "step": 11942
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2638860620898057e-05,
+      "loss": 0.6357,
+      "step": 11943
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2637729115750995e-05,
+      "loss": 0.5936,
+      "step": 11944
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.263659757430646e-05,
+      "loss": 0.6736,
+      "step": 11945
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2635465996580028e-05,
+      "loss": 0.6659,
+      "step": 11946
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2634334382587264e-05,
+      "loss": 0.6669,
+      "step": 11947
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2633202732343746e-05,
+      "loss": 0.6128,
+      "step": 11948
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2632071045865048e-05,
+      "loss": 0.6625,
+      "step": 11949
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2630939323166735e-05,
+      "loss": 0.6231,
+      "step": 11950
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2629807564264389e-05,
+      "loss": 0.6413,
+      "step": 11951
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262867576917358e-05,
+      "loss": 0.6073,
+      "step": 11952
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2627543937909882e-05,
+      "loss": 0.6556,
+      "step": 11953
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262641207048887e-05,
+      "loss": 0.6567,
+      "step": 11954
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2625280166926119e-05,
+      "loss": 0.5816,
+      "step": 11955
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262414822723721e-05,
+      "loss": 0.6344,
+      "step": 11956
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262301625143771e-05,
+      "loss": 0.5839,
+      "step": 11957
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2621884239543208e-05,
+      "loss": 0.5549,
+      "step": 11958
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2620752191569267e-05,
+      "loss": 0.6081,
+      "step": 11959
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.261962010753148e-05,
+      "loss": 0.6271,
+      "step": 11960
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2618487987445417e-05,
+      "loss": 0.6876,
+      "step": 11961
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2617355831326657e-05,
+      "loss": 0.5435,
+      "step": 11962
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2616223639190782e-05,
+      "loss": 0.5842,
+      "step": 11963
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2615091411053369e-05,
+      "loss": 0.7174,
+      "step": 11964
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2613959146930004e-05,
+      "loss": 0.6589,
+      "step": 11965
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.261282684683626e-05,
+      "loss": 0.6368,
+      "step": 11966
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2611694510787727e-05,
+      "loss": 0.617,
+      "step": 11967
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2610562138799977e-05,
+      "loss": 0.5843,
+      "step": 11968
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2609429730888605e-05,
+      "loss": 0.5979,
+      "step": 11969
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2608297287069182e-05,
+      "loss": 0.615,
+      "step": 11970
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2607164807357299e-05,
+      "loss": 0.6412,
+      "step": 11971
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2606032291768537e-05,
+      "loss": 0.5868,
+      "step": 11972
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.260489974031848e-05,
+      "loss": 0.656,
+      "step": 11973
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2603767153022715e-05,
+      "loss": 0.6163,
+      "step": 11974
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2602634529896824e-05,
+      "loss": 0.622,
+      "step": 11975
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2601501870956396e-05,
+      "loss": 0.614,
+      "step": 11976
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2600369176217018e-05,
+      "loss": 0.6369,
+      "step": 11977
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2599236445694274e-05,
+      "loss": 0.5623,
+      "step": 11978
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.259810367940375e-05,
+      "loss": 0.6048,
+      "step": 11979
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2596970877361041e-05,
+      "loss": 0.634,
+      "step": 11980
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2595838039581729e-05,
+      "loss": 0.6536,
+      "step": 11981
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2594705166081404e-05,
+      "loss": 0.5468,
+      "step": 11982
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2593572256875656e-05,
+      "loss": 0.5625,
+      "step": 11983
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2592439311980078e-05,
+      "loss": 0.6089,
+      "step": 11984
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2591306331410254e-05,
+      "loss": 0.5891,
+      "step": 11985
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2590173315181776e-05,
+      "loss": 0.6584,
+      "step": 11986
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2589040263310241e-05,
+      "loss": 0.6246,
+      "step": 11987
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2587907175811234e-05,
+      "loss": 0.7027,
+      "step": 11988
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2586774052700354e-05,
+      "loss": 0.6168,
+      "step": 11989
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2585640893993187e-05,
+      "loss": 0.6485,
+      "step": 11990
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2584507699705332e-05,
+      "loss": 0.6508,
+      "step": 11991
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2583374469852379e-05,
+      "loss": 0.6749,
+      "step": 11992
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2582241204449923e-05,
+      "loss": 0.6805,
+      "step": 11993
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2581107903513558e-05,
+      "loss": 0.6688,
+      "step": 11994
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2579974567058879e-05,
+      "loss": 0.5208,
+      "step": 11995
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2578841195101487e-05,
+      "loss": 0.5888,
+      "step": 11996
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2577707787656972e-05,
+      "loss": 0.6747,
+      "step": 11997
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2576574344740935e-05,
+      "loss": 0.6176,
+      "step": 11998
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2575440866368969e-05,
+      "loss": 0.6446,
+      "step": 11999
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2574307352556675e-05,
+      "loss": 0.5789,
+      "step": 12000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2573173803319647e-05,
+      "loss": 0.5693,
+      "step": 12001
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2572040218673492e-05,
+      "loss": 0.6982,
+      "step": 12002
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2570906598633798e-05,
+      "loss": 0.6615,
+      "step": 12003
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2569772943216174e-05,
+      "loss": 0.6725,
+      "step": 12004
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2568639252436212e-05,
+      "loss": 0.5949,
+      "step": 12005
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.256750552630952e-05,
+      "loss": 0.5844,
+      "step": 12006
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2566371764851694e-05,
+      "loss": 0.5845,
+      "step": 12007
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2565237968078339e-05,
+      "loss": 0.6469,
+      "step": 12008
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2564104136005058e-05,
+      "loss": 0.6263,
+      "step": 12009
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2562970268647447e-05,
+      "loss": 0.6765,
+      "step": 12010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2561836366021116e-05,
+      "loss": 0.5414,
+      "step": 12011
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2560702428141662e-05,
+      "loss": 0.5999,
+      "step": 12012
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2559568455024698e-05,
+      "loss": 0.7071,
+      "step": 12013
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2558434446685816e-05,
+      "loss": 0.651,
+      "step": 12014
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2557300403140636e-05,
+      "loss": 0.5758,
+      "step": 12015
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2556166324404747e-05,
+      "loss": 0.6549,
+      "step": 12016
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2555032210493768e-05,
+      "loss": 0.557,
+      "step": 12017
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2553898061423302e-05,
+      "loss": 0.6555,
+      "step": 12018
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2552763877208952e-05,
+      "loss": 0.6041,
+      "step": 12019
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2551629657866329e-05,
+      "loss": 0.5901,
+      "step": 12020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2550495403411038e-05,
+      "loss": 0.5535,
+      "step": 12021
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2549361113858693e-05,
+      "loss": 0.6569,
+      "step": 12022
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2548226789224896e-05,
+      "loss": 0.6678,
+      "step": 12023
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2547092429525261e-05,
+      "loss": 0.6206,
+      "step": 12024
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2545958034775393e-05,
+      "loss": 0.5857,
+      "step": 12025
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2544823604990908e-05,
+      "loss": 0.6622,
+      "step": 12026
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2543689140187417e-05,
+      "loss": 0.6277,
+      "step": 12027
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2542554640380522e-05,
+      "loss": 0.5682,
+      "step": 12028
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2541420105585846e-05,
+      "loss": 0.659,
+      "step": 12029
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2540285535818992e-05,
+      "loss": 0.6602,
+      "step": 12030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253915093109558e-05,
+      "loss": 0.5735,
+      "step": 12031
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253801629143122e-05,
+      "loss": 0.6277,
+      "step": 12032
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2536881616841527e-05,
+      "loss": 0.7422,
+      "step": 12033
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253574690734211e-05,
+      "loss": 0.6109,
+      "step": 12034
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2534612162948591e-05,
+      "loss": 0.6669,
+      "step": 12035
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253347738367658e-05,
+      "loss": 0.5589,
+      "step": 12036
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2532342569541691e-05,
+      "loss": 0.6419,
+      "step": 12037
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2531207720559548e-05,
+      "loss": 0.596,
+      "step": 12038
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2530072836745762e-05,
+      "loss": 0.566,
+      "step": 12039
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2528937918115948e-05,
+      "loss": 0.6177,
+      "step": 12040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252780296468573e-05,
+      "loss": 0.6054,
+      "step": 12041
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252666797647072e-05,
+      "loss": 0.6789,
+      "step": 12042
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2525532953486537e-05,
+      "loss": 0.5535,
+      "step": 12043
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2524397895748804e-05,
+      "loss": 0.6586,
+      "step": 12044
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2523262803273135e-05,
+      "loss": 0.6203,
+      "step": 12045
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2522127676075155e-05,
+      "loss": 0.6637,
+      "step": 12046
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252099251417048e-05,
+      "loss": 0.5936,
+      "step": 12047
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2519857317574733e-05,
+      "loss": 0.6447,
+      "step": 12048
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.251872208630354e-05,
+      "loss": 0.6656,
+      "step": 12049
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2517586820372511e-05,
+      "loss": 0.6267,
+      "step": 12050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.251645151979728e-05,
+      "loss": 0.565,
+      "step": 12051
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2515316184593462e-05,
+      "loss": 0.6662,
+      "step": 12052
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2514180814776686e-05,
+      "loss": 0.6285,
+      "step": 12053
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.251304541036257e-05,
+      "loss": 0.6137,
+      "step": 12054
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2511909971366744e-05,
+      "loss": 0.5939,
+      "step": 12055
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2510774497804827e-05,
+      "loss": 0.6117,
+      "step": 12056
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2509638989692447e-05,
+      "loss": 0.6201,
+      "step": 12057
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.250850344704523e-05,
+      "loss": 0.6895,
+      "step": 12058
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2507367869878804e-05,
+      "loss": 0.6155,
+      "step": 12059
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2506232258208793e-05,
+      "loss": 0.5792,
+      "step": 12060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.250509661205082e-05,
+      "loss": 0.6125,
+      "step": 12061
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2503960931420519e-05,
+      "loss": 0.6534,
+      "step": 12062
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2502825216333512e-05,
+      "loss": 0.6558,
+      "step": 12063
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2501689466805433e-05,
+      "loss": 0.6718,
+      "step": 12064
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.250055368285191e-05,
+      "loss": 0.6388,
+      "step": 12065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.249941786448857e-05,
+      "loss": 0.6642,
+      "step": 12066
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2498282011731045e-05,
+      "loss": 0.6375,
+      "step": 12067
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.249714612459496e-05,
+      "loss": 0.629,
+      "step": 12068
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2496010203095955e-05,
+      "loss": 0.5828,
+      "step": 12069
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2494874247249656e-05,
+      "loss": 0.6337,
+      "step": 12070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2493738257071695e-05,
+      "loss": 0.6188,
+      "step": 12071
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2492602232577702e-05,
+      "loss": 0.6483,
+      "step": 12072
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2491466173783313e-05,
+      "loss": 0.627,
+      "step": 12073
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2490330080704163e-05,
+      "loss": 0.7417,
+      "step": 12074
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2489193953355879e-05,
+      "loss": 0.6506,
+      "step": 12075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2488057791754102e-05,
+      "loss": 0.6443,
+      "step": 12076
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2486921595914461e-05,
+      "loss": 0.6312,
+      "step": 12077
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2485785365852596e-05,
+      "loss": 0.6155,
+      "step": 12078
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.248464910158414e-05,
+      "loss": 0.6175,
+      "step": 12079
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2483512803124729e-05,
+      "loss": 0.5961,
+      "step": 12080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2482376470489996e-05,
+      "loss": 0.6058,
+      "step": 12081
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2481240103695587e-05,
+      "loss": 0.6152,
+      "step": 12082
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.248010370275713e-05,
+      "loss": 0.6892,
+      "step": 12083
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2478967267690267e-05,
+      "loss": 0.6278,
+      "step": 12084
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247783079851064e-05,
+      "loss": 0.6097,
+      "step": 12085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247669429523388e-05,
+      "loss": 0.5514,
+      "step": 12086
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2475557757875632e-05,
+      "loss": 0.6444,
+      "step": 12087
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2474421186451533e-05,
+      "loss": 0.6507,
+      "step": 12088
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2473284580977224e-05,
+      "loss": 0.5859,
+      "step": 12089
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247214794146835e-05,
+      "loss": 0.5919,
+      "step": 12090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2471011267940545e-05,
+      "loss": 0.687,
+      "step": 12091
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2469874560409452e-05,
+      "loss": 0.5552,
+      "step": 12092
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2468737818890717e-05,
+      "loss": 0.6434,
+      "step": 12093
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.246760104339998e-05,
+      "loss": 0.6576,
+      "step": 12094
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2466464233952883e-05,
+      "loss": 0.6449,
+      "step": 12095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2465327390565074e-05,
+      "loss": 0.6724,
+      "step": 12096
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2464190513252192e-05,
+      "loss": 0.5901,
+      "step": 12097
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2463053602029884e-05,
+      "loss": 0.6414,
+      "step": 12098
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2461916656913792e-05,
+      "loss": 0.5654,
+      "step": 12099
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2460779677919567e-05,
+      "loss": 0.6516,
+      "step": 12100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245964266506285e-05,
+      "loss": 0.6177,
+      "step": 12101
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245850561835929e-05,
+      "loss": 0.6691,
+      "step": 12102
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2457368537824528e-05,
+      "loss": 0.7164,
+      "step": 12103
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245623142347422e-05,
+      "loss": 0.5895,
+      "step": 12104
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245509427532401e-05,
+      "loss": 0.5809,
+      "step": 12105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2453957093389542e-05,
+      "loss": 0.5655,
+      "step": 12106
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2452819877686469e-05,
+      "loss": 0.6565,
+      "step": 12107
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245168262823044e-05,
+      "loss": 0.6574,
+      "step": 12108
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2450545345037104e-05,
+      "loss": 0.5759,
+      "step": 12109
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244940802812211e-05,
+      "loss": 0.671,
+      "step": 12110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244827067750111e-05,
+      "loss": 0.6272,
+      "step": 12111
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2447133293189753e-05,
+      "loss": 0.5943,
+      "step": 12112
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2445995875203693e-05,
+      "loss": 0.6241,
+      "step": 12113
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2444858423558581e-05,
+      "loss": 0.5769,
+      "step": 12114
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2443720938270067e-05,
+      "loss": 0.6058,
+      "step": 12115
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2442583419353808e-05,
+      "loss": 0.5381,
+      "step": 12116
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2441445866825452e-05,
+      "loss": 0.6395,
+      "step": 12117
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244030828070066e-05,
+      "loss": 0.6087,
+      "step": 12118
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2439170660995077e-05,
+      "loss": 0.6287,
+      "step": 12119
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2438033007724367e-05,
+      "loss": 0.6029,
+      "step": 12120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2436895320904176e-05,
+      "loss": 0.592,
+      "step": 12121
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2435757600550167e-05,
+      "loss": 0.732,
+      "step": 12122
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2434619846677991e-05,
+      "loss": 0.6802,
+      "step": 12123
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.243348205930331e-05,
+      "loss": 0.5889,
+      "step": 12124
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2432344238441774e-05,
+      "loss": 0.6364,
+      "step": 12125
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2431206384109045e-05,
+      "loss": 0.6296,
+      "step": 12126
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2430068496320782e-05,
+      "loss": 0.6349,
+      "step": 12127
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.242893057509264e-05,
+      "loss": 0.607,
+      "step": 12128
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.242779262044028e-05,
+      "loss": 0.6625,
+      "step": 12129
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2426654632379359e-05,
+      "loss": 0.5801,
+      "step": 12130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2425516610925539e-05,
+      "loss": 0.7105,
+      "step": 12131
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2424378556094476e-05,
+      "loss": 0.6026,
+      "step": 12132
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2423240467901838e-05,
+      "loss": 0.7141,
+      "step": 12133
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2422102346363278e-05,
+      "loss": 0.6001,
+      "step": 12134
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2420964191494464e-05,
+      "loss": 0.6622,
+      "step": 12135
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2419826003311056e-05,
+      "loss": 0.5806,
+      "step": 12136
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2418687781828713e-05,
+      "loss": 0.6169,
+      "step": 12137
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2417549527063106e-05,
+      "loss": 0.6762,
+      "step": 12138
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241641123902989e-05,
+      "loss": 0.6285,
+      "step": 12139
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2415272917744731e-05,
+      "loss": 0.6652,
+      "step": 12140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2414134563223294e-05,
+      "loss": 0.602,
+      "step": 12141
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241299617548125e-05,
+      "loss": 0.6236,
+      "step": 12142
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2411857754534254e-05,
+      "loss": 0.578,
+      "step": 12143
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2410719300397975e-05,
+      "loss": 0.6278,
+      "step": 12144
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2409580813088082e-05,
+      "loss": 0.5527,
+      "step": 12145
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2408442292620237e-05,
+      "loss": 0.6025,
+      "step": 12146
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2407303739010113e-05,
+      "loss": 0.6171,
+      "step": 12147
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2406165152273374e-05,
+      "loss": 0.5585,
+      "step": 12148
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2405026532425688e-05,
+      "loss": 0.6472,
+      "step": 12149
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2403887879482723e-05,
+      "loss": 0.6568,
+      "step": 12150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2402749193460148e-05,
+      "loss": 0.7668,
+      "step": 12151
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2401610474373635e-05,
+      "loss": 0.617,
+      "step": 12152
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2400471722238851e-05,
+      "loss": 0.5283,
+      "step": 12153
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2399332937071465e-05,
+      "loss": 0.6327,
+      "step": 12154
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2398194118887152e-05,
+      "loss": 0.6559,
+      "step": 12155
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2397055267701579e-05,
+      "loss": 0.6028,
+      "step": 12156
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2395916383530418e-05,
+      "loss": 0.6584,
+      "step": 12157
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2394777466389345e-05,
+      "loss": 0.6614,
+      "step": 12158
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2393638516294029e-05,
+      "loss": 0.6411,
+      "step": 12159
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2392499533260145e-05,
+      "loss": 0.6444,
+      "step": 12160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2391360517303362e-05,
+      "loss": 0.6014,
+      "step": 12161
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.239022146843936e-05,
+      "loss": 0.5411,
+      "step": 12162
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2389082386683809e-05,
+      "loss": 0.6472,
+      "step": 12163
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2387943272052385e-05,
+      "loss": 0.6763,
+      "step": 12164
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2386804124560764e-05,
+      "loss": 0.6091,
+      "step": 12165
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2385664944224622e-05,
+      "loss": 0.6205,
+      "step": 12166
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2384525731059634e-05,
+      "loss": 0.6933,
+      "step": 12167
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2383386485081475e-05,
+      "loss": 0.6497,
+      "step": 12168
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2382247206305825e-05,
+      "loss": 0.5228,
+      "step": 12169
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.238110789474836e-05,
+      "loss": 0.5731,
+      "step": 12170
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2379968550424756e-05,
+      "loss": 0.5718,
+      "step": 12171
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2378829173350695e-05,
+      "loss": 0.6227,
+      "step": 12172
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2377689763541853e-05,
+      "loss": 0.62,
+      "step": 12173
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2376550321013914e-05,
+      "loss": 0.6003,
+      "step": 12174
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2375410845782551e-05,
+      "loss": 0.6094,
+      "step": 12175
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.237427133786345e-05,
+      "loss": 0.6088,
+      "step": 12176
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2373131797272289e-05,
+      "loss": 0.5669,
+      "step": 12177
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2371992224024749e-05,
+      "loss": 0.6371,
+      "step": 12178
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.237085261813651e-05,
+      "loss": 0.6135,
+      "step": 12179
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2369712979623259e-05,
+      "loss": 0.6726,
+      "step": 12180
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2368573308500673e-05,
+      "loss": 0.6274,
+      "step": 12181
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2367433604784438e-05,
+      "loss": 0.5808,
+      "step": 12182
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2366293868490236e-05,
+      "loss": 0.6044,
+      "step": 12183
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2365154099633749e-05,
+      "loss": 0.6319,
+      "step": 12184
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2364014298230667e-05,
+      "loss": 0.6567,
+      "step": 12185
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2362874464296669e-05,
+      "loss": 0.6252,
+      "step": 12186
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2361734597847442e-05,
+      "loss": 0.5622,
+      "step": 12187
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236059469889867e-05,
+      "loss": 0.6521,
+      "step": 12188
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2359454767466045e-05,
+      "loss": 0.6317,
+      "step": 12189
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2358314803565246e-05,
+      "loss": 0.5734,
+      "step": 12190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2357174807211966e-05,
+      "loss": 0.6478,
+      "step": 12191
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2356034778421887e-05,
+      "loss": 0.5899,
+      "step": 12192
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2354894717210699e-05,
+      "loss": 0.6537,
+      "step": 12193
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2353754623594092e-05,
+      "loss": 0.6527,
+      "step": 12194
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235261449758775e-05,
+      "loss": 0.6468,
+      "step": 12195
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2351474339207372e-05,
+      "loss": 0.6458,
+      "step": 12196
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2350334148468635e-05,
+      "loss": 0.5452,
+      "step": 12197
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.234919392538724e-05,
+      "loss": 0.6142,
+      "step": 12198
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2348053669978866e-05,
+      "loss": 0.6045,
+      "step": 12199
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2346913382259216e-05,
+      "loss": 0.578,
+      "step": 12200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2345773062243971e-05,
+      "loss": 0.6853,
+      "step": 12201
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2344632709948831e-05,
+      "loss": 0.6383,
+      "step": 12202
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2343492325389483e-05,
+      "loss": 0.6149,
+      "step": 12203
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2342351908581622e-05,
+      "loss": 0.6565,
+      "step": 12204
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.234121145954094e-05,
+      "loss": 0.6169,
+      "step": 12205
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2340070978283132e-05,
+      "loss": 0.5844,
+      "step": 12206
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.233893046482389e-05,
+      "loss": 0.6254,
+      "step": 12207
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2337789919178911e-05,
+      "loss": 0.6339,
+      "step": 12208
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2336649341363889e-05,
+      "loss": 0.6458,
+      "step": 12209
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2335508731394516e-05,
+      "loss": 0.6345,
+      "step": 12210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2334368089286494e-05,
+      "loss": 0.6655,
+      "step": 12211
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2333227415055515e-05,
+      "loss": 0.594,
+      "step": 12212
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2332086708717277e-05,
+      "loss": 0.6124,
+      "step": 12213
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2330945970287473e-05,
+      "loss": 0.6517,
+      "step": 12214
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2329805199781808e-05,
+      "loss": 0.5785,
+      "step": 12215
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2328664397215976e-05,
+      "loss": 0.6103,
+      "step": 12216
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2327523562605675e-05,
+      "loss": 0.5993,
+      "step": 12217
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2326382695966605e-05,
+      "loss": 0.578,
+      "step": 12218
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2325241797314465e-05,
+      "loss": 0.6224,
+      "step": 12219
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2324100866664955e-05,
+      "loss": 0.6237,
+      "step": 12220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2322959904033774e-05,
+      "loss": 0.6469,
+      "step": 12221
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2321818909436625e-05,
+      "loss": 0.5893,
+      "step": 12222
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2320677882889205e-05,
+      "loss": 0.606,
+      "step": 12223
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.231953682440722e-05,
+      "loss": 0.5593,
+      "step": 12224
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2318395734006372e-05,
+      "loss": 0.6595,
+      "step": 12225
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2317254611702358e-05,
+      "loss": 0.647,
+      "step": 12226
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2316113457510887e-05,
+      "loss": 0.7199,
+      "step": 12227
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2314972271447657e-05,
+      "loss": 0.6307,
+      "step": 12228
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2313831053528378e-05,
+      "loss": 0.6752,
+      "step": 12229
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2312689803768749e-05,
+      "loss": 0.606,
+      "step": 12230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2311548522184476e-05,
+      "loss": 0.6191,
+      "step": 12231
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2310407208791264e-05,
+      "loss": 0.6709,
+      "step": 12232
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.230926586360482e-05,
+      "loss": 0.7471,
+      "step": 12233
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2308124486640847e-05,
+      "loss": 0.6579,
+      "step": 12234
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2306983077915052e-05,
+      "loss": 0.7423,
+      "step": 12235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2305841637443146e-05,
+      "loss": 0.6986,
+      "step": 12236
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.230470016524083e-05,
+      "loss": 0.6064,
+      "step": 12237
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2303558661323818e-05,
+      "loss": 0.6785,
+      "step": 12238
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.230241712570781e-05,
+      "loss": 0.6944,
+      "step": 12239
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2301275558408526e-05,
+      "loss": 0.6452,
+      "step": 12240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2300133959441663e-05,
+      "loss": 0.6177,
+      "step": 12241
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2298992328822937e-05,
+      "loss": 0.5993,
+      "step": 12242
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2297850666568054e-05,
+      "loss": 0.6358,
+      "step": 12243
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2296708972692728e-05,
+      "loss": 0.5792,
+      "step": 12244
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2295567247212672e-05,
+      "loss": 0.5857,
+      "step": 12245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2294425490143591e-05,
+      "loss": 0.6434,
+      "step": 12246
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.22932837015012e-05,
+      "loss": 0.5752,
+      "step": 12247
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2292141881301208e-05,
+      "loss": 0.6819,
+      "step": 12248
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2291000029559332e-05,
+      "loss": 0.6027,
+      "step": 12249
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2289858146291283e-05,
+      "loss": 0.6394,
+      "step": 12250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2288716231512772e-05,
+      "loss": 0.6424,
+      "step": 12251
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2287574285239513e-05,
+      "loss": 0.7194,
+      "step": 12252
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2286432307487226e-05,
+      "loss": 0.6521,
+      "step": 12253
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2285290298271619e-05,
+      "loss": 0.6572,
+      "step": 12254
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2284148257608411e-05,
+      "loss": 0.6235,
+      "step": 12255
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2283006185513315e-05,
+      "loss": 0.5772,
+      "step": 12256
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2281864082002046e-05,
+      "loss": 0.5968,
+      "step": 12257
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2280721947090327e-05,
+      "loss": 0.6015,
+      "step": 12258
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2279579780793864e-05,
+      "loss": 0.6217,
+      "step": 12259
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2278437583128384e-05,
+      "loss": 0.6485,
+      "step": 12260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.22772953541096e-05,
+      "loss": 0.6397,
+      "step": 12261
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2276153093753231e-05,
+      "loss": 0.6456,
+      "step": 12262
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2275010802074996e-05,
+      "loss": 0.6338,
+      "step": 12263
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2273868479090614e-05,
+      "loss": 0.6053,
+      "step": 12264
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2272726124815802e-05,
+      "loss": 0.5932,
+      "step": 12265
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227158373926628e-05,
+      "loss": 0.6533,
+      "step": 12266
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2270441322457773e-05,
+      "loss": 0.6482,
+      "step": 12267
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2269298874405995e-05,
+      "loss": 0.6641,
+      "step": 12268
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2268156395126675e-05,
+      "loss": 0.6378,
+      "step": 12269
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2267013884635528e-05,
+      "loss": 0.6345,
+      "step": 12270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2265871342948276e-05,
+      "loss": 0.6564,
+      "step": 12271
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2264728770080647e-05,
+      "loss": 0.5752,
+      "step": 12272
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2263586166048359e-05,
+      "loss": 0.6676,
+      "step": 12273
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2262443530867136e-05,
+      "loss": 0.582,
+      "step": 12274
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2261300864552701e-05,
+      "loss": 0.7279,
+      "step": 12275
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.226015816712078e-05,
+      "loss": 0.6381,
+      "step": 12276
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2259015438587096e-05,
+      "loss": 0.6639,
+      "step": 12277
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2257872678967375e-05,
+      "loss": 0.623,
+      "step": 12278
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2256729888277341e-05,
+      "loss": 0.6016,
+      "step": 12279
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2255587066532723e-05,
+      "loss": 0.5852,
+      "step": 12280
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2254444213749243e-05,
+      "loss": 0.5931,
+      "step": 12281
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2253301329942631e-05,
+      "loss": 0.6353,
+      "step": 12282
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2252158415128612e-05,
+      "loss": 0.6264,
+      "step": 12283
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2251015469322915e-05,
+      "loss": 0.6804,
+      "step": 12284
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2249872492541268e-05,
+      "loss": 0.5857,
+      "step": 12285
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2248729484799395e-05,
+      "loss": 0.594,
+      "step": 12286
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2247586446113034e-05,
+      "loss": 0.561,
+      "step": 12287
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2246443376497903e-05,
+      "loss": 0.6204,
+      "step": 12288
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2245300275969742e-05,
+      "loss": 0.6144,
+      "step": 12289
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2244157144544273e-05,
+      "loss": 0.565,
+      "step": 12290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.224301398223723e-05,
+      "loss": 0.6076,
+      "step": 12291
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2241870789064345e-05,
+      "loss": 0.6841,
+      "step": 12292
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2240727565041346e-05,
+      "loss": 0.6253,
+      "step": 12293
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.223958431018397e-05,
+      "loss": 0.7089,
+      "step": 12294
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2238441024507943e-05,
+      "loss": 0.5992,
+      "step": 12295
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2237297708029004e-05,
+      "loss": 0.6133,
+      "step": 12296
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.223615436076288e-05,
+      "loss": 0.6712,
+      "step": 12297
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2235010982725308e-05,
+      "loss": 0.5903,
+      "step": 12298
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2233867573932022e-05,
+      "loss": 0.674,
+      "step": 12299
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2232724134398755e-05,
+      "loss": 0.5864,
+      "step": 12300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2231580664141241e-05,
+      "loss": 0.5967,
+      "step": 12301
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2230437163175216e-05,
+      "loss": 0.6172,
+      "step": 12302
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2229293631516414e-05,
+      "loss": 0.5575,
+      "step": 12303
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2228150069180575e-05,
+      "loss": 0.6479,
+      "step": 12304
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2227006476183435e-05,
+      "loss": 0.57,
+      "step": 12305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2225862852540727e-05,
+      "loss": 0.5369,
+      "step": 12306
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.222471919826819e-05,
+      "loss": 0.6152,
+      "step": 12307
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2223575513381563e-05,
+      "loss": 0.6332,
+      "step": 12308
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2222431797896586e-05,
+      "loss": 0.7015,
+      "step": 12309
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.222128805182899e-05,
+      "loss": 0.6092,
+      "step": 12310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.222014427519452e-05,
+      "loss": 0.6522,
+      "step": 12311
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2219000468008915e-05,
+      "loss": 0.6051,
+      "step": 12312
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2217856630287912e-05,
+      "loss": 0.6684,
+      "step": 12313
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2216712762047254e-05,
+      "loss": 0.6666,
+      "step": 12314
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2215568863302681e-05,
+      "loss": 0.5701,
+      "step": 12315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2214424934069933e-05,
+      "loss": 0.5674,
+      "step": 12316
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2213280974364751e-05,
+      "loss": 0.6142,
+      "step": 12317
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.221213698420288e-05,
+      "loss": 0.6641,
+      "step": 12318
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.221099296360006e-05,
+      "loss": 0.6134,
+      "step": 12319
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2209848912572036e-05,
+      "loss": 0.7023,
+      "step": 12320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2208704831134545e-05,
+      "loss": 0.5773,
+      "step": 12321
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2207560719303337e-05,
+      "loss": 0.4955,
+      "step": 12322
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2206416577094153e-05,
+      "loss": 0.5443,
+      "step": 12323
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2205272404522739e-05,
+      "loss": 0.5373,
+      "step": 12324
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.220412820160484e-05,
+      "loss": 0.6559,
+      "step": 12325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2202983968356197e-05,
+      "loss": 0.5845,
+      "step": 12326
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2201839704792562e-05,
+      "loss": 0.6109,
+      "step": 12327
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2200695410929677e-05,
+      "loss": 0.6308,
+      "step": 12328
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219955108678329e-05,
+      "loss": 0.5795,
+      "step": 12329
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2198406732369147e-05,
+      "loss": 0.6243,
+      "step": 12330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2197262347702993e-05,
+      "loss": 0.6269,
+      "step": 12331
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2196117932800581e-05,
+      "loss": 0.6609,
+      "step": 12332
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2194973487677655e-05,
+      "loss": 0.5386,
+      "step": 12333
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2193829012349968e-05,
+      "loss": 0.6343,
+      "step": 12334
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2192684506833264e-05,
+      "loss": 0.6591,
+      "step": 12335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2191539971143296e-05,
+      "loss": 0.6602,
+      "step": 12336
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2190395405295811e-05,
+      "loss": 0.6145,
+      "step": 12337
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.218925080930656e-05,
+      "loss": 0.6058,
+      "step": 12338
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2188106183191298e-05,
+      "loss": 0.596,
+      "step": 12339
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2186961526965768e-05,
+      "loss": 0.5989,
+      "step": 12340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2185816840645727e-05,
+      "loss": 0.7124,
+      "step": 12341
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2184672124246928e-05,
+      "loss": 0.5931,
+      "step": 12342
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2183527377785116e-05,
+      "loss": 0.6653,
+      "step": 12343
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2182382601276053e-05,
+      "loss": 0.6449,
+      "step": 12344
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2181237794735486e-05,
+      "loss": 0.6041,
+      "step": 12345
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2180092958179171e-05,
+      "loss": 0.6693,
+      "step": 12346
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.217894809162286e-05,
+      "loss": 0.5929,
+      "step": 12347
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.217780319508231e-05,
+      "loss": 0.6099,
+      "step": 12348
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2176658268573273e-05,
+      "loss": 0.5879,
+      "step": 12349
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2175513312111507e-05,
+      "loss": 0.6388,
+      "step": 12350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2174368325712766e-05,
+      "loss": 0.5831,
+      "step": 12351
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2173223309392808e-05,
+      "loss": 0.62,
+      "step": 12352
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2172078263167387e-05,
+      "loss": 0.6205,
+      "step": 12353
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2170933187052257e-05,
+      "loss": 0.6305,
+      "step": 12354
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2169788081063181e-05,
+      "loss": 0.5722,
+      "step": 12355
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2168642945215916e-05,
+      "loss": 0.6719,
+      "step": 12356
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2167497779526215e-05,
+      "loss": 0.6163,
+      "step": 12357
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2166352584009843e-05,
+      "loss": 0.6206,
+      "step": 12358
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2165207358682555e-05,
+      "loss": 0.5336,
+      "step": 12359
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2164062103560112e-05,
+      "loss": 0.6555,
+      "step": 12360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2162916818658272e-05,
+      "loss": 0.5871,
+      "step": 12361
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2161771503992798e-05,
+      "loss": 0.6483,
+      "step": 12362
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2160626159579447e-05,
+      "loss": 0.5682,
+      "step": 12363
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2159480785433978e-05,
+      "loss": 0.5953,
+      "step": 12364
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2158335381572162e-05,
+      "loss": 0.6661,
+      "step": 12365
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2157189948009752e-05,
+      "loss": 0.6596,
+      "step": 12366
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2156044484762514e-05,
+      "loss": 0.642,
+      "step": 12367
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2154898991846207e-05,
+      "loss": 0.5949,
+      "step": 12368
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2153753469276599e-05,
+      "loss": 0.5918,
+      "step": 12369
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.215260791706945e-05,
+      "loss": 0.6181,
+      "step": 12370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2151462335240524e-05,
+      "loss": 0.5975,
+      "step": 12371
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2150316723805586e-05,
+      "loss": 0.6516,
+      "step": 12372
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2149171082780402e-05,
+      "loss": 0.6084,
+      "step": 12373
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2148025412180733e-05,
+      "loss": 0.6686,
+      "step": 12374
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2146879712022348e-05,
+      "loss": 0.6155,
+      "step": 12375
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2145733982321012e-05,
+      "loss": 0.6093,
+      "step": 12376
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.214458822309249e-05,
+      "loss": 0.6384,
+      "step": 12377
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2143442434352552e-05,
+      "loss": 0.647,
+      "step": 12378
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2142296616116963e-05,
+      "loss": 0.7007,
+      "step": 12379
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2141150768401486e-05,
+      "loss": 0.5709,
+      "step": 12380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2140004891221895e-05,
+      "loss": 0.609,
+      "step": 12381
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2138858984593956e-05,
+      "loss": 0.5608,
+      "step": 12382
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2137713048533439e-05,
+      "loss": 0.6855,
+      "step": 12383
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2136567083056112e-05,
+      "loss": 0.591,
+      "step": 12384
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2135421088177746e-05,
+      "loss": 0.6616,
+      "step": 12385
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2134275063914104e-05,
+      "loss": 0.6246,
+      "step": 12386
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.213312901028097e-05,
+      "loss": 0.6547,
+      "step": 12387
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2131982927294099e-05,
+      "loss": 0.6792,
+      "step": 12388
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2130836814969275e-05,
+      "loss": 0.643,
+      "step": 12389
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2129690673322261e-05,
+      "loss": 0.5814,
+      "step": 12390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212854450236883e-05,
+      "loss": 0.6624,
+      "step": 12391
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2127398302124761e-05,
+      "loss": 0.6107,
+      "step": 12392
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212625207260582e-05,
+      "loss": 0.6875,
+      "step": 12393
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2125105813827782e-05,
+      "loss": 0.6174,
+      "step": 12394
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2123959525806421e-05,
+      "loss": 0.611,
+      "step": 12395
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2122813208557512e-05,
+      "loss": 0.5944,
+      "step": 12396
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2121666862096827e-05,
+      "loss": 0.6555,
+      "step": 12397
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2120520486440142e-05,
+      "loss": 0.6862,
+      "step": 12398
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.211937408160323e-05,
+      "loss": 0.5878,
+      "step": 12399
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.211822764760187e-05,
+      "loss": 0.6536,
+      "step": 12400
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2117081184451835e-05,
+      "loss": 0.618,
+      "step": 12401
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2115934692168902e-05,
+      "loss": 0.6306,
+      "step": 12402
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2114788170768852e-05,
+      "loss": 0.6366,
+      "step": 12403
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2113641620267456e-05,
+      "loss": 0.5532,
+      "step": 12404
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2112495040680496e-05,
+      "loss": 0.6553,
+      "step": 12405
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2111348432023746e-05,
+      "loss": 0.579,
+      "step": 12406
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2110201794312987e-05,
+      "loss": 0.669,
+      "step": 12407
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2109055127563998e-05,
+      "loss": 0.5687,
+      "step": 12408
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2107908431792556e-05,
+      "loss": 0.6221,
+      "step": 12409
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2106761707014443e-05,
+      "loss": 0.6214,
+      "step": 12410
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2105614953245436e-05,
+      "loss": 0.6494,
+      "step": 12411
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210446817050132e-05,
+      "loss": 0.6868,
+      "step": 12412
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2103321358797868e-05,
+      "loss": 0.6541,
+      "step": 12413
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210217451815087e-05,
+      "loss": 0.6227,
+      "step": 12414
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2101027648576103e-05,
+      "loss": 0.5776,
+      "step": 12415
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209988075008935e-05,
+      "loss": 0.6344,
+      "step": 12416
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209873382270639e-05,
+      "loss": 0.6617,
+      "step": 12417
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2097586866443011e-05,
+      "loss": 0.5908,
+      "step": 12418
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2096439881314991e-05,
+      "loss": 0.6006,
+      "step": 12419
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2095292867338122e-05,
+      "loss": 0.6087,
+      "step": 12420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2094145824528176e-05,
+      "loss": 0.5599,
+      "step": 12421
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2092998752900945e-05,
+      "loss": 0.5989,
+      "step": 12422
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2091851652472213e-05,
+      "loss": 0.5892,
+      "step": 12423
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2090704523257763e-05,
+      "loss": 0.5867,
+      "step": 12424
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2089557365273383e-05,
+      "loss": 0.6851,
+      "step": 12425
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2088410178534857e-05,
+      "loss": 0.5733,
+      "step": 12426
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2087262963057973e-05,
+      "loss": 0.597,
+      "step": 12427
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2086115718858513e-05,
+      "loss": 0.6176,
+      "step": 12428
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2084968445952272e-05,
+      "loss": 0.6113,
+      "step": 12429
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.208382114435503e-05,
+      "loss": 0.5894,
+      "step": 12430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2082673814082581e-05,
+      "loss": 0.6221,
+      "step": 12431
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2081526455150707e-05,
+      "loss": 0.5941,
+      "step": 12432
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.20803790675752e-05,
+      "loss": 0.5942,
+      "step": 12433
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2079231651371854e-05,
+      "loss": 0.5751,
+      "step": 12434
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2078084206556448e-05,
+      "loss": 0.6541,
+      "step": 12435
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2076936733144781e-05,
+      "loss": 0.6241,
+      "step": 12436
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2075789231152637e-05,
+      "loss": 0.6093,
+      "step": 12437
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2074641700595813e-05,
+      "loss": 0.6557,
+      "step": 12438
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2073494141490092e-05,
+      "loss": 0.6794,
+      "step": 12439
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2072346553851273e-05,
+      "loss": 0.5871,
+      "step": 12440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2071198937695141e-05,
+      "loss": 0.661,
+      "step": 12441
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2070051293037493e-05,
+      "loss": 0.6318,
+      "step": 12442
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.206890361989412e-05,
+      "loss": 0.6569,
+      "step": 12443
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2067755918280817e-05,
+      "loss": 0.649,
+      "step": 12444
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2066608188213376e-05,
+      "loss": 0.5293,
+      "step": 12445
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2065460429707589e-05,
+      "loss": 0.622,
+      "step": 12446
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2064312642779253e-05,
+      "loss": 0.6101,
+      "step": 12447
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2063164827444158e-05,
+      "loss": 0.6152,
+      "step": 12448
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2062016983718106e-05,
+      "loss": 0.7232,
+      "step": 12449
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2060869111616886e-05,
+      "loss": 0.6613,
+      "step": 12450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2059721211156296e-05,
+      "loss": 0.6695,
+      "step": 12451
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2058573282352132e-05,
+      "loss": 0.6316,
+      "step": 12452
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2057425325220193e-05,
+      "loss": 0.6049,
+      "step": 12453
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2056277339776275e-05,
+      "loss": 0.6182,
+      "step": 12454
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.205512932603617e-05,
+      "loss": 0.5853,
+      "step": 12455
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2053981284015682e-05,
+      "loss": 0.6534,
+      "step": 12456
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2052833213730607e-05,
+      "loss": 0.6311,
+      "step": 12457
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.205168511519674e-05,
+      "loss": 0.6694,
+      "step": 12458
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2050536988429886e-05,
+      "loss": 0.6381,
+      "step": 12459
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2049388833445842e-05,
+      "loss": 0.572,
+      "step": 12460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2048240650260402e-05,
+      "loss": 0.6031,
+      "step": 12461
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2047092438889378e-05,
+      "loss": 0.6608,
+      "step": 12462
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2045944199348558e-05,
+      "loss": 0.6481,
+      "step": 12463
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.204479593165375e-05,
+      "loss": 0.6869,
+      "step": 12464
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2043647635820755e-05,
+      "loss": 0.6406,
+      "step": 12465
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2042499311865367e-05,
+      "loss": 0.5561,
+      "step": 12466
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.20413509598034e-05,
+      "loss": 0.6034,
+      "step": 12467
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2040202579650649e-05,
+      "loss": 0.635,
+      "step": 12468
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2039054171422916e-05,
+      "loss": 0.575,
+      "step": 12469
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2037905735136006e-05,
+      "loss": 0.6325,
+      "step": 12470
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2036757270805724e-05,
+      "loss": 0.5987,
+      "step": 12471
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2035608778447869e-05,
+      "loss": 0.6858,
+      "step": 12472
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.203446025807825e-05,
+      "loss": 0.5684,
+      "step": 12473
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2033311709712675e-05,
+      "loss": 0.6347,
+      "step": 12474
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2032163133366939e-05,
+      "loss": 0.6248,
+      "step": 12475
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2031014529056852e-05,
+      "loss": 0.5837,
+      "step": 12476
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2029865896798222e-05,
+      "loss": 0.6311,
+      "step": 12477
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2028717236606857e-05,
+      "loss": 0.5739,
+      "step": 12478
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2027568548498556e-05,
+      "loss": 0.6017,
+      "step": 12479
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2026419832489133e-05,
+      "loss": 0.5711,
+      "step": 12480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2025271088594392e-05,
+      "loss": 0.6286,
+      "step": 12481
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202412231683014e-05,
+      "loss": 0.6658,
+      "step": 12482
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2022973517212186e-05,
+      "loss": 0.6597,
+      "step": 12483
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202182468975634e-05,
+      "loss": 0.5877,
+      "step": 12484
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202067583447841e-05,
+      "loss": 0.6219,
+      "step": 12485
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2019526951394203e-05,
+      "loss": 0.6617,
+      "step": 12486
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2018378040519534e-05,
+      "loss": 0.6234,
+      "step": 12487
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2017229101870206e-05,
+      "loss": 0.5877,
+      "step": 12488
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2016080135462034e-05,
+      "loss": 0.6874,
+      "step": 12489
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.201493114131083e-05,
+      "loss": 0.6055,
+      "step": 12490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.20137821194324e-05,
+      "loss": 0.6526,
+      "step": 12491
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.201263306984256e-05,
+      "loss": 0.6147,
+      "step": 12492
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.201148399255712e-05,
+      "loss": 0.611,
+      "step": 12493
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2010334887591894e-05,
+      "loss": 0.6804,
+      "step": 12494
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2009185754962693e-05,
+      "loss": 0.6192,
+      "step": 12495
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2008036594685333e-05,
+      "loss": 0.6424,
+      "step": 12496
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2006887406775621e-05,
+      "loss": 0.58,
+      "step": 12497
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2005738191249379e-05,
+      "loss": 0.5733,
+      "step": 12498
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2004588948122414e-05,
+      "loss": 0.6531,
+      "step": 12499
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2003439677410542e-05,
+      "loss": 0.6582,
+      "step": 12500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2002290379129582e-05,
+      "loss": 0.5566,
+      "step": 12501
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2001141053295347e-05,
+      "loss": 0.6763,
+      "step": 12502
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1999991699923654e-05,
+      "loss": 0.6414,
+      "step": 12503
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1998842319030315e-05,
+      "loss": 0.6878,
+      "step": 12504
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199769291063115e-05,
+      "loss": 0.5582,
+      "step": 12505
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1996543474741977e-05,
+      "loss": 0.6175,
+      "step": 12506
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1995394011378611e-05,
+      "loss": 0.6247,
+      "step": 12507
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199424452055687e-05,
+      "loss": 0.6567,
+      "step": 12508
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1993095002292573e-05,
+      "loss": 0.6217,
+      "step": 12509
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1991945456601534e-05,
+      "loss": 0.5623,
+      "step": 12510
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1990795883499578e-05,
+      "loss": 0.6587,
+      "step": 12511
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198964628300252e-05,
+      "loss": 0.5767,
+      "step": 12512
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1988496655126182e-05,
+      "loss": 0.6943,
+      "step": 12513
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1987346999886384e-05,
+      "loss": 0.6781,
+      "step": 12514
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1986197317298943e-05,
+      "loss": 0.7172,
+      "step": 12515
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1985047607379683e-05,
+      "loss": 0.6477,
+      "step": 12516
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1983897870144423e-05,
+      "loss": 0.67,
+      "step": 12517
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1982748105608985e-05,
+      "loss": 0.6386,
+      "step": 12518
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1981598313789192e-05,
+      "loss": 0.5548,
+      "step": 12519
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1980448494700861e-05,
+      "loss": 0.6312,
+      "step": 12520
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1979298648359823e-05,
+      "loss": 0.6322,
+      "step": 12521
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1978148774781895e-05,
+      "loss": 0.6586,
+      "step": 12522
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1976998873982902e-05,
+      "loss": 0.6622,
+      "step": 12523
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1975848945978665e-05,
+      "loss": 0.6088,
+      "step": 12524
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1974698990785015e-05,
+      "loss": 0.579,
+      "step": 12525
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1973549008417767e-05,
+      "loss": 0.6591,
+      "step": 12526
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1972398998892753e-05,
+      "loss": 0.723,
+      "step": 12527
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1971248962225797e-05,
+      "loss": 0.7037,
+      "step": 12528
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1970098898432721e-05,
+      "loss": 0.5907,
+      "step": 12529
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1968948807529353e-05,
+      "loss": 0.5954,
+      "step": 12530
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196779868953152e-05,
+      "loss": 0.6479,
+      "step": 12531
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1966648544455048e-05,
+      "loss": 0.5624,
+      "step": 12532
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1965498372315762e-05,
+      "loss": 0.6772,
+      "step": 12533
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1964348173129497e-05,
+      "loss": 0.5866,
+      "step": 12534
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1963197946912068e-05,
+      "loss": 0.6205,
+      "step": 12535
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1962047693679315e-05,
+      "loss": 0.6639,
+      "step": 12536
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1960897413447059e-05,
+      "loss": 0.6389,
+      "step": 12537
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1959747106231134e-05,
+      "loss": 0.7245,
+      "step": 12538
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1958596772047367e-05,
+      "loss": 0.6224,
+      "step": 12539
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1957446410911586e-05,
+      "loss": 0.6268,
+      "step": 12540
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1956296022839622e-05,
+      "loss": 0.6029,
+      "step": 12541
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1955145607847306e-05,
+      "loss": 0.5656,
+      "step": 12542
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1953995165950469e-05,
+      "loss": 0.6209,
+      "step": 12543
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.195284469716494e-05,
+      "loss": 0.6691,
+      "step": 12544
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1951694201506556e-05,
+      "loss": 0.6514,
+      "step": 12545
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1950543678991142e-05,
+      "loss": 0.6567,
+      "step": 12546
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1949393129634534e-05,
+      "loss": 0.6167,
+      "step": 12547
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1948242553452563e-05,
+      "loss": 0.6492,
+      "step": 12548
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1947091950461062e-05,
+      "loss": 0.5788,
+      "step": 12549
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1945941320675866e-05,
+      "loss": 0.5866,
+      "step": 12550
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1944790664112807e-05,
+      "loss": 0.61,
+      "step": 12551
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.194363998078772e-05,
+      "loss": 0.6804,
+      "step": 12552
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1942489270716437e-05,
+      "loss": 0.5998,
+      "step": 12553
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1941338533914798e-05,
+      "loss": 0.5541,
+      "step": 12554
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1940187770398633e-05,
+      "loss": 0.6539,
+      "step": 12555
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1939036980183781e-05,
+      "loss": 0.6103,
+      "step": 12556
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1937886163286075e-05,
+      "loss": 0.5585,
+      "step": 12557
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1936735319721354e-05,
+      "loss": 0.6757,
+      "step": 12558
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1935584449505451e-05,
+      "loss": 0.6789,
+      "step": 12559
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1934433552654208e-05,
+      "loss": 0.5804,
+      "step": 12560
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1933282629183458e-05,
+      "loss": 0.598,
+      "step": 12561
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1932131679109038e-05,
+      "loss": 0.5746,
+      "step": 12562
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1930980702446793e-05,
+      "loss": 0.6146,
+      "step": 12563
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1929829699212554e-05,
+      "loss": 0.6429,
+      "step": 12564
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1928678669422164e-05,
+      "loss": 0.6358,
+      "step": 12565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1927527613091459e-05,
+      "loss": 0.6231,
+      "step": 12566
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1926376530236283e-05,
+      "loss": 0.6627,
+      "step": 12567
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1925225420872471e-05,
+      "loss": 0.5948,
+      "step": 12568
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1924074285015867e-05,
+      "loss": 0.6006,
+      "step": 12569
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1922923122682308e-05,
+      "loss": 0.638,
+      "step": 12570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192177193388764e-05,
+      "loss": 0.555,
+      "step": 12571
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.19206207186477e-05,
+      "loss": 0.5504,
+      "step": 12572
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1919469476978329e-05,
+      "loss": 0.6847,
+      "step": 12573
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1918318208895377e-05,
+      "loss": 0.6041,
+      "step": 12574
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1917166914414678e-05,
+      "loss": 0.6488,
+      "step": 12575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1916015593552077e-05,
+      "loss": 0.6332,
+      "step": 12576
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1914864246323417e-05,
+      "loss": 0.6805,
+      "step": 12577
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1913712872744545e-05,
+      "loss": 0.6779,
+      "step": 12578
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1912561472831301e-05,
+      "loss": 0.5759,
+      "step": 12579
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1911410046599533e-05,
+      "loss": 0.6615,
+      "step": 12580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1910258594065079e-05,
+      "loss": 0.6211,
+      "step": 12581
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1909107115243788e-05,
+      "loss": 0.6108,
+      "step": 12582
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1907955610151509e-05,
+      "loss": 0.5643,
+      "step": 12583
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1906804078804081e-05,
+      "loss": 0.6444,
+      "step": 12584
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1905652521217359e-05,
+      "loss": 0.6283,
+      "step": 12585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1904500937407181e-05,
+      "loss": 0.6853,
+      "step": 12586
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1903349327389397e-05,
+      "loss": 0.5943,
+      "step": 12587
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1902197691179853e-05,
+      "loss": 0.5717,
+      "step": 12588
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1901046028794396e-05,
+      "loss": 0.6017,
+      "step": 12589
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1899894340248879e-05,
+      "loss": 0.6253,
+      "step": 12590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1898742625559141e-05,
+      "loss": 0.6396,
+      "step": 12591
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1897590884741042e-05,
+      "loss": 0.6147,
+      "step": 12592
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1896439117810421e-05,
+      "loss": 0.6485,
+      "step": 12593
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1895287324783134e-05,
+      "loss": 0.6012,
+      "step": 12594
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1894135505675025e-05,
+      "loss": 0.5903,
+      "step": 12595
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1892983660501948e-05,
+      "loss": 0.638,
+      "step": 12596
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1891831789279752e-05,
+      "loss": 0.618,
+      "step": 12597
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.189067989202429e-05,
+      "loss": 0.654,
+      "step": 12598
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1889527968751413e-05,
+      "loss": 0.6008,
+      "step": 12599
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1888376019476966e-05,
+      "loss": 0.586,
+      "step": 12600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188722404421681e-05,
+      "loss": 0.6281,
+      "step": 12601
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1886072042986791e-05,
+      "loss": 0.6753,
+      "step": 12602
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1884920015802765e-05,
+      "loss": 0.6312,
+      "step": 12603
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1883767962680581e-05,
+      "loss": 0.524,
+      "step": 12604
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1882615883636097e-05,
+      "loss": 0.6116,
+      "step": 12605
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1881463778685163e-05,
+      "loss": 0.6253,
+      "step": 12606
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1880311647843635e-05,
+      "loss": 0.5614,
+      "step": 12607
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1879159491127363e-05,
+      "loss": 0.6149,
+      "step": 12608
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1878007308552208e-05,
+      "loss": 0.6328,
+      "step": 12609
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1876855100134024e-05,
+      "loss": 0.6069,
+      "step": 12610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.187570286588866e-05,
+      "loss": 0.6126,
+      "step": 12611
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1874550605831982e-05,
+      "loss": 0.6828,
+      "step": 12612
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1873398319979837e-05,
+      "loss": 0.6477,
+      "step": 12613
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1872246008348088e-05,
+      "loss": 0.675,
+      "step": 12614
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1871093670952585e-05,
+      "loss": 0.6114,
+      "step": 12615
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1869941307809191e-05,
+      "loss": 0.6515,
+      "step": 12616
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186878891893376e-05,
+      "loss": 0.6403,
+      "step": 12617
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1867636504342155e-05,
+      "loss": 0.6379,
+      "step": 12618
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1866484064050228e-05,
+      "loss": 0.7354,
+      "step": 12619
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186533159807384e-05,
+      "loss": 0.6243,
+      "step": 12620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186417910642885e-05,
+      "loss": 0.6434,
+      "step": 12621
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1863026589131118e-05,
+      "loss": 0.5948,
+      "step": 12622
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1861874046196504e-05,
+      "loss": 0.5825,
+      "step": 12623
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1860721477640866e-05,
+      "loss": 0.6979,
+      "step": 12624
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1859568883480067e-05,
+      "loss": 0.6492,
+      "step": 12625
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1858416263729965e-05,
+      "loss": 0.6259,
+      "step": 12626
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1857263618406423e-05,
+      "loss": 0.652,
+      "step": 12627
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.18561109475253e-05,
+      "loss": 0.6226,
+      "step": 12628
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1854958251102459e-05,
+      "loss": 0.645,
+      "step": 12629
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1853805529153766e-05,
+      "loss": 0.6035,
+      "step": 12630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1852652781695078e-05,
+      "loss": 0.628,
+      "step": 12631
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1851500008742261e-05,
+      "loss": 0.6179,
+      "step": 12632
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1850347210311175e-05,
+      "loss": 0.6091,
+      "step": 12633
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1849194386417686e-05,
+      "loss": 0.6448,
+      "step": 12634
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1848041537077657e-05,
+      "loss": 0.664,
+      "step": 12635
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1846888662306954e-05,
+      "loss": 0.573,
+      "step": 12636
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184573576212144e-05,
+      "loss": 0.6575,
+      "step": 12637
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1844582836536982e-05,
+      "loss": 0.6966,
+      "step": 12638
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184342988556944e-05,
+      "loss": 0.5673,
+      "step": 12639
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1842276909234682e-05,
+      "loss": 0.6118,
+      "step": 12640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184112390754858e-05,
+      "loss": 0.5813,
+      "step": 12641
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.183997088052699e-05,
+      "loss": 0.6365,
+      "step": 12642
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1838817828185787e-05,
+      "loss": 0.6064,
+      "step": 12643
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1837664750540834e-05,
+      "loss": 0.6372,
+      "step": 12644
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1836511647608e-05,
+      "loss": 0.6524,
+      "step": 12645
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1835358519403153e-05,
+      "loss": 0.5909,
+      "step": 12646
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1834205365942158e-05,
+      "loss": 0.6212,
+      "step": 12647
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1833052187240886e-05,
+      "loss": 0.6506,
+      "step": 12648
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1831898983315206e-05,
+      "loss": 0.6192,
+      "step": 12649
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1830745754180985e-05,
+      "loss": 0.6175,
+      "step": 12650
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1829592499854094e-05,
+      "loss": 0.615,
+      "step": 12651
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1828439220350405e-05,
+      "loss": 0.6391,
+      "step": 12652
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1827285915685783e-05,
+      "loss": 0.5971,
+      "step": 12653
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1826132585876102e-05,
+      "loss": 0.7158,
+      "step": 12654
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1824979230937233e-05,
+      "loss": 0.6305,
+      "step": 12655
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1823825850885045e-05,
+      "loss": 0.6484,
+      "step": 12656
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.182267244573541e-05,
+      "loss": 0.6404,
+      "step": 12657
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1821519015504203e-05,
+      "loss": 0.5551,
+      "step": 12658
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1820365560207292e-05,
+      "loss": 0.6641,
+      "step": 12659
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1819212079860551e-05,
+      "loss": 0.628,
+      "step": 12660
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1818058574479855e-05,
+      "loss": 0.5659,
+      "step": 12661
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1816905044081076e-05,
+      "loss": 0.6671,
+      "step": 12662
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1815751488680084e-05,
+      "loss": 0.6123,
+      "step": 12663
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1814597908292757e-05,
+      "loss": 0.6606,
+      "step": 12664
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181344430293497e-05,
+      "loss": 0.6584,
+      "step": 12665
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1812290672622595e-05,
+      "loss": 0.627,
+      "step": 12666
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181113701737151e-05,
+      "loss": 0.6787,
+      "step": 12667
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1809983337197586e-05,
+      "loss": 0.6564,
+      "step": 12668
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1808829632116703e-05,
+      "loss": 0.6561,
+      "step": 12669
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1807675902144731e-05,
+      "loss": 0.6165,
+      "step": 12670
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1806522147297552e-05,
+      "loss": 0.6763,
+      "step": 12671
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1805368367591043e-05,
+      "loss": 0.6324,
+      "step": 12672
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1804214563041077e-05,
+      "loss": 0.6255,
+      "step": 12673
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1803060733663535e-05,
+      "loss": 0.6613,
+      "step": 12674
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.180190687947429e-05,
+      "loss": 0.6739,
+      "step": 12675
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1800753000489225e-05,
+      "loss": 0.6032,
+      "step": 12676
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1799599096724215e-05,
+      "loss": 0.5814,
+      "step": 12677
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1798445168195141e-05,
+      "loss": 0.5919,
+      "step": 12678
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1797291214917882e-05,
+      "loss": 0.6845,
+      "step": 12679
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1796137236908315e-05,
+      "loss": 0.6167,
+      "step": 12680
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1794983234182323e-05,
+      "loss": 0.6357,
+      "step": 12681
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1793829206755782e-05,
+      "loss": 0.6085,
+      "step": 12682
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.179267515464458e-05,
+      "loss": 0.6214,
+      "step": 12683
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1791521077864587e-05,
+      "loss": 0.6549,
+      "step": 12684
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1790366976431692e-05,
+      "loss": 0.6726,
+      "step": 12685
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1789212850361775e-05,
+      "loss": 0.6476,
+      "step": 12686
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1788058699670719e-05,
+      "loss": 0.6625,
+      "step": 12687
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1786904524374401e-05,
+      "loss": 0.6022,
+      "step": 12688
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1785750324488708e-05,
+      "loss": 0.59,
+      "step": 12689
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1784596100029519e-05,
+      "loss": 0.5407,
+      "step": 12690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1783441851012722e-05,
+      "loss": 0.6194,
+      "step": 12691
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1782287577454199e-05,
+      "loss": 0.655,
+      "step": 12692
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1781133279369831e-05,
+      "loss": 0.6474,
+      "step": 12693
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1779978956775507e-05,
+      "loss": 0.6886,
+      "step": 12694
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1778824609687103e-05,
+      "loss": 0.6481,
+      "step": 12695
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1777670238120513e-05,
+      "loss": 0.6179,
+      "step": 12696
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1776515842091619e-05,
+      "loss": 0.6466,
+      "step": 12697
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1775361421616303e-05,
+      "loss": 0.6707,
+      "step": 12698
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1774206976710455e-05,
+      "loss": 0.6082,
+      "step": 12699
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1773052507389962e-05,
+      "loss": 0.6463,
+      "step": 12700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1771898013670708e-05,
+      "loss": 0.6717,
+      "step": 12701
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1770743495568579e-05,
+      "loss": 0.5423,
+      "step": 12702
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1769588953099464e-05,
+      "loss": 0.6587,
+      "step": 12703
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.176843438627925e-05,
+      "loss": 0.5929,
+      "step": 12704
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1767279795123827e-05,
+      "loss": 0.6488,
+      "step": 12705
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1766125179649079e-05,
+      "loss": 0.5999,
+      "step": 12706
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1764970539870899e-05,
+      "loss": 0.5827,
+      "step": 12707
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.176381587580517e-05,
+      "loss": 0.7167,
+      "step": 12708
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1762661187467785e-05,
+      "loss": 0.656,
+      "step": 12709
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1761506474874635e-05,
+      "loss": 0.6048,
+      "step": 12710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1760351738041609e-05,
+      "loss": 0.6686,
+      "step": 12711
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1759196976984595e-05,
+      "loss": 0.6953,
+      "step": 12712
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1758042191719485e-05,
+      "loss": 0.601,
+      "step": 12713
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.175688738226217e-05,
+      "loss": 0.6253,
+      "step": 12714
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.175573254862854e-05,
+      "loss": 0.6662,
+      "step": 12715
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1754577690834488e-05,
+      "loss": 0.6652,
+      "step": 12716
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1753422808895904e-05,
+      "loss": 0.6039,
+      "step": 12717
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1752267902828682e-05,
+      "loss": 0.6197,
+      "step": 12718
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1751112972648717e-05,
+      "loss": 0.6667,
+      "step": 12719
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1749958018371895e-05,
+      "loss": 0.7082,
+      "step": 12720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1748803040014116e-05,
+      "loss": 0.5437,
+      "step": 12721
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1747648037591268e-05,
+      "loss": 0.6719,
+      "step": 12722
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1746493011119248e-05,
+      "loss": 0.618,
+      "step": 12723
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.174533796061395e-05,
+      "loss": 0.6556,
+      "step": 12724
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1744182886091269e-05,
+      "loss": 0.5955,
+      "step": 12725
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1743027787567099e-05,
+      "loss": 0.6483,
+      "step": 12726
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1741872665057334e-05,
+      "loss": 0.6703,
+      "step": 12727
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1740717518577869e-05,
+      "loss": 0.6268,
+      "step": 12728
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1739562348144602e-05,
+      "loss": 0.6067,
+      "step": 12729
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.173840715377343e-05,
+      "loss": 0.6657,
+      "step": 12730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1737251935480247e-05,
+      "loss": 0.5959,
+      "step": 12731
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1736096693280953e-05,
+      "loss": 0.5928,
+      "step": 12732
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1734941427191442e-05,
+      "loss": 0.6723,
+      "step": 12733
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1733786137227612e-05,
+      "loss": 0.6357,
+      "step": 12734
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1732630823405362e-05,
+      "loss": 0.6406,
+      "step": 12735
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1731475485740589e-05,
+      "loss": 0.586,
+      "step": 12736
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1730320124249191e-05,
+      "loss": 0.5596,
+      "step": 12737
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.172916473894707e-05,
+      "loss": 0.6418,
+      "step": 12738
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.172800932985012e-05,
+      "loss": 0.632,
+      "step": 12739
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1726853896974245e-05,
+      "loss": 0.6472,
+      "step": 12740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1725698440335343e-05,
+      "loss": 0.6935,
+      "step": 12741
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1724542959949313e-05,
+      "loss": 0.5145,
+      "step": 12742
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1723387455832058e-05,
+      "loss": 0.5984,
+      "step": 12743
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1722231927999474e-05,
+      "loss": 0.5921,
+      "step": 12744
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1721076376467471e-05,
+      "loss": 0.583,
+      "step": 12745
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.171992080125194e-05,
+      "loss": 0.5782,
+      "step": 12746
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1718765202368791e-05,
+      "loss": 0.5889,
+      "step": 12747
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.171760957983392e-05,
+      "loss": 0.6322,
+      "step": 12748
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.171645393366323e-05,
+      "loss": 0.5738,
+      "step": 12749
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1715298263872632e-05,
+      "loss": 0.6428,
+      "step": 12750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1714142570478018e-05,
+      "loss": 0.6068,
+      "step": 12751
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1712986853495298e-05,
+      "loss": 0.6211,
+      "step": 12752
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1711831112940371e-05,
+      "loss": 0.6415,
+      "step": 12753
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1710675348829148e-05,
+      "loss": 0.6938,
+      "step": 12754
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1709519561177524e-05,
+      "loss": 0.6555,
+      "step": 12755
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1708363750001415e-05,
+      "loss": 0.648,
+      "step": 12756
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1707207915316714e-05,
+      "loss": 0.6637,
+      "step": 12757
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1706052057139335e-05,
+      "loss": 0.6454,
+      "step": 12758
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1704896175485179e-05,
+      "loss": 0.5976,
+      "step": 12759
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1703740270370154e-05,
+      "loss": 0.6648,
+      "step": 12760
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1702584341810166e-05,
+      "loss": 0.7031,
+      "step": 12761
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1701428389821121e-05,
+      "loss": 0.6473,
+      "step": 12762
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1700272414418929e-05,
+      "loss": 0.6548,
+      "step": 12763
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1699116415619492e-05,
+      "loss": 0.6658,
+      "step": 12764
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1697960393438721e-05,
+      "loss": 0.6533,
+      "step": 12765
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1696804347892524e-05,
+      "loss": 0.7525,
+      "step": 12766
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1695648278996808e-05,
+      "loss": 0.6424,
+      "step": 12767
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.169449218676748e-05,
+      "loss": 0.6226,
+      "step": 12768
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1693336071220452e-05,
+      "loss": 0.6512,
+      "step": 12769
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.169217993237163e-05,
+      "loss": 0.7032,
+      "step": 12770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1691023770236925e-05,
+      "loss": 0.6048,
+      "step": 12771
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1689867584832248e-05,
+      "loss": 0.6447,
+      "step": 12772
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1688711376173509e-05,
+      "loss": 0.5501,
+      "step": 12773
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.168755514427662e-05,
+      "loss": 0.6526,
+      "step": 12774
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1686398889157485e-05,
+      "loss": 0.6186,
+      "step": 12775
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1685242610832022e-05,
+      "loss": 0.6585,
+      "step": 12776
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1684086309316136e-05,
+      "loss": 0.5786,
+      "step": 12777
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1682929984625749e-05,
+      "loss": 0.6618,
+      "step": 12778
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.168177363677676e-05,
+      "loss": 0.6961,
+      "step": 12779
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1680617265785093e-05,
+      "loss": 0.6875,
+      "step": 12780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1679460871666652e-05,
+      "loss": 0.6814,
+      "step": 12781
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1678304454437358e-05,
+      "loss": 0.5611,
+      "step": 12782
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1677148014113117e-05,
+      "loss": 0.6425,
+      "step": 12783
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1675991550709846e-05,
+      "loss": 0.6693,
+      "step": 12784
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1674835064243459e-05,
+      "loss": 0.6169,
+      "step": 12785
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1673678554729871e-05,
+      "loss": 0.6163,
+      "step": 12786
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1672522022184993e-05,
+      "loss": 0.6811,
+      "step": 12787
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1671365466624744e-05,
+      "loss": 0.6346,
+      "step": 12788
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1670208888065038e-05,
+      "loss": 0.6027,
+      "step": 12789
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.166905228652179e-05,
+      "loss": 0.6124,
+      "step": 12790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1667895662010913e-05,
+      "loss": 0.6618,
+      "step": 12791
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.166673901454833e-05,
+      "loss": 0.6425,
+      "step": 12792
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1665582344149952e-05,
+      "loss": 0.6035,
+      "step": 12793
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1664425650831698e-05,
+      "loss": 0.6588,
+      "step": 12794
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1663268934609484e-05,
+      "loss": 0.5522,
+      "step": 12795
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1662112195499229e-05,
+      "loss": 0.5934,
+      "step": 12796
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1660955433516849e-05,
+      "loss": 0.6606,
+      "step": 12797
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1659798648678262e-05,
+      "loss": 0.6611,
+      "step": 12798
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.165864184099939e-05,
+      "loss": 0.6718,
+      "step": 12799
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1657485010496146e-05,
+      "loss": 0.6465,
+      "step": 12800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1656328157184453e-05,
+      "loss": 0.579,
+      "step": 12801
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1655171281080229e-05,
+      "loss": 0.6532,
+      "step": 12802
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1654014382199395e-05,
+      "loss": 0.6019,
+      "step": 12803
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1652857460557867e-05,
+      "loss": 0.6407,
+      "step": 12804
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1651700516171569e-05,
+      "loss": 0.7227,
+      "step": 12805
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1650543549056422e-05,
+      "loss": 0.6368,
+      "step": 12806
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1649386559228342e-05,
+      "loss": 0.6405,
+      "step": 12807
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1648229546703258e-05,
+      "loss": 0.5891,
+      "step": 12808
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1647072511497084e-05,
+      "loss": 0.6467,
+      "step": 12809
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1645915453625746e-05,
+      "loss": 0.6552,
+      "step": 12810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1644758373105165e-05,
+      "loss": 0.6336,
+      "step": 12811
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1643601269951266e-05,
+      "loss": 0.6466,
+      "step": 12812
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1642444144179965e-05,
+      "loss": 0.5855,
+      "step": 12813
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1641286995807193e-05,
+      "loss": 0.6524,
+      "step": 12814
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.164012982484887e-05,
+      "loss": 0.6694,
+      "step": 12815
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1638972631320916e-05,
+      "loss": 0.5805,
+      "step": 12816
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.163781541523926e-05,
+      "loss": 0.5917,
+      "step": 12817
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1636658176619824e-05,
+      "loss": 0.6127,
+      "step": 12818
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1635500915478536e-05,
+      "loss": 0.5956,
+      "step": 12819
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1634343631831317e-05,
+      "loss": 0.6563,
+      "step": 12820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1633186325694094e-05,
+      "loss": 0.5987,
+      "step": 12821
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1632028997082792e-05,
+      "loss": 0.6253,
+      "step": 12822
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1630871646013335e-05,
+      "loss": 0.5977,
+      "step": 12823
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1629714272501653e-05,
+      "loss": 0.5955,
+      "step": 12824
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162855687656367e-05,
+      "loss": 0.6315,
+      "step": 12825
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1627399458215314e-05,
+      "loss": 0.6673,
+      "step": 12826
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1626242017472511e-05,
+      "loss": 0.5736,
+      "step": 12827
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162508455435119e-05,
+      "loss": 0.5954,
+      "step": 12828
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1623927068867276e-05,
+      "loss": 0.555,
+      "step": 12829
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.16227695610367e-05,
+      "loss": 0.523,
+      "step": 12830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1621612030875386e-05,
+      "loss": 0.5962,
+      "step": 12831
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162045447839927e-05,
+      "loss": 0.5857,
+      "step": 12832
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1619296903624274e-05,
+      "loss": 0.6633,
+      "step": 12833
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1618139306566331e-05,
+      "loss": 0.6695,
+      "step": 12834
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1616981687241366e-05,
+      "loss": 0.6506,
+      "step": 12835
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1615824045665315e-05,
+      "loss": 0.6101,
+      "step": 12836
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1614666381854107e-05,
+      "loss": 0.6775,
+      "step": 12837
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1613508695823668e-05,
+      "loss": 0.6259,
+      "step": 12838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1612350987589932e-05,
+      "loss": 0.6756,
+      "step": 12839
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.161119325716883e-05,
+      "loss": 0.5941,
+      "step": 12840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1610035504576293e-05,
+      "loss": 0.6968,
+      "step": 12841
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1608877729828252e-05,
+      "loss": 0.5525,
+      "step": 12842
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1607719932940643e-05,
+      "loss": 0.6984,
+      "step": 12843
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1606562113929392e-05,
+      "loss": 0.6917,
+      "step": 12844
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1605404272810437e-05,
+      "loss": 0.5713,
+      "step": 12845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1604246409599704e-05,
+      "loss": 0.6276,
+      "step": 12846
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1603088524313135e-05,
+      "loss": 0.5687,
+      "step": 12847
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1601930616966656e-05,
+      "loss": 0.5658,
+      "step": 12848
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1600772687576207e-05,
+      "loss": 0.6243,
+      "step": 12849
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159961473615772e-05,
+      "loss": 0.6286,
+      "step": 12850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1598456762727127e-05,
+      "loss": 0.5664,
+      "step": 12851
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1597298767300365e-05,
+      "loss": 0.6588,
+      "step": 12852
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1596140749893366e-05,
+      "loss": 0.5628,
+      "step": 12853
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159498271052207e-05,
+      "loss": 0.6121,
+      "step": 12854
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159382464920241e-05,
+      "loss": 0.6253,
+      "step": 12855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1592666565950324e-05,
+      "loss": 0.6021,
+      "step": 12856
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1591508460781743e-05,
+      "loss": 0.5933,
+      "step": 12857
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159035033371261e-05,
+      "loss": 0.6151,
+      "step": 12858
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1589192184758859e-05,
+      "loss": 0.6367,
+      "step": 12859
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1588034013936427e-05,
+      "loss": 0.6189,
+      "step": 12860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1586875821261252e-05,
+      "loss": 0.6267,
+      "step": 12861
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.158571760674927e-05,
+      "loss": 0.6309,
+      "step": 12862
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1584559370416422e-05,
+      "loss": 0.631,
+      "step": 12863
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.158340111227864e-05,
+      "loss": 0.6847,
+      "step": 12864
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1582242832351873e-05,
+      "loss": 0.5697,
+      "step": 12865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1581084530652051e-05,
+      "loss": 0.5573,
+      "step": 12866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.157992620719512e-05,
+      "loss": 0.6194,
+      "step": 12867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1578767861997014e-05,
+      "loss": 0.5927,
+      "step": 12868
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1577609495073673e-05,
+      "loss": 0.5862,
+      "step": 12869
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1576451106441043e-05,
+      "loss": 0.5697,
+      "step": 12870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1575292696115056e-05,
+      "loss": 0.6588,
+      "step": 12871
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1574134264111662e-05,
+      "loss": 0.6531,
+      "step": 12872
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1572975810446796e-05,
+      "loss": 0.5917,
+      "step": 12873
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1571817335136402e-05,
+      "loss": 0.5961,
+      "step": 12874
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1570658838196416e-05,
+      "loss": 0.6626,
+      "step": 12875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1569500319642789e-05,
+      "loss": 0.6401,
+      "step": 12876
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1568341779491454e-05,
+      "loss": 0.6072,
+      "step": 12877
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1567183217758363e-05,
+      "loss": 0.6595,
+      "step": 12878
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.156602463445945e-05,
+      "loss": 0.5904,
+      "step": 12879
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1564866029610661e-05,
+      "loss": 0.5726,
+      "step": 12880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1563707403227944e-05,
+      "loss": 0.6636,
+      "step": 12881
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1562548755327236e-05,
+      "loss": 0.6775,
+      "step": 12882
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.156139008592449e-05,
+      "loss": 0.6384,
+      "step": 12883
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1560231395035638e-05,
+      "loss": 0.5886,
+      "step": 12884
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1559072682676635e-05,
+      "loss": 0.6859,
+      "step": 12885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.155791394886342e-05,
+      "loss": 0.6917,
+      "step": 12886
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1556755193611941e-05,
+      "loss": 0.5937,
+      "step": 12887
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1555596416938142e-05,
+      "loss": 0.6448,
+      "step": 12888
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1554437618857967e-05,
+      "loss": 0.6407,
+      "step": 12889
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.155327879938737e-05,
+      "loss": 0.5908,
+      "step": 12890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1552119958542288e-05,
+      "loss": 0.6734,
+      "step": 12891
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1550961096338673e-05,
+      "loss": 0.5967,
+      "step": 12892
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1549802212792469e-05,
+      "loss": 0.6392,
+      "step": 12893
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1548643307919627e-05,
+      "loss": 0.6962,
+      "step": 12894
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1547484381736091e-05,
+      "loss": 0.6928,
+      "step": 12895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1546325434257812e-05,
+      "loss": 0.5948,
+      "step": 12896
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1545166465500732e-05,
+      "loss": 0.6155,
+      "step": 12897
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1544007475480809e-05,
+      "loss": 0.6081,
+      "step": 12898
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1542848464213983e-05,
+      "loss": 0.6576,
+      "step": 12899
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1541689431716208e-05,
+      "loss": 0.5726,
+      "step": 12900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1540530378003431e-05,
+      "loss": 0.6365,
+      "step": 12901
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1539371303091605e-05,
+      "loss": 0.5794,
+      "step": 12902
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1538212206996675e-05,
+      "loss": 0.6545,
+      "step": 12903
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153705308973459e-05,
+      "loss": 0.6796,
+      "step": 12904
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1535893951321309e-05,
+      "loss": 0.5714,
+      "step": 12905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1534734791772776e-05,
+      "loss": 0.5949,
+      "step": 12906
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1533575611104943e-05,
+      "loss": 0.6509,
+      "step": 12907
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1532416409333764e-05,
+      "loss": 0.5702,
+      "step": 12908
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1531257186475186e-05,
+      "loss": 0.6242,
+      "step": 12909
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1530097942545167e-05,
+      "loss": 0.5932,
+      "step": 12910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1528938677559652e-05,
+      "loss": 0.658,
+      "step": 12911
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1527779391534601e-05,
+      "loss": 0.5953,
+      "step": 12912
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152662008448596e-05,
+      "loss": 0.5988,
+      "step": 12913
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1525460756429687e-05,
+      "loss": 0.5841,
+      "step": 12914
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1524301407381733e-05,
+      "loss": 0.6618,
+      "step": 12915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152314203735805e-05,
+      "loss": 0.6717,
+      "step": 12916
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.15219826463746e-05,
+      "loss": 0.6325,
+      "step": 12917
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1520823234447327e-05,
+      "loss": 0.6469,
+      "step": 12918
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151966380159219e-05,
+      "loss": 0.635,
+      "step": 12919
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1518504347825146e-05,
+      "loss": 0.691,
+      "step": 12920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1517344873162148e-05,
+      "loss": 0.6052,
+      "step": 12921
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1516185377619149e-05,
+      "loss": 0.5803,
+      "step": 12922
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151502586121211e-05,
+      "loss": 0.5759,
+      "step": 12923
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1513866323956979e-05,
+      "loss": 0.5847,
+      "step": 12924
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1512706765869724e-05,
+      "loss": 0.6457,
+      "step": 12925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1511547186966287e-05,
+      "loss": 0.6882,
+      "step": 12926
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1510387587262637e-05,
+      "loss": 0.6186,
+      "step": 12927
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1509227966774728e-05,
+      "loss": 0.5712,
+      "step": 12928
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1508068325518514e-05,
+      "loss": 0.6055,
+      "step": 12929
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1506908663509954e-05,
+      "loss": 0.6823,
+      "step": 12930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1505748980765006e-05,
+      "loss": 0.6123,
+      "step": 12931
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1504589277299632e-05,
+      "loss": 0.6226,
+      "step": 12932
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1503429553129786e-05,
+      "loss": 0.6905,
+      "step": 12933
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1502269808271429e-05,
+      "loss": 0.6271,
+      "step": 12934
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1501110042740518e-05,
+      "loss": 0.6331,
+      "step": 12935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1499950256553013e-05,
+      "loss": 0.5928,
+      "step": 12936
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1498790449724878e-05,
+      "loss": 0.6076,
+      "step": 12937
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1497630622272067e-05,
+      "loss": 0.6298,
+      "step": 12938
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1496470774210544e-05,
+      "loss": 0.5972,
+      "step": 12939
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1495310905556267e-05,
+      "loss": 0.7023,
+      "step": 12940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.14941510163252e-05,
+      "loss": 0.645,
+      "step": 12941
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1492991106533297e-05,
+      "loss": 0.6133,
+      "step": 12942
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1491831176196531e-05,
+      "loss": 0.6332,
+      "step": 12943
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1490671225330855e-05,
+      "loss": 0.6361,
+      "step": 12944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1489511253952232e-05,
+      "loss": 0.665,
+      "step": 12945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1488351262076625e-05,
+      "loss": 0.6185,
+      "step": 12946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1487191249719998e-05,
+      "loss": 0.5561,
+      "step": 12947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1486031216898312e-05,
+      "loss": 0.6119,
+      "step": 12948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.148487116362753e-05,
+      "loss": 0.5879,
+      "step": 12949
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.148371108992362e-05,
+      "loss": 0.6351,
+      "step": 12950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1482550995802537e-05,
+      "loss": 0.6468,
+      "step": 12951
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1481390881280252e-05,
+      "loss": 0.5839,
+      "step": 12952
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1480230746372727e-05,
+      "loss": 0.5998,
+      "step": 12953
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1479070591095924e-05,
+      "loss": 0.5537,
+      "step": 12954
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1477910415465811e-05,
+      "loss": 0.6487,
+      "step": 12955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1476750219498354e-05,
+      "loss": 0.651,
+      "step": 12956
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1475590003209516e-05,
+      "loss": 0.611,
+      "step": 12957
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.147442976661526e-05,
+      "loss": 0.6483,
+      "step": 12958
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1473269509731557e-05,
+      "loss": 0.5906,
+      "step": 12959
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.147210923257437e-05,
+      "loss": 0.6077,
+      "step": 12960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.147094893515967e-05,
+      "loss": 0.6993,
+      "step": 12961
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1469788617503417e-05,
+      "loss": 0.6156,
+      "step": 12962
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146862827962158e-05,
+      "loss": 0.619,
+      "step": 12963
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1467467921530132e-05,
+      "loss": 0.5891,
+      "step": 12964
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1466307543245033e-05,
+      "loss": 0.6145,
+      "step": 12965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1465147144782252e-05,
+      "loss": 0.6778,
+      "step": 12966
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146398672615776e-05,
+      "loss": 0.6619,
+      "step": 12967
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1462826287387528e-05,
+      "loss": 0.6191,
+      "step": 12968
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1461665828487517e-05,
+      "loss": 0.5546,
+      "step": 12969
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.14605053494737e-05,
+      "loss": 0.5736,
+      "step": 12970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1459344850362047e-05,
+      "loss": 0.5581,
+      "step": 12971
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1458184331168526e-05,
+      "loss": 0.604,
+      "step": 12972
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1457023791909107e-05,
+      "loss": 0.6497,
+      "step": 12973
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1455863232599764e-05,
+      "loss": 0.6277,
+      "step": 12974
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145470265325646e-05,
+      "loss": 0.6214,
+      "step": 12975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145354205389517e-05,
+      "loss": 0.5833,
+      "step": 12976
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1452381434531863e-05,
+      "loss": 0.6707,
+      "step": 12977
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1451220795182511e-05,
+      "loss": 0.6407,
+      "step": 12978
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145006013586309e-05,
+      "loss": 0.6302,
+      "step": 12979
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1448899456589563e-05,
+      "loss": 0.6066,
+      "step": 12980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1447738757377911e-05,
+      "loss": 0.7082,
+      "step": 12981
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1446578038244096e-05,
+      "loss": 0.5935,
+      "step": 12982
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1445417299204101e-05,
+      "loss": 0.6141,
+      "step": 12983
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.144425654027389e-05,
+      "loss": 0.6399,
+      "step": 12984
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1443095761469443e-05,
+      "loss": 0.6671,
+      "step": 12985
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.144193496280673e-05,
+      "loss": 0.6883,
+      "step": 12986
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1440774144301723e-05,
+      "loss": 0.6088,
+      "step": 12987
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1439613305970398e-05,
+      "loss": 0.6213,
+      "step": 12988
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.143845244782873e-05,
+      "loss": 0.7005,
+      "step": 12989
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1437291569892693e-05,
+      "loss": 0.6359,
+      "step": 12990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1436130672178258e-05,
+      "loss": 0.6492,
+      "step": 12991
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1434969754701406e-05,
+      "loss": 0.6556,
+      "step": 12992
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1433808817478107e-05,
+      "loss": 0.666,
+      "step": 12993
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1432647860524342e-05,
+      "loss": 0.5737,
+      "step": 12994
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1431486883856082e-05,
+      "loss": 0.6569,
+      "step": 12995
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1430325887489304e-05,
+      "loss": 0.6493,
+      "step": 12996
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1429164871439985e-05,
+      "loss": 0.5593,
+      "step": 12997
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1428003835724102e-05,
+      "loss": 0.6199,
+      "step": 12998
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1426842780357633e-05,
+      "loss": 0.5675,
+      "step": 12999
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.142568170535655e-05,
+      "loss": 0.6587,
+      "step": 13000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1424520610736837e-05,
+      "loss": 0.6237,
+      "step": 13001
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1423359496514469e-05,
+      "loss": 0.6397,
+      "step": 13002
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1422198362705424e-05,
+      "loss": 0.6105,
+      "step": 13003
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1421037209325676e-05,
+      "loss": 0.5765,
+      "step": 13004
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141987603639121e-05,
+      "loss": 0.5992,
+      "step": 13005
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1418714843918002e-05,
+      "loss": 0.6409,
+      "step": 13006
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141755363192203e-05,
+      "loss": 0.6953,
+      "step": 13007
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1416392400419279e-05,
+      "loss": 0.6619,
+      "step": 13008
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1415231149425721e-05,
+      "loss": 0.672,
+      "step": 13009
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141406987895734e-05,
+      "loss": 0.6333,
+      "step": 13010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1412908589030112e-05,
+      "loss": 0.6031,
+      "step": 13011
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1411747279660021e-05,
+      "loss": 0.5982,
+      "step": 13012
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1410585950863047e-05,
+      "loss": 0.6387,
+      "step": 13013
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140942460265517e-05,
+      "loss": 0.638,
+      "step": 13014
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1408263235052374e-05,
+      "loss": 0.6627,
+      "step": 13015
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1407101848070638e-05,
+      "loss": 0.7127,
+      "step": 13016
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1405940441725947e-05,
+      "loss": 0.6573,
+      "step": 13017
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1404779016034275e-05,
+      "loss": 0.6061,
+      "step": 13018
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1403617571011614e-05,
+      "loss": 0.5298,
+      "step": 13019
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1402456106673938e-05,
+      "loss": 0.6153,
+      "step": 13020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1401294623037238e-05,
+      "loss": 0.6453,
+      "step": 13021
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1400133120117486e-05,
+      "loss": 0.6232,
+      "step": 13022
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1398971597930679e-05,
+      "loss": 0.6177,
+      "step": 13023
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1397810056492791e-05,
+      "loss": 0.6411,
+      "step": 13024
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1396648495819807e-05,
+      "loss": 0.6471,
+      "step": 13025
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1395486915927713e-05,
+      "loss": 0.5695,
+      "step": 13026
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1394325316832493e-05,
+      "loss": 0.6395,
+      "step": 13027
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1393163698550133e-05,
+      "loss": 0.6676,
+      "step": 13028
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1392002061096614e-05,
+      "loss": 0.5659,
+      "step": 13029
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1390840404487926e-05,
+      "loss": 0.665,
+      "step": 13030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.138967872874005e-05,
+      "loss": 0.6535,
+      "step": 13031
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1388517033868976e-05,
+      "loss": 0.6822,
+      "step": 13032
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1387355319890685e-05,
+      "loss": 0.585,
+      "step": 13033
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1386193586821165e-05,
+      "loss": 0.6256,
+      "step": 13034
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1385031834676405e-05,
+      "loss": 0.5366,
+      "step": 13035
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.138387006347239e-05,
+      "loss": 0.7269,
+      "step": 13036
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1382708273225108e-05,
+      "loss": 0.5335,
+      "step": 13037
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1381546463950542e-05,
+      "loss": 0.6332,
+      "step": 13038
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1380384635664688e-05,
+      "loss": 0.664,
+      "step": 13039
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1379222788383522e-05,
+      "loss": 0.6306,
+      "step": 13040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1378060922123044e-05,
+      "loss": 0.6454,
+      "step": 13041
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1376899036899234e-05,
+      "loss": 0.6183,
+      "step": 13042
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1375737132728084e-05,
+      "loss": 0.6398,
+      "step": 13043
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137457520962558e-05,
+      "loss": 0.6396,
+      "step": 13044
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1373413267607716e-05,
+      "loss": 0.5981,
+      "step": 13045
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1372251306690478e-05,
+      "loss": 0.577,
+      "step": 13046
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1371089326889857e-05,
+      "loss": 0.6695,
+      "step": 13047
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1369927328221842e-05,
+      "loss": 0.6054,
+      "step": 13048
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1368765310702422e-05,
+      "loss": 0.6075,
+      "step": 13049
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1367603274347591e-05,
+      "loss": 0.6362,
+      "step": 13050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1366441219173335e-05,
+      "loss": 0.6609,
+      "step": 13051
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1365279145195649e-05,
+      "loss": 0.5785,
+      "step": 13052
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.136411705243052e-05,
+      "loss": 0.6545,
+      "step": 13053
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1362954940893944e-05,
+      "loss": 0.6025,
+      "step": 13054
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1361792810601908e-05,
+      "loss": 0.6284,
+      "step": 13055
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1360630661570406e-05,
+      "loss": 0.6785,
+      "step": 13056
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1359468493815434e-05,
+      "loss": 0.5735,
+      "step": 13057
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1358306307352978e-05,
+      "loss": 0.5635,
+      "step": 13058
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1357144102199035e-05,
+      "loss": 0.6569,
+      "step": 13059
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1355981878369598e-05,
+      "loss": 0.6061,
+      "step": 13060
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1354819635880658e-05,
+      "loss": 0.6511,
+      "step": 13061
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1353657374748206e-05,
+      "loss": 0.6055,
+      "step": 13062
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1352495094988245e-05,
+      "loss": 0.618,
+      "step": 13063
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1351332796616759e-05,
+      "loss": 0.6074,
+      "step": 13064
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1350170479649747e-05,
+      "loss": 0.5856,
+      "step": 13065
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1349008144103202e-05,
+      "loss": 0.685,
+      "step": 13066
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134784578999312e-05,
+      "loss": 0.6462,
+      "step": 13067
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1346683417335497e-05,
+      "loss": 0.5044,
+      "step": 13068
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1345521026146327e-05,
+      "loss": 0.6668,
+      "step": 13069
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1344358616441605e-05,
+      "loss": 0.6024,
+      "step": 13070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1343196188237327e-05,
+      "loss": 0.6489,
+      "step": 13071
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134203374154949e-05,
+      "loss": 0.666,
+      "step": 13072
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1340871276394085e-05,
+      "loss": 0.5939,
+      "step": 13073
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1339708792787119e-05,
+      "loss": 0.638,
+      "step": 13074
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1338546290744578e-05,
+      "loss": 0.5822,
+      "step": 13075
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1337383770282466e-05,
+      "loss": 0.5781,
+      "step": 13076
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.133622123141678e-05,
+      "loss": 0.6066,
+      "step": 13077
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1335058674163512e-05,
+      "loss": 0.6201,
+      "step": 13078
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1333896098538669e-05,
+      "loss": 0.6047,
+      "step": 13079
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1332733504558239e-05,
+      "loss": 0.6126,
+      "step": 13080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1331570892238228e-05,
+      "loss": 0.5556,
+      "step": 13081
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1330408261594627e-05,
+      "loss": 0.6033,
+      "step": 13082
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1329245612643445e-05,
+      "loss": 0.6275,
+      "step": 13083
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.132808294540067e-05,
+      "loss": 0.6594,
+      "step": 13084
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1326920259882312e-05,
+      "loss": 0.6797,
+      "step": 13085
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1325757556104362e-05,
+      "loss": 0.6525,
+      "step": 13086
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1324594834082823e-05,
+      "loss": 0.5584,
+      "step": 13087
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1323432093833695e-05,
+      "loss": 0.6927,
+      "step": 13088
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.132226933537298e-05,
+      "loss": 0.6117,
+      "step": 13089
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1321106558716676e-05,
+      "loss": 0.6075,
+      "step": 13090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1319943763880786e-05,
+      "loss": 0.6099,
+      "step": 13091
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.131878095088131e-05,
+      "loss": 0.6593,
+      "step": 13092
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1317618119734249e-05,
+      "loss": 0.6506,
+      "step": 13093
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1316455270455604e-05,
+      "loss": 0.6762,
+      "step": 13094
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1315292403061378e-05,
+      "loss": 0.6466,
+      "step": 13095
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1314129517567576e-05,
+      "loss": 0.5792,
+      "step": 13096
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1312966613990192e-05,
+      "loss": 0.5484,
+      "step": 13097
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1311803692345236e-05,
+      "loss": 0.6551,
+      "step": 13098
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1310640752648709e-05,
+      "loss": 0.6005,
+      "step": 13099
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1309477794916612e-05,
+      "loss": 0.6095,
+      "step": 13100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1308314819164951e-05,
+      "loss": 0.6369,
+      "step": 13101
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130715182540973e-05,
+      "loss": 0.5656,
+      "step": 13102
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1305988813666951e-05,
+      "loss": 0.6347,
+      "step": 13103
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1304825783952615e-05,
+      "loss": 0.603,
+      "step": 13104
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1303662736282735e-05,
+      "loss": 0.625,
+      "step": 13105
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1302499670673306e-05,
+      "loss": 0.6567,
+      "step": 13106
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1301336587140337e-05,
+      "loss": 0.5985,
+      "step": 13107
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1300173485699834e-05,
+      "loss": 0.6078,
+      "step": 13108
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1299010366367801e-05,
+      "loss": 0.544,
+      "step": 13109
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1297847229160248e-05,
+      "loss": 0.6112,
+      "step": 13110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1296684074093171e-05,
+      "loss": 0.6692,
+      "step": 13111
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1295520901182586e-05,
+      "loss": 0.536,
+      "step": 13112
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1294357710444494e-05,
+      "loss": 0.6364,
+      "step": 13113
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1293194501894901e-05,
+      "loss": 0.6252,
+      "step": 13114
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1292031275549817e-05,
+      "loss": 0.653,
+      "step": 13115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1290868031425246e-05,
+      "loss": 0.5697,
+      "step": 13116
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.12897047695372e-05,
+      "loss": 0.6771,
+      "step": 13117
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1288541489901679e-05,
+      "loss": 0.6476,
+      "step": 13118
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1287378192534697e-05,
+      "loss": 0.6708,
+      "step": 13119
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1286214877452259e-05,
+      "loss": 0.645,
+      "step": 13120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1285051544670375e-05,
+      "loss": 0.6533,
+      "step": 13121
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.128388819420505e-05,
+      "loss": 0.6336,
+      "step": 13122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1282724826072297e-05,
+      "loss": 0.5484,
+      "step": 13123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1281561440288125e-05,
+      "loss": 0.667,
+      "step": 13124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1280398036868538e-05,
+      "loss": 0.6331,
+      "step": 13125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1279234615829553e-05,
+      "loss": 0.6374,
+      "step": 13126
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1278071177187174e-05,
+      "loss": 0.585,
+      "step": 13127
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1276907720957414e-05,
+      "loss": 0.6691,
+      "step": 13128
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1275744247156277e-05,
+      "loss": 0.5521,
+      "step": 13129
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1274580755799785e-05,
+      "loss": 0.6217,
+      "step": 13130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1273417246903935e-05,
+      "loss": 0.5941,
+      "step": 13131
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.127225372048475e-05,
+      "loss": 0.6096,
+      "step": 13132
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1271090176558233e-05,
+      "loss": 0.6288,
+      "step": 13133
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1269926615140397e-05,
+      "loss": 0.6003,
+      "step": 13134
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1268763036247259e-05,
+      "loss": 0.644,
+      "step": 13135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1267599439894822e-05,
+      "loss": 0.6344,
+      "step": 13136
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126643582609911e-05,
+      "loss": 0.5792,
+      "step": 13137
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126527219487612e-05,
+      "loss": 0.574,
+      "step": 13138
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126410854624188e-05,
+      "loss": 0.5786,
+      "step": 13139
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126294488021239e-05,
+      "loss": 0.7127,
+      "step": 13140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1261781196803674e-05,
+      "loss": 0.6473,
+      "step": 13141
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1260617496031733e-05,
+      "loss": 0.6241,
+      "step": 13142
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1259453777912595e-05,
+      "loss": 0.6118,
+      "step": 13143
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1258290042462261e-05,
+      "loss": 0.5522,
+      "step": 13144
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.125712628969675e-05,
+      "loss": 0.5859,
+      "step": 13145
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1255962519632082e-05,
+      "loss": 0.6884,
+      "step": 13146
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1254798732284263e-05,
+      "loss": 0.5967,
+      "step": 13147
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1253634927669313e-05,
+      "loss": 0.6239,
+      "step": 13148
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1252471105803241e-05,
+      "loss": 0.6655,
+      "step": 13149
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.125130726670207e-05,
+      "loss": 0.6369,
+      "step": 13150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1250143410381808e-05,
+      "loss": 0.6687,
+      "step": 13151
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1248979536858477e-05,
+      "loss": 0.6003,
+      "step": 13152
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1247815646148088e-05,
+      "loss": 0.6311,
+      "step": 13153
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1246651738266659e-05,
+      "loss": 0.5658,
+      "step": 13154
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1245487813230209e-05,
+      "loss": 0.6701,
+      "step": 13155
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.124432387105475e-05,
+      "loss": 0.6029,
+      "step": 13156
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1243159911756306e-05,
+      "loss": 0.5867,
+      "step": 13157
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1241995935350882e-05,
+      "loss": 0.5663,
+      "step": 13158
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1240831941854509e-05,
+      "loss": 0.6479,
+      "step": 13159
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1239667931283193e-05,
+      "loss": 0.5477,
+      "step": 13160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1238503903652962e-05,
+      "loss": 0.5899,
+      "step": 13161
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1237339858979825e-05,
+      "loss": 0.6094,
+      "step": 13162
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1236175797279807e-05,
+      "loss": 0.63,
+      "step": 13163
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1235011718568922e-05,
+      "loss": 0.6488,
+      "step": 13164
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.123384762286319e-05,
+      "loss": 0.6092,
+      "step": 13165
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1232683510178632e-05,
+      "loss": 0.6087,
+      "step": 13166
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1231519380531265e-05,
+      "loss": 0.6642,
+      "step": 13167
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1230355233937112e-05,
+      "loss": 0.6103,
+      "step": 13168
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1229191070412187e-05,
+      "loss": 0.6452,
+      "step": 13169
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1228026889972514e-05,
+      "loss": 0.591,
+      "step": 13170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.122686269263411e-05,
+      "loss": 0.6297,
+      "step": 13171
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1225698478413e-05,
+      "loss": 0.6429,
+      "step": 13172
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.12245342473252e-05,
+      "loss": 0.5827,
+      "step": 13173
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1223369999386736e-05,
+      "loss": 0.6433,
+      "step": 13174
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1222205734613621e-05,
+      "loss": 0.584,
+      "step": 13175
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.122104145302188e-05,
+      "loss": 0.5837,
+      "step": 13176
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1219877154627543e-05,
+      "loss": 0.6293,
+      "step": 13177
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121871283944662e-05,
+      "loss": 0.5655,
+      "step": 13178
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1217548507495137e-05,
+      "loss": 0.5583,
+      "step": 13179
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1216384158789117e-05,
+      "loss": 0.6243,
+      "step": 13180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1215219793344585e-05,
+      "loss": 0.6161,
+      "step": 13181
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1214055411177558e-05,
+      "loss": 0.6327,
+      "step": 13182
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1212891012304064e-05,
+      "loss": 0.6714,
+      "step": 13183
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1211726596740123e-05,
+      "loss": 0.6212,
+      "step": 13184
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1210562164501757e-05,
+      "loss": 0.644,
+      "step": 13185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1209397715604994e-05,
+      "loss": 0.6397,
+      "step": 13186
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1208233250065857e-05,
+      "loss": 0.6632,
+      "step": 13187
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1207068767900367e-05,
+      "loss": 0.6303,
+      "step": 13188
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.120590426912455e-05,
+      "loss": 0.6552,
+      "step": 13189
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1204739753754433e-05,
+      "loss": 0.5858,
+      "step": 13190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1203575221806035e-05,
+      "loss": 0.5705,
+      "step": 13191
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1202410673295387e-05,
+      "loss": 0.6181,
+      "step": 13192
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.120124610823851e-05,
+      "loss": 0.6319,
+      "step": 13193
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1200081526651432e-05,
+      "loss": 0.5556,
+      "step": 13194
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1198916928550176e-05,
+      "loss": 0.6483,
+      "step": 13195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1197752313950771e-05,
+      "loss": 0.6601,
+      "step": 13196
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1196587682869243e-05,
+      "loss": 0.5987,
+      "step": 13197
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1195423035321615e-05,
+      "loss": 0.6214,
+      "step": 13198
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1194258371323917e-05,
+      "loss": 0.6727,
+      "step": 13199
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1193093690892173e-05,
+      "loss": 0.6754,
+      "step": 13200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1191928994042414e-05,
+      "loss": 0.6455,
+      "step": 13201
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1190764280790663e-05,
+      "loss": 0.6364,
+      "step": 13202
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.118959955115295e-05,
+      "loss": 0.6774,
+      "step": 13203
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1188434805145299e-05,
+      "loss": 0.6034,
+      "step": 13204
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1187270042783744e-05,
+      "loss": 0.6876,
+      "step": 13205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1186105264084309e-05,
+      "loss": 0.6478,
+      "step": 13206
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1184940469063022e-05,
+      "loss": 0.6593,
+      "step": 13207
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1183775657735916e-05,
+      "loss": 0.5953,
+      "step": 13208
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1182610830119013e-05,
+      "loss": 0.6645,
+      "step": 13209
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1181445986228349e-05,
+      "loss": 0.558,
+      "step": 13210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1180281126079948e-05,
+      "loss": 0.5884,
+      "step": 13211
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1179116249689846e-05,
+      "loss": 0.5743,
+      "step": 13212
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1177951357074062e-05,
+      "loss": 0.5971,
+      "step": 13213
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1176786448248637e-05,
+      "loss": 0.6484,
+      "step": 13214
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1175621523229595e-05,
+      "loss": 0.5733,
+      "step": 13215
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1174456582032965e-05,
+      "loss": 0.5725,
+      "step": 13216
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1173291624674785e-05,
+      "loss": 0.6321,
+      "step": 13217
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117212665117108e-05,
+      "loss": 0.5792,
+      "step": 13218
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1170961661537882e-05,
+      "loss": 0.6852,
+      "step": 13219
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1169796655791222e-05,
+      "loss": 0.6128,
+      "step": 13220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1168631633947134e-05,
+      "loss": 0.5709,
+      "step": 13221
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1167466596021645e-05,
+      "loss": 0.5752,
+      "step": 13222
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.116630154203079e-05,
+      "loss": 0.5363,
+      "step": 13223
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1165136471990603e-05,
+      "loss": 0.6138,
+      "step": 13224
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1163971385917114e-05,
+      "loss": 0.6062,
+      "step": 13225
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1162806283826355e-05,
+      "loss": 0.5632,
+      "step": 13226
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.116164116573436e-05,
+      "loss": 0.7111,
+      "step": 13227
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1160476031657164e-05,
+      "loss": 0.6241,
+      "step": 13228
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1159310881610795e-05,
+      "loss": 0.6015,
+      "step": 13229
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1158145715611293e-05,
+      "loss": 0.649,
+      "step": 13230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1156980533674684e-05,
+      "loss": 0.6327,
+      "step": 13231
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.115581533581701e-05,
+      "loss": 0.6364,
+      "step": 13232
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.11546501220543e-05,
+      "loss": 0.5812,
+      "step": 13233
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1153484892402589e-05,
+      "loss": 0.5951,
+      "step": 13234
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1152319646877913e-05,
+      "loss": 0.5703,
+      "step": 13235
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1151154385496305e-05,
+      "loss": 0.6445,
+      "step": 13236
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1149989108273804e-05,
+      "loss": 0.6112,
+      "step": 13237
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.114882381522644e-05,
+      "loss": 0.6796,
+      "step": 13238
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1147658506370252e-05,
+      "loss": 0.6123,
+      "step": 13239
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1146493181721271e-05,
+      "loss": 0.6069,
+      "step": 13240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1145327841295543e-05,
+      "loss": 0.5493,
+      "step": 13241
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1144162485109093e-05,
+      "loss": 0.6188,
+      "step": 13242
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1142997113177962e-05,
+      "loss": 0.5956,
+      "step": 13243
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1141831725518189e-05,
+      "loss": 0.6446,
+      "step": 13244
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1140666322145804e-05,
+      "loss": 0.6256,
+      "step": 13245
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1139500903076852e-05,
+      "loss": 0.654,
+      "step": 13246
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1138335468327363e-05,
+      "loss": 0.6004,
+      "step": 13247
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1137170017913382e-05,
+      "loss": 0.7143,
+      "step": 13248
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.113600455185094e-05,
+      "loss": 0.6144,
+      "step": 13249
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1134839070156075e-05,
+      "loss": 0.6379,
+      "step": 13250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1133673572844829e-05,
+      "loss": 0.6485,
+      "step": 13251
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.113250805993324e-05,
+      "loss": 0.5934,
+      "step": 13252
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1131342531437344e-05,
+      "loss": 0.6863,
+      "step": 13253
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1130176987373181e-05,
+      "loss": 0.5786,
+      "step": 13254
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1129011427756788e-05,
+      "loss": 0.644,
+      "step": 13255
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1127845852604207e-05,
+      "loss": 0.5793,
+      "step": 13256
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1126680261931478e-05,
+      "loss": 0.5837,
+      "step": 13257
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1125514655754634e-05,
+      "loss": 0.6067,
+      "step": 13258
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1124349034089724e-05,
+      "loss": 0.6148,
+      "step": 13259
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1123183396952779e-05,
+      "loss": 0.5718,
+      "step": 13260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1122017744359848e-05,
+      "loss": 0.6229,
+      "step": 13261
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1120852076326965e-05,
+      "loss": 0.5754,
+      "step": 13262
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1119686392870172e-05,
+      "loss": 0.5914,
+      "step": 13263
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.111852069400551e-05,
+      "loss": 0.6188,
+      "step": 13264
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.111735497974902e-05,
+      "loss": 0.6123,
+      "step": 13265
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1116189250116744e-05,
+      "loss": 0.5633,
+      "step": 13266
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1115023505124724e-05,
+      "loss": 0.5861,
+      "step": 13267
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1113857744789002e-05,
+      "loss": 0.6082,
+      "step": 13268
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1112691969125618e-05,
+      "loss": 0.6442,
+      "step": 13269
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1111526178150615e-05,
+      "loss": 0.6377,
+      "step": 13270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1110360371880034e-05,
+      "loss": 0.6523,
+      "step": 13271
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1109194550329919e-05,
+      "loss": 0.6073,
+      "step": 13272
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1108028713516312e-05,
+      "loss": 0.5976,
+      "step": 13273
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1106862861455255e-05,
+      "loss": 0.6151,
+      "step": 13274
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1105696994162796e-05,
+      "loss": 0.6527,
+      "step": 13275
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1104531111654973e-05,
+      "loss": 0.6058,
+      "step": 13276
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1103365213947833e-05,
+      "loss": 0.6144,
+      "step": 13277
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1102199301057413e-05,
+      "loss": 0.6276,
+      "step": 13278
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1101033372999767e-05,
+      "loss": 0.6223,
+      "step": 13279
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1099867429790931e-05,
+      "loss": 0.5656,
+      "step": 13280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1098701471446958e-05,
+      "loss": 0.6579,
+      "step": 13281
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1097535497983883e-05,
+      "loss": 0.6411,
+      "step": 13282
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1096369509417757e-05,
+      "loss": 0.5921,
+      "step": 13283
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.109520350576462e-05,
+      "loss": 0.6544,
+      "step": 13284
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1094037487040523e-05,
+      "loss": 0.6262,
+      "step": 13285
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1092871453261509e-05,
+      "loss": 0.5994,
+      "step": 13286
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1091705404443622e-05,
+      "loss": 0.6096,
+      "step": 13287
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1090539340602912e-05,
+      "loss": 0.6372,
+      "step": 13288
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.108937326175542e-05,
+      "loss": 0.5736,
+      "step": 13289
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1088207167917198e-05,
+      "loss": 0.6317,
+      "step": 13290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1087041059104287e-05,
+      "loss": 0.6509,
+      "step": 13291
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1085874935332736e-05,
+      "loss": 0.6281,
+      "step": 13292
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.108470879661859e-05,
+      "loss": 0.6249,
+      "step": 13293
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1083542642977902e-05,
+      "loss": 0.5862,
+      "step": 13294
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.108237647442671e-05,
+      "loss": 0.6171,
+      "step": 13295
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1081210290981068e-05,
+      "loss": 0.6632,
+      "step": 13296
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1080044092657025e-05,
+      "loss": 0.5749,
+      "step": 13297
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1078877879470625e-05,
+      "loss": 0.7291,
+      "step": 13298
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107771165143792e-05,
+      "loss": 0.6321,
+      "step": 13299
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1076545408574951e-05,
+      "loss": 0.611,
+      "step": 13300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1075379150897774e-05,
+      "loss": 0.6238,
+      "step": 13301
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1074212878422432e-05,
+      "loss": 0.6605,
+      "step": 13302
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1073046591164982e-05,
+      "loss": 0.6051,
+      "step": 13303
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1071880289141463e-05,
+      "loss": 0.6052,
+      "step": 13304
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107071397236793e-05,
+      "loss": 0.6523,
+      "step": 13305
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1069547640860437e-05,
+      "loss": 0.6818,
+      "step": 13306
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1068381294635022e-05,
+      "loss": 0.5638,
+      "step": 13307
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1067214933707746e-05,
+      "loss": 0.568,
+      "step": 13308
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1066048558094652e-05,
+      "loss": 0.6129,
+      "step": 13309
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1064882167811796e-05,
+      "loss": 0.6877,
+      "step": 13310
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1063715762875225e-05,
+      "loss": 0.6694,
+      "step": 13311
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106254934330099e-05,
+      "loss": 0.6148,
+      "step": 13312
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106138290910514e-05,
+      "loss": 0.6532,
+      "step": 13313
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1060216460303735e-05,
+      "loss": 0.5749,
+      "step": 13314
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1059049996912814e-05,
+      "loss": 0.6177,
+      "step": 13315
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1057883518948433e-05,
+      "loss": 0.6493,
+      "step": 13316
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.105671702642665e-05,
+      "loss": 0.6218,
+      "step": 13317
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1055550519363511e-05,
+      "loss": 0.5912,
+      "step": 13318
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1054383997775068e-05,
+      "loss": 0.6005,
+      "step": 13319
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1053217461677377e-05,
+      "loss": 0.6664,
+      "step": 13320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1052050911086487e-05,
+      "loss": 0.6669,
+      "step": 13321
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1050884346018452e-05,
+      "loss": 0.595,
+      "step": 13322
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1049717766489325e-05,
+      "loss": 0.6275,
+      "step": 13323
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.104855117251516e-05,
+      "loss": 0.5706,
+      "step": 13324
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.104738456411201e-05,
+      "loss": 0.6095,
+      "step": 13325
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1046217941295926e-05,
+      "loss": 0.6325,
+      "step": 13326
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1045051304082966e-05,
+      "loss": 0.6608,
+      "step": 13327
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.104388465248918e-05,
+      "loss": 0.5968,
+      "step": 13328
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1042717986530624e-05,
+      "loss": 0.673,
+      "step": 13329
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1041551306223353e-05,
+      "loss": 0.5822,
+      "step": 13330
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1040384611583419e-05,
+      "loss": 0.5963,
+      "step": 13331
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103921790262688e-05,
+      "loss": 0.6074,
+      "step": 13332
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103805117936979e-05,
+      "loss": 0.6511,
+      "step": 13333
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1036884441828203e-05,
+      "loss": 0.6832,
+      "step": 13334
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1035717690018177e-05,
+      "loss": 0.6409,
+      "step": 13335
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1034550923955762e-05,
+      "loss": 0.576,
+      "step": 13336
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103338414365702e-05,
+      "loss": 0.6392,
+      "step": 13337
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1032217349138003e-05,
+      "loss": 0.6151,
+      "step": 13338
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1031050540414769e-05,
+      "loss": 0.6382,
+      "step": 13339
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1029883717503374e-05,
+      "loss": 0.5146,
+      "step": 13340
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1028716880419874e-05,
+      "loss": 0.5878,
+      "step": 13341
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1027550029180323e-05,
+      "loss": 0.5626,
+      "step": 13342
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1026383163800782e-05,
+      "loss": 0.6605,
+      "step": 13343
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.102521628429731e-05,
+      "loss": 0.6002,
+      "step": 13344
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1024049390685957e-05,
+      "loss": 0.5752,
+      "step": 13345
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1022882482982786e-05,
+      "loss": 0.6334,
+      "step": 13346
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1021715561203854e-05,
+      "loss": 0.5319,
+      "step": 13347
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.102054862536522e-05,
+      "loss": 0.6334,
+      "step": 13348
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1019381675482937e-05,
+      "loss": 0.6399,
+      "step": 13349
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1018214711573067e-05,
+      "loss": 0.639,
+      "step": 13350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1017047733651666e-05,
+      "loss": 0.6422,
+      "step": 13351
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1015880741734796e-05,
+      "loss": 0.6654,
+      "step": 13352
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1014713735838516e-05,
+      "loss": 0.5808,
+      "step": 13353
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1013546715978883e-05,
+      "loss": 0.6651,
+      "step": 13354
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1012379682171956e-05,
+      "loss": 0.5882,
+      "step": 13355
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1011212634433795e-05,
+      "loss": 0.699,
+      "step": 13356
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1010045572780461e-05,
+      "loss": 0.6476,
+      "step": 13357
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1008878497228009e-05,
+      "loss": 0.6177,
+      "step": 13358
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1007711407792504e-05,
+      "loss": 0.5934,
+      "step": 13359
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1006544304490005e-05,
+      "loss": 0.5594,
+      "step": 13360
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1005377187336575e-05,
+      "loss": 0.6244,
+      "step": 13361
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1004210056348265e-05,
+      "loss": 0.6245,
+      "step": 13362
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1003042911541146e-05,
+      "loss": 0.6919,
+      "step": 13363
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1001875752931277e-05,
+      "loss": 0.5667,
+      "step": 13364
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1000708580534713e-05,
+      "loss": 0.6876,
+      "step": 13365
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0999541394367523e-05,
+      "loss": 0.5887,
+      "step": 13366
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0998374194445764e-05,
+      "loss": 0.6023,
+      "step": 13367
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.09972069807855e-05,
+      "loss": 0.5324,
+      "step": 13368
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.099603975340279e-05,
+      "loss": 0.5565,
+      "step": 13369
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.09948725123137e-05,
+      "loss": 0.5951,
+      "step": 13370
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0993705257534286e-05,
+      "loss": 0.635,
+      "step": 13371
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0992537989080618e-05,
+      "loss": 0.606,
+      "step": 13372
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0991370706968754e-05,
+      "loss": 0.5562,
+      "step": 13373
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0990203411214757e-05,
+      "loss": 0.642,
+      "step": 13374
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0989036101834693e-05,
+      "loss": 0.6546,
+      "step": 13375
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0987868778844621e-05,
+      "loss": 0.6662,
+      "step": 13376
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.098670144226061e-05,
+      "loss": 0.5551,
+      "step": 13377
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0985534092098715e-05,
+      "loss": 0.584,
+      "step": 13378
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0984366728375012e-05,
+      "loss": 0.6215,
+      "step": 13379
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.098319935110555e-05,
+      "loss": 0.6268,
+      "step": 13380
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0982031960306408e-05,
+      "loss": 0.6263,
+      "step": 13381
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0980864555993638e-05,
+      "loss": 0.5286,
+      "step": 13382
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0979697138183311e-05,
+      "loss": 0.5569,
+      "step": 13383
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0978529706891493e-05,
+      "loss": 0.6548,
+      "step": 13384
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0977362262134243e-05,
+      "loss": 0.6017,
+      "step": 13385
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0976194803927634e-05,
+      "loss": 0.57,
+      "step": 13386
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0975027332287721e-05,
+      "loss": 0.5799,
+      "step": 13387
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0973859847230582e-05,
+      "loss": 0.59,
+      "step": 13388
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0972692348772271e-05,
+      "loss": 0.6782,
+      "step": 13389
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.097152483692886e-05,
+      "loss": 0.6256,
+      "step": 13390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0970357311716414e-05,
+      "loss": 0.5921,
+      "step": 13391
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0969189773150998e-05,
+      "loss": 0.6699,
+      "step": 13392
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0968022221248678e-05,
+      "loss": 0.6911,
+      "step": 13393
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0966854656025523e-05,
+      "loss": 0.6417,
+      "step": 13394
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0965687077497597e-05,
+      "loss": 0.6246,
+      "step": 13395
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.096451948568097e-05,
+      "loss": 0.6204,
+      "step": 13396
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0963351880591708e-05,
+      "loss": 0.6574,
+      "step": 13397
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0962184262245876e-05,
+      "loss": 0.5946,
+      "step": 13398
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0961016630659544e-05,
+      "loss": 0.643,
+      "step": 13399
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0959848985848778e-05,
+      "loss": 0.6052,
+      "step": 13400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0958681327829647e-05,
+      "loss": 0.6171,
+      "step": 13401
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.095751365661822e-05,
+      "loss": 0.6257,
+      "step": 13402
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0956345972230561e-05,
+      "loss": 0.6108,
+      "step": 13403
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0955178274682744e-05,
+      "loss": 0.6637,
+      "step": 13404
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0954010563990831e-05,
+      "loss": 0.5711,
+      "step": 13405
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0952842840170896e-05,
+      "loss": 0.5521,
+      "step": 13406
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0951675103239009e-05,
+      "loss": 0.6147,
+      "step": 13407
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0950507353211233e-05,
+      "loss": 0.634,
+      "step": 13408
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.094933959010364e-05,
+      "loss": 0.5974,
+      "step": 13409
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0948171813932304e-05,
+      "loss": 0.5734,
+      "step": 13410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0947004024713289e-05,
+      "loss": 0.5079,
+      "step": 13411
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0945836222462664e-05,
+      "loss": 0.6443,
+      "step": 13412
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0944668407196503e-05,
+      "loss": 0.6159,
+      "step": 13413
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0943500578930874e-05,
+      "loss": 0.6567,
+      "step": 13414
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.094233273768185e-05,
+      "loss": 0.6209,
+      "step": 13415
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0941164883465499e-05,
+      "loss": 0.666,
+      "step": 13416
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0939997016297892e-05,
+      "loss": 0.6241,
+      "step": 13417
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0938829136195099e-05,
+      "loss": 0.6146,
+      "step": 13418
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0937661243173195e-05,
+      "loss": 0.643,
+      "step": 13419
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0936493337248247e-05,
+      "loss": 0.6053,
+      "step": 13420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0935325418436326e-05,
+      "loss": 0.6555,
+      "step": 13421
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0934157486753507e-05,
+      "loss": 0.5735,
+      "step": 13422
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0932989542215861e-05,
+      "loss": 0.6651,
+      "step": 13423
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0931821584839457e-05,
+      "loss": 0.6665,
+      "step": 13424
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.093065361464037e-05,
+      "loss": 0.5641,
+      "step": 13425
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0929485631634673e-05,
+      "loss": 0.6505,
+      "step": 13426
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0928317635838434e-05,
+      "loss": 0.6112,
+      "step": 13427
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092714962726773e-05,
+      "loss": 0.6691,
+      "step": 13428
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0925981605938635e-05,
+      "loss": 0.6431,
+      "step": 13429
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0924813571867216e-05,
+      "loss": 0.647,
+      "step": 13430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0923645525069552e-05,
+      "loss": 0.625,
+      "step": 13431
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0922477465561712e-05,
+      "loss": 0.5283,
+      "step": 13432
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0921309393359774e-05,
+      "loss": 0.6054,
+      "step": 13433
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0920141308479805e-05,
+      "loss": 0.559,
+      "step": 13434
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0918973210937888e-05,
+      "loss": 0.6475,
+      "step": 13435
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.091780510075009e-05,
+      "loss": 0.5699,
+      "step": 13436
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0916636977932489e-05,
+      "loss": 0.5798,
+      "step": 13437
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0915468842501153e-05,
+      "loss": 0.5578,
+      "step": 13438
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0914300694472166e-05,
+      "loss": 0.5435,
+      "step": 13439
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0913132533861597e-05,
+      "loss": 0.6657,
+      "step": 13440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.091196436068552e-05,
+      "loss": 0.6368,
+      "step": 13441
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0910796174960015e-05,
+      "loss": 0.6102,
+      "step": 13442
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0909627976701152e-05,
+      "loss": 0.6191,
+      "step": 13443
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.090845976592501e-05,
+      "loss": 0.5722,
+      "step": 13444
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.090729154264766e-05,
+      "loss": 0.6049,
+      "step": 13445
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0906123306885185e-05,
+      "loss": 0.5668,
+      "step": 13446
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0904955058653657e-05,
+      "loss": 0.5783,
+      "step": 13447
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0903786797969152e-05,
+      "loss": 0.641,
+      "step": 13448
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0902618524847746e-05,
+      "loss": 0.4754,
+      "step": 13449
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0901450239305516e-05,
+      "loss": 0.5781,
+      "step": 13450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0900281941358537e-05,
+      "loss": 0.6159,
+      "step": 13451
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0899113631022889e-05,
+      "loss": 0.6382,
+      "step": 13452
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0897945308314647e-05,
+      "loss": 0.5984,
+      "step": 13453
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0896776973249889e-05,
+      "loss": 0.644,
+      "step": 13454
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0895608625844694e-05,
+      "loss": 0.6402,
+      "step": 13455
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0894440266115134e-05,
+      "loss": 0.5675,
+      "step": 13456
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0893271894077291e-05,
+      "loss": 0.6503,
+      "step": 13457
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.089210350974724e-05,
+      "loss": 0.6376,
+      "step": 13458
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0890935113141066e-05,
+      "loss": 0.5945,
+      "step": 13459
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0889766704274838e-05,
+      "loss": 0.6308,
+      "step": 13460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0888598283164638e-05,
+      "loss": 0.6222,
+      "step": 13461
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0887429849826548e-05,
+      "loss": 0.649,
+      "step": 13462
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088626140427664e-05,
+      "loss": 0.6955,
+      "step": 13463
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0885092946530997e-05,
+      "loss": 0.6037,
+      "step": 13464
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0883924476605698e-05,
+      "loss": 0.6418,
+      "step": 13465
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0882755994516822e-05,
+      "loss": 0.6913,
+      "step": 13466
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0881587500280448e-05,
+      "loss": 0.6332,
+      "step": 13467
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0880418993912656e-05,
+      "loss": 0.5961,
+      "step": 13468
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0879250475429523e-05,
+      "loss": 0.7316,
+      "step": 13469
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0878081944847133e-05,
+      "loss": 0.5311,
+      "step": 13470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0876913402181562e-05,
+      "loss": 0.6227,
+      "step": 13471
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0875744847448893e-05,
+      "loss": 0.6211,
+      "step": 13472
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0874576280665206e-05,
+      "loss": 0.623,
+      "step": 13473
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087340770184658e-05,
+      "loss": 0.5949,
+      "step": 13474
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0872239111009095e-05,
+      "loss": 0.6224,
+      "step": 13475
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0871070508168836e-05,
+      "loss": 0.5777,
+      "step": 13476
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.086990189334188e-05,
+      "loss": 0.6449,
+      "step": 13477
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.086873326654431e-05,
+      "loss": 0.6987,
+      "step": 13478
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0867564627792208e-05,
+      "loss": 0.678,
+      "step": 13479
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0866395977101654e-05,
+      "loss": 0.6194,
+      "step": 13480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0865227314488728e-05,
+      "loss": 0.6533,
+      "step": 13481
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0864058639969514e-05,
+      "loss": 0.5748,
+      "step": 13482
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0862889953560096e-05,
+      "loss": 0.6695,
+      "step": 13483
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0861721255276554e-05,
+      "loss": 0.5981,
+      "step": 13484
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.086055254513497e-05,
+      "loss": 0.6521,
+      "step": 13485
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0859383823151426e-05,
+      "loss": 0.625,
+      "step": 13486
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0858215089342006e-05,
+      "loss": 0.5994,
+      "step": 13487
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0857046343722792e-05,
+      "loss": 0.5418,
+      "step": 13488
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0855877586309868e-05,
+      "loss": 0.5433,
+      "step": 13489
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0854708817119318e-05,
+      "loss": 0.6388,
+      "step": 13490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.085354003616722e-05,
+      "loss": 0.6266,
+      "step": 13491
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0852371243469662e-05,
+      "loss": 0.6246,
+      "step": 13492
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0851202439042727e-05,
+      "loss": 0.6292,
+      "step": 13493
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0850033622902498e-05,
+      "loss": 0.598,
+      "step": 13494
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0848864795065062e-05,
+      "loss": 0.6509,
+      "step": 13495
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0847695955546498e-05,
+      "loss": 0.6238,
+      "step": 13496
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0846527104362896e-05,
+      "loss": 0.6012,
+      "step": 13497
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0845358241530333e-05,
+      "loss": 0.5846,
+      "step": 13498
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0844189367064901e-05,
+      "loss": 0.5796,
+      "step": 13499
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.084302048098268e-05,
+      "loss": 0.6382,
+      "step": 13500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0841851583299755e-05,
+      "loss": 0.5629,
+      "step": 13501
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0840682674032214e-05,
+      "loss": 0.592,
+      "step": 13502
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0839513753196141e-05,
+      "loss": 0.7035,
+      "step": 13503
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0838344820807623e-05,
+      "loss": 0.5544,
+      "step": 13504
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0837175876882739e-05,
+      "loss": 0.6576,
+      "step": 13505
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0836006921437586e-05,
+      "loss": 0.656,
+      "step": 13506
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0834837954488236e-05,
+      "loss": 0.6991,
+      "step": 13507
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0833668976050787e-05,
+      "loss": 0.5783,
+      "step": 13508
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0832499986141318e-05,
+      "loss": 0.6688,
+      "step": 13509
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0831330984775918e-05,
+      "loss": 0.6022,
+      "step": 13510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0830161971970675e-05,
+      "loss": 0.6412,
+      "step": 13511
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0828992947741672e-05,
+      "loss": 0.5382,
+      "step": 13512
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0827823912104998e-05,
+      "loss": 0.5854,
+      "step": 13513
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0826654865076738e-05,
+      "loss": 0.5827,
+      "step": 13514
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0825485806672983e-05,
+      "loss": 0.5746,
+      "step": 13515
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0824316736909817e-05,
+      "loss": 0.6237,
+      "step": 13516
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.082314765580333e-05,
+      "loss": 0.5965,
+      "step": 13517
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0821978563369605e-05,
+      "loss": 0.6671,
+      "step": 13518
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0820809459624735e-05,
+      "loss": 0.5703,
+      "step": 13519
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0819640344584805e-05,
+      "loss": 0.616,
+      "step": 13520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0818471218265905e-05,
+      "loss": 0.7179,
+      "step": 13521
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0817302080684118e-05,
+      "loss": 0.5907,
+      "step": 13522
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0816132931855539e-05,
+      "loss": 0.5956,
+      "step": 13523
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0814963771796255e-05,
+      "loss": 0.6585,
+      "step": 13524
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.081379460052235e-05,
+      "loss": 0.5926,
+      "step": 13525
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0812625418049918e-05,
+      "loss": 0.6319,
+      "step": 13526
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0811456224395048e-05,
+      "loss": 0.5918,
+      "step": 13527
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0810287019573823e-05,
+      "loss": 0.6252,
+      "step": 13528
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080911780360234e-05,
+      "loss": 0.6504,
+      "step": 13529
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0807948576496685e-05,
+      "loss": 0.643,
+      "step": 13530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0806779338272945e-05,
+      "loss": 0.6302,
+      "step": 13531
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0805610088947219e-05,
+      "loss": 0.5485,
+      "step": 13532
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0804440828535583e-05,
+      "loss": 0.6357,
+      "step": 13533
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0803271557054136e-05,
+      "loss": 0.6417,
+      "step": 13534
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080210227451897e-05,
+      "loss": 0.5289,
+      "step": 13535
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080093298094617e-05,
+      "loss": 0.6045,
+      "step": 13536
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0799763676351829e-05,
+      "loss": 0.677,
+      "step": 13537
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0798594360752034e-05,
+      "loss": 0.5984,
+      "step": 13538
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0797425034162881e-05,
+      "loss": 0.651,
+      "step": 13539
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0796255696600459e-05,
+      "loss": 0.5789,
+      "step": 13540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0795086348080859e-05,
+      "loss": 0.6386,
+      "step": 13541
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0793916988620172e-05,
+      "loss": 0.5602,
+      "step": 13542
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0792747618234489e-05,
+      "loss": 0.6817,
+      "step": 13543
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0791578236939903e-05,
+      "loss": 0.643,
+      "step": 13544
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0790408844752505e-05,
+      "loss": 0.5577,
+      "step": 13545
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0789239441688388e-05,
+      "loss": 0.5336,
+      "step": 13546
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.078807002776364e-05,
+      "loss": 0.6321,
+      "step": 13547
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.078690060299436e-05,
+      "loss": 0.5965,
+      "step": 13548
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0785731167396632e-05,
+      "loss": 0.5964,
+      "step": 13549
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0784561720986552e-05,
+      "loss": 0.6436,
+      "step": 13550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0783392263780218e-05,
+      "loss": 0.6439,
+      "step": 13551
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0782222795793714e-05,
+      "loss": 0.5608,
+      "step": 13552
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.078105331704314e-05,
+      "loss": 0.6424,
+      "step": 13553
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0779883827544584e-05,
+      "loss": 0.56,
+      "step": 13554
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0778714327314143e-05,
+      "loss": 0.6555,
+      "step": 13555
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0777544816367907e-05,
+      "loss": 0.6363,
+      "step": 13556
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0776375294721972e-05,
+      "loss": 0.5806,
+      "step": 13557
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.077520576239243e-05,
+      "loss": 0.6643,
+      "step": 13558
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0774036219395376e-05,
+      "loss": 0.5967,
+      "step": 13559
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0772866665746904e-05,
+      "loss": 0.5578,
+      "step": 13560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0771697101463106e-05,
+      "loss": 0.5712,
+      "step": 13561
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.077052752656008e-05,
+      "loss": 0.544,
+      "step": 13562
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0769357941053916e-05,
+      "loss": 0.6675,
+      "step": 13563
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0768188344960712e-05,
+      "loss": 0.6569,
+      "step": 13564
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0767018738296561e-05,
+      "loss": 0.6057,
+      "step": 13565
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0765849121077561e-05,
+      "loss": 0.5996,
+      "step": 13566
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0764679493319801e-05,
+      "loss": 0.621,
+      "step": 13567
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.076350985503938e-05,
+      "loss": 0.5765,
+      "step": 13568
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.076234020625239e-05,
+      "loss": 0.6528,
+      "step": 13569
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.076117054697493e-05,
+      "loss": 0.6139,
+      "step": 13570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0760000877223097e-05,
+      "loss": 0.6456,
+      "step": 13571
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0758831197012985e-05,
+      "loss": 0.5492,
+      "step": 13572
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0757661506360686e-05,
+      "loss": 0.565,
+      "step": 13573
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0756491805282298e-05,
+      "loss": 0.6312,
+      "step": 13574
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0755322093793924e-05,
+      "loss": 0.6088,
+      "step": 13575
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0754152371911648e-05,
+      "loss": 0.5576,
+      "step": 13576
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0752982639651575e-05,
+      "loss": 0.6477,
+      "step": 13577
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.07518128970298e-05,
+      "loss": 0.5764,
+      "step": 13578
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0750643144062418e-05,
+      "loss": 0.5673,
+      "step": 13579
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0749473380765525e-05,
+      "loss": 0.6331,
+      "step": 13580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0748303607155222e-05,
+      "loss": 0.5868,
+      "step": 13581
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0747133823247604e-05,
+      "loss": 0.6297,
+      "step": 13582
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0745964029058766e-05,
+      "loss": 0.5858,
+      "step": 13583
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.074479422460481e-05,
+      "loss": 0.6565,
+      "step": 13584
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0743624409901829e-05,
+      "loss": 0.6174,
+      "step": 13585
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0742454584965924e-05,
+      "loss": 0.5952,
+      "step": 13586
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0741284749813192e-05,
+      "loss": 0.6347,
+      "step": 13587
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.074011490445973e-05,
+      "loss": 0.6462,
+      "step": 13588
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0738945048921636e-05,
+      "loss": 0.6054,
+      "step": 13589
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0737775183215011e-05,
+      "loss": 0.6012,
+      "step": 13590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0736605307355948e-05,
+      "loss": 0.6842,
+      "step": 13591
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0735435421360549e-05,
+      "loss": 0.5759,
+      "step": 13592
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0734265525244916e-05,
+      "loss": 0.6362,
+      "step": 13593
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0733095619025141e-05,
+      "loss": 0.6323,
+      "step": 13594
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0731925702717327e-05,
+      "loss": 0.5518,
+      "step": 13595
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0730755776337571e-05,
+      "loss": 0.5974,
+      "step": 13596
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0729585839901975e-05,
+      "loss": 0.6377,
+      "step": 13597
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0728415893426636e-05,
+      "loss": 0.561,
+      "step": 13598
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0727245936927655e-05,
+      "loss": 0.6747,
+      "step": 13599
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072607597042113e-05,
+      "loss": 0.6651,
+      "step": 13600
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072490599392316e-05,
+      "loss": 0.7011,
+      "step": 13601
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072373600744985e-05,
+      "loss": 0.6523,
+      "step": 13602
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0722566011017295e-05,
+      "loss": 0.5504,
+      "step": 13603
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0721396004641598e-05,
+      "loss": 0.6082,
+      "step": 13604
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0720225988338857e-05,
+      "loss": 0.579,
+      "step": 13605
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0719055962125173e-05,
+      "loss": 0.5565,
+      "step": 13606
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0717885926016647e-05,
+      "loss": 0.5883,
+      "step": 13607
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0716715880029383e-05,
+      "loss": 0.6719,
+      "step": 13608
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0715545824179474e-05,
+      "loss": 0.6048,
+      "step": 13609
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.071437575848303e-05,
+      "loss": 0.5985,
+      "step": 13610
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0713205682956145e-05,
+      "loss": 0.581,
+      "step": 13611
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0712035597614923e-05,
+      "loss": 0.6071,
+      "step": 13612
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0710865502475469e-05,
+      "loss": 0.6845,
+      "step": 13613
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0709695397553877e-05,
+      "loss": 0.6038,
+      "step": 13614
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0708525282866256e-05,
+      "loss": 0.6159,
+      "step": 13615
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0707355158428702e-05,
+      "loss": 0.5893,
+      "step": 13616
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0706185024257321e-05,
+      "loss": 0.631,
+      "step": 13617
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0705014880368212e-05,
+      "loss": 0.6634,
+      "step": 13618
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0703844726777482e-05,
+      "loss": 0.5668,
+      "step": 13619
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0702674563501227e-05,
+      "loss": 0.6545,
+      "step": 13620
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0701504390555555e-05,
+      "loss": 0.6799,
+      "step": 13621
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0700334207956563e-05,
+      "loss": 0.6506,
+      "step": 13622
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0699164015720358e-05,
+      "loss": 0.6732,
+      "step": 13623
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0697993813863043e-05,
+      "loss": 0.6126,
+      "step": 13624
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0696823602400718e-05,
+      "loss": 0.6623,
+      "step": 13625
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0695653381349491e-05,
+      "loss": 0.62,
+      "step": 13626
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0694483150725458e-05,
+      "loss": 0.6116,
+      "step": 13627
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.069331291054473e-05,
+      "loss": 0.5705,
+      "step": 13628
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0692142660823406e-05,
+      "loss": 0.5645,
+      "step": 13629
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0690972401577593e-05,
+      "loss": 0.5807,
+      "step": 13630
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.068980213282339e-05,
+      "loss": 0.6401,
+      "step": 13631
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0688631854576903e-05,
+      "loss": 0.6697,
+      "step": 13632
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0687461566854237e-05,
+      "loss": 0.5563,
+      "step": 13633
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0686291269671496e-05,
+      "loss": 0.6608,
+      "step": 13634
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0685120963044785e-05,
+      "loss": 0.5941,
+      "step": 13635
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0683950646990209e-05,
+      "loss": 0.5816,
+      "step": 13636
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0682780321523868e-05,
+      "loss": 0.6389,
+      "step": 13637
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0681609986661872e-05,
+      "loss": 0.6717,
+      "step": 13638
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0680439642420323e-05,
+      "loss": 0.6781,
+      "step": 13639
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0679269288815328e-05,
+      "loss": 0.6033,
+      "step": 13640
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0678098925862988e-05,
+      "loss": 0.5964,
+      "step": 13641
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0676928553579414e-05,
+      "loss": 0.7016,
+      "step": 13642
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0675758171980706e-05,
+      "loss": 0.6201,
+      "step": 13643
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0674587781082973e-05,
+      "loss": 0.53,
+      "step": 13644
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0673417380902317e-05,
+      "loss": 0.6178,
+      "step": 13645
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0672246971454849e-05,
+      "loss": 0.6314,
+      "step": 13646
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0671076552756669e-05,
+      "loss": 0.663,
+      "step": 13647
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066990612482389e-05,
+      "loss": 0.6383,
+      "step": 13648
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0668735687672611e-05,
+      "loss": 0.5732,
+      "step": 13649
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066756524131894e-05,
+      "loss": 0.6427,
+      "step": 13650
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066639478577899e-05,
+      "loss": 0.5902,
+      "step": 13651
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0665224321068857e-05,
+      "loss": 0.666,
+      "step": 13652
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0664053847204656e-05,
+      "loss": 0.5794,
+      "step": 13653
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0662883364202488e-05,
+      "loss": 0.624,
+      "step": 13654
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0661712872078464e-05,
+      "loss": 0.6223,
+      "step": 13655
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066054237084869e-05,
+      "loss": 0.669,
+      "step": 13656
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0659371860529273e-05,
+      "loss": 0.6496,
+      "step": 13657
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0658201341136316e-05,
+      "loss": 0.6578,
+      "step": 13658
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0657030812685934e-05,
+      "loss": 0.6385,
+      "step": 13659
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.065586027519423e-05,
+      "loss": 0.5675,
+      "step": 13660
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.065468972867731e-05,
+      "loss": 0.5558,
+      "step": 13661
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0653519173151287e-05,
+      "loss": 0.6319,
+      "step": 13662
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0652348608632264e-05,
+      "loss": 0.5854,
+      "step": 13663
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0651178035136353e-05,
+      "loss": 0.6527,
+      "step": 13664
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0650007452679658e-05,
+      "loss": 0.5122,
+      "step": 13665
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.064883686127829e-05,
+      "loss": 0.6182,
+      "step": 13666
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0647666260948357e-05,
+      "loss": 0.6119,
+      "step": 13667
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0646495651705967e-05,
+      "loss": 0.5848,
+      "step": 13668
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.064532503356723e-05,
+      "loss": 0.6541,
+      "step": 13669
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.064415440654825e-05,
+      "loss": 0.5957,
+      "step": 13670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0642983770665144e-05,
+      "loss": 0.5556,
+      "step": 13671
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0641813125934013e-05,
+      "loss": 0.6224,
+      "step": 13672
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0640642472370972e-05,
+      "loss": 0.6988,
+      "step": 13673
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0639471809992126e-05,
+      "loss": 0.6105,
+      "step": 13674
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0638301138813588e-05,
+      "loss": 0.6572,
+      "step": 13675
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0637130458851462e-05,
+      "loss": 0.587,
+      "step": 13676
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0635959770121862e-05,
+      "loss": 0.6153,
+      "step": 13677
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0634789072640897e-05,
+      "loss": 0.5673,
+      "step": 13678
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0633618366424676e-05,
+      "loss": 0.5767,
+      "step": 13679
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0632447651489313e-05,
+      "loss": 0.6335,
+      "step": 13680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.063127692785091e-05,
+      "loss": 0.5785,
+      "step": 13681
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0630106195525585e-05,
+      "loss": 0.6278,
+      "step": 13682
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0628935454529443e-05,
+      "loss": 0.633,
+      "step": 13683
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0627764704878598e-05,
+      "loss": 0.5785,
+      "step": 13684
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.062659394658916e-05,
+      "loss": 0.5306,
+      "step": 13685
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0625423179677236e-05,
+      "loss": 0.6407,
+      "step": 13686
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0624252404158941e-05,
+      "loss": 0.662,
+      "step": 13687
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0623081620050386e-05,
+      "loss": 0.6465,
+      "step": 13688
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0621910827367676e-05,
+      "loss": 0.5923,
+      "step": 13689
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.062074002612693e-05,
+      "loss": 0.5015,
+      "step": 13690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0619569216344257e-05,
+      "loss": 0.6107,
+      "step": 13691
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0618398398035764e-05,
+      "loss": 0.6161,
+      "step": 13692
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0617227571217567e-05,
+      "loss": 0.5773,
+      "step": 13693
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0616056735905775e-05,
+      "loss": 0.5849,
+      "step": 13694
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0614885892116504e-05,
+      "loss": 0.6322,
+      "step": 13695
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0613715039865859e-05,
+      "loss": 0.5544,
+      "step": 13696
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.061254417916996e-05,
+      "loss": 0.589,
+      "step": 13697
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.061137331004491e-05,
+      "loss": 0.5696,
+      "step": 13698
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0610202432506831e-05,
+      "loss": 0.5756,
+      "step": 13699
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0609031546571827e-05,
+      "loss": 0.5909,
+      "step": 13700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0607860652256015e-05,
+      "loss": 0.6855,
+      "step": 13701
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0606689749575504e-05,
+      "loss": 0.5848,
+      "step": 13702
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0605518838546412e-05,
+      "loss": 0.6464,
+      "step": 13703
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060434791918485e-05,
+      "loss": 0.5595,
+      "step": 13704
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0603176991506925e-05,
+      "loss": 0.5554,
+      "step": 13705
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060200605552876e-05,
+      "loss": 0.6501,
+      "step": 13706
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0600835111266457e-05,
+      "loss": 0.6377,
+      "step": 13707
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.059966415873614e-05,
+      "loss": 0.5822,
+      "step": 13708
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0598493197953913e-05,
+      "loss": 0.6694,
+      "step": 13709
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0597322228935895e-05,
+      "loss": 0.6108,
+      "step": 13710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05961512516982e-05,
+      "loss": 0.6954,
+      "step": 13711
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0594980266256938e-05,
+      "loss": 0.6446,
+      "step": 13712
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0593809272628226e-05,
+      "loss": 0.6271,
+      "step": 13713
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0592638270828175e-05,
+      "loss": 0.6068,
+      "step": 13714
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0591467260872905e-05,
+      "loss": 0.6376,
+      "step": 13715
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0590296242778521e-05,
+      "loss": 0.6289,
+      "step": 13716
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0589125216561144e-05,
+      "loss": 0.6807,
+      "step": 13717
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0587954182236886e-05,
+      "loss": 0.6125,
+      "step": 13718
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0586783139821863e-05,
+      "loss": 0.677,
+      "step": 13719
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0585612089332186e-05,
+      "loss": 0.6141,
+      "step": 13720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0584441030783971e-05,
+      "loss": 0.6234,
+      "step": 13721
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058326996419334e-05,
+      "loss": 0.572,
+      "step": 13722
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0582098889576396e-05,
+      "loss": 0.6115,
+      "step": 13723
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0580927806949264e-05,
+      "loss": 0.6711,
+      "step": 13724
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0579756716328052e-05,
+      "loss": 0.6977,
+      "step": 13725
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.057858561772888e-05,
+      "loss": 0.6635,
+      "step": 13726
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0577414511167858e-05,
+      "loss": 0.5739,
+      "step": 13727
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.057624339666111e-05,
+      "loss": 0.5583,
+      "step": 13728
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0575072274224743e-05,
+      "loss": 0.6329,
+      "step": 13729
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0573901143874878e-05,
+      "loss": 0.6221,
+      "step": 13730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0572730005627628e-05,
+      "loss": 0.6099,
+      "step": 13731
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0571558859499106e-05,
+      "loss": 0.6033,
+      "step": 13732
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0570387705505438e-05,
+      "loss": 0.6252,
+      "step": 13733
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0569216543662733e-05,
+      "loss": 0.6835,
+      "step": 13734
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0568045373987107e-05,
+      "loss": 0.555,
+      "step": 13735
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0566874196494677e-05,
+      "loss": 0.6025,
+      "step": 13736
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0565703011201563e-05,
+      "loss": 0.6252,
+      "step": 13737
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0564531818123874e-05,
+      "loss": 0.5698,
+      "step": 13738
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0563360617277734e-05,
+      "loss": 0.6541,
+      "step": 13739
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0562189408679255e-05,
+      "loss": 0.5855,
+      "step": 13740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0561018192344557e-05,
+      "loss": 0.576,
+      "step": 13741
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0559846968289755e-05,
+      "loss": 0.608,
+      "step": 13742
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0558675736530968e-05,
+      "loss": 0.5775,
+      "step": 13743
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0557504497084312e-05,
+      "loss": 0.6051,
+      "step": 13744
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0556333249965902e-05,
+      "loss": 0.6047,
+      "step": 13745
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0555161995191863e-05,
+      "loss": 0.6162,
+      "step": 13746
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0553990732778303e-05,
+      "loss": 0.5559,
+      "step": 13747
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0552819462741343e-05,
+      "loss": 0.6334,
+      "step": 13748
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0551648185097103e-05,
+      "loss": 0.629,
+      "step": 13749
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0550476899861699e-05,
+      "loss": 0.6679,
+      "step": 13750
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054930560705125e-05,
+      "loss": 0.603,
+      "step": 13751
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0548134306681871e-05,
+      "loss": 0.6088,
+      "step": 13752
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0546962998769686e-05,
+      "loss": 0.5644,
+      "step": 13753
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0545791683330803e-05,
+      "loss": 0.608,
+      "step": 13754
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0544620360381355e-05,
+      "loss": 0.5916,
+      "step": 13755
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0543449029937447e-05,
+      "loss": 0.652,
+      "step": 13756
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0542277692015204e-05,
+      "loss": 0.6354,
+      "step": 13757
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0541106346630744e-05,
+      "loss": 0.6083,
+      "step": 13758
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0539934993800183e-05,
+      "loss": 0.6413,
+      "step": 13759
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0538763633539643e-05,
+      "loss": 0.6715,
+      "step": 13760
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0537592265865245e-05,
+      "loss": 0.6504,
+      "step": 13761
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0536420890793101e-05,
+      "loss": 0.5844,
+      "step": 13762
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0535249508339337e-05,
+      "loss": 0.6663,
+      "step": 13763
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0534078118520068e-05,
+      "loss": 0.5944,
+      "step": 13764
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0532906721351415e-05,
+      "loss": 0.613,
+      "step": 13765
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0531735316849498e-05,
+      "loss": 0.561,
+      "step": 13766
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0530563905030436e-05,
+      "loss": 0.6976,
+      "step": 13767
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0529392485910346e-05,
+      "loss": 0.603,
+      "step": 13768
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0528221059505353e-05,
+      "loss": 0.5767,
+      "step": 13769
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0527049625831572e-05,
+      "loss": 0.6076,
+      "step": 13770
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0525878184905127e-05,
+      "loss": 0.6195,
+      "step": 13771
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0524706736742135e-05,
+      "loss": 0.5226,
+      "step": 13772
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0523535281358718e-05,
+      "loss": 0.5537,
+      "step": 13773
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0522363818770995e-05,
+      "loss": 0.5748,
+      "step": 13774
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.052119234899509e-05,
+      "loss": 0.6893,
+      "step": 13775
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0520020872047115e-05,
+      "loss": 0.5554,
+      "step": 13776
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0518849387943198e-05,
+      "loss": 0.6395,
+      "step": 13777
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0517677896699456e-05,
+      "loss": 0.6107,
+      "step": 13778
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0516506398332014e-05,
+      "loss": 0.6387,
+      "step": 13779
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.051533489285699e-05,
+      "loss": 0.7128,
+      "step": 13780
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0514163380290502e-05,
+      "loss": 0.6165,
+      "step": 13781
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0512991860648678e-05,
+      "loss": 0.6112,
+      "step": 13782
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0511820333947632e-05,
+      "loss": 0.6379,
+      "step": 13783
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0510648800203491e-05,
+      "loss": 0.6065,
+      "step": 13784
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0509477259432372e-05,
+      "loss": 0.6534,
+      "step": 13785
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05083057116504e-05,
+      "loss": 0.5983,
+      "step": 13786
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0507134156873693e-05,
+      "loss": 0.5701,
+      "step": 13787
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0505962595118376e-05,
+      "loss": 0.593,
+      "step": 13788
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0504791026400567e-05,
+      "loss": 0.5474,
+      "step": 13789
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0503619450736392e-05,
+      "loss": 0.6177,
+      "step": 13790
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0502447868141969e-05,
+      "loss": 0.6165,
+      "step": 13791
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0501276278633423e-05,
+      "loss": 0.5657,
+      "step": 13792
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0500104682226876e-05,
+      "loss": 0.6147,
+      "step": 13793
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0498933078938447e-05,
+      "loss": 0.5944,
+      "step": 13794
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0497761468784262e-05,
+      "loss": 0.6589,
+      "step": 13795
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049658985178044e-05,
+      "loss": 0.5612,
+      "step": 13796
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0495418227943109e-05,
+      "loss": 0.6789,
+      "step": 13797
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0494246597288382e-05,
+      "loss": 0.5892,
+      "step": 13798
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049307495983239e-05,
+      "loss": 0.6386,
+      "step": 13799
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0491903315591256e-05,
+      "loss": 0.5739,
+      "step": 13800
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0490731664581098e-05,
+      "loss": 0.6404,
+      "step": 13801
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0489560006818042e-05,
+      "loss": 0.5442,
+      "step": 13802
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.048838834231821e-05,
+      "loss": 0.6533,
+      "step": 13803
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0487216671097725e-05,
+      "loss": 0.6719,
+      "step": 13804
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0486044993172708e-05,
+      "loss": 0.5962,
+      "step": 13805
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.048487330855929e-05,
+      "loss": 0.6405,
+      "step": 13806
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0483701617273585e-05,
+      "loss": 0.6336,
+      "step": 13807
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0482529919331723e-05,
+      "loss": 0.6863,
+      "step": 13808
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0481358214749824e-05,
+      "loss": 0.6332,
+      "step": 13809
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0480186503544013e-05,
+      "loss": 0.6719,
+      "step": 13810
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0479014785730414e-05,
+      "loss": 0.6103,
+      "step": 13811
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.047784306132515e-05,
+      "loss": 0.552,
+      "step": 13812
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0476671330344349e-05,
+      "loss": 0.6037,
+      "step": 13813
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0475499592804128e-05,
+      "loss": 0.6278,
+      "step": 13814
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0474327848720616e-05,
+      "loss": 0.6123,
+      "step": 13815
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0473156098109935e-05,
+      "loss": 0.6806,
+      "step": 13816
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0471984340988212e-05,
+      "loss": 0.5862,
+      "step": 13817
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.047081257737157e-05,
+      "loss": 0.6311,
+      "step": 13818
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0469640807276133e-05,
+      "loss": 0.7038,
+      "step": 13819
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0468469030718023e-05,
+      "loss": 0.7073,
+      "step": 13820
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0467297247713368e-05,
+      "loss": 0.6808,
+      "step": 13821
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0466125458278295e-05,
+      "loss": 0.5818,
+      "step": 13822
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0464953662428925e-05,
+      "loss": 0.5602,
+      "step": 13823
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0463781860181385e-05,
+      "loss": 0.6254,
+      "step": 13824
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0462610051551798e-05,
+      "loss": 0.657,
+      "step": 13825
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0461438236556292e-05,
+      "loss": 0.5837,
+      "step": 13826
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0460266415210986e-05,
+      "loss": 0.6166,
+      "step": 13827
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0459094587532015e-05,
+      "loss": 0.7324,
+      "step": 13828
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0457922753535494e-05,
+      "loss": 0.6456,
+      "step": 13829
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0456750913237557e-05,
+      "loss": 0.5669,
+      "step": 13830
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0455579066654327e-05,
+      "loss": 0.6706,
+      "step": 13831
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0454407213801924e-05,
+      "loss": 0.5597,
+      "step": 13832
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0453235354696482e-05,
+      "loss": 0.6426,
+      "step": 13833
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0452063489354124e-05,
+      "loss": 0.6137,
+      "step": 13834
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0450891617790975e-05,
+      "loss": 0.632,
+      "step": 13835
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0449719740023158e-05,
+      "loss": 0.6502,
+      "step": 13836
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0448547856066806e-05,
+      "loss": 0.626,
+      "step": 13837
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0447375965938037e-05,
+      "loss": 0.6048,
+      "step": 13838
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0446204069652986e-05,
+      "loss": 0.6401,
+      "step": 13839
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0445032167227773e-05,
+      "loss": 0.5595,
+      "step": 13840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0443860258678525e-05,
+      "loss": 0.5702,
+      "step": 13841
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0442688344021372e-05,
+      "loss": 0.6061,
+      "step": 13842
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0441516423272437e-05,
+      "loss": 0.624,
+      "step": 13843
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.044034449644785e-05,
+      "loss": 0.6365,
+      "step": 13844
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0439172563563734e-05,
+      "loss": 0.6032,
+      "step": 13845
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043800062463622e-05,
+      "loss": 0.6441,
+      "step": 13846
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043682867968143e-05,
+      "loss": 0.639,
+      "step": 13847
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0435656728715496e-05,
+      "loss": 0.633,
+      "step": 13848
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0434484771754538e-05,
+      "loss": 0.6066,
+      "step": 13849
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0433312808814694e-05,
+      "loss": 0.6417,
+      "step": 13850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0432140839912081e-05,
+      "loss": 0.6093,
+      "step": 13851
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043096886506283e-05,
+      "loss": 0.6366,
+      "step": 13852
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0429796884283071e-05,
+      "loss": 0.7059,
+      "step": 13853
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0428624897588926e-05,
+      "loss": 0.6188,
+      "step": 13854
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0427452904996528e-05,
+      "loss": 0.6144,
+      "step": 13855
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0426280906522003e-05,
+      "loss": 0.5773,
+      "step": 13856
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0425108902181475e-05,
+      "loss": 0.5593,
+      "step": 13857
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.042393689199108e-05,
+      "loss": 0.5886,
+      "step": 13858
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0422764875966938e-05,
+      "loss": 0.6488,
+      "step": 13859
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.042159285412518e-05,
+      "loss": 0.5907,
+      "step": 13860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0420420826481935e-05,
+      "loss": 0.5782,
+      "step": 13861
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.041924879305333e-05,
+      "loss": 0.58,
+      "step": 13862
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.041807675385549e-05,
+      "loss": 0.6131,
+      "step": 13863
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.041690470890455e-05,
+      "loss": 0.6524,
+      "step": 13864
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0415732658216634e-05,
+      "loss": 0.5608,
+      "step": 13865
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.041456060180787e-05,
+      "loss": 0.5915,
+      "step": 13866
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0413388539694388e-05,
+      "loss": 0.6716,
+      "step": 13867
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0412216471892316e-05,
+      "loss": 0.5986,
+      "step": 13868
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0411044398417784e-05,
+      "loss": 0.6387,
+      "step": 13869
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.040987231928692e-05,
+      "loss": 0.6809,
+      "step": 13870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0408700234515855e-05,
+      "loss": 0.6754,
+      "step": 13871
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0407528144120713e-05,
+      "loss": 0.608,
+      "step": 13872
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0406356048117628e-05,
+      "loss": 0.7252,
+      "step": 13873
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0405183946522723e-05,
+      "loss": 0.5866,
+      "step": 13874
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0404011839352133e-05,
+      "loss": 0.6151,
+      "step": 13875
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0402839726621986e-05,
+      "loss": 0.6618,
+      "step": 13876
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0401667608348408e-05,
+      "loss": 0.5474,
+      "step": 13877
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0400495484547533e-05,
+      "loss": 0.629,
+      "step": 13878
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0399323355235486e-05,
+      "loss": 0.5848,
+      "step": 13879
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.03981512204284e-05,
+      "loss": 0.6085,
+      "step": 13880
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0396979080142405e-05,
+      "loss": 0.5873,
+      "step": 13881
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0395806934393626e-05,
+      "loss": 0.6426,
+      "step": 13882
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0394634783198198e-05,
+      "loss": 0.6193,
+      "step": 13883
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0393462626572249e-05,
+      "loss": 0.7106,
+      "step": 13884
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0392290464531906e-05,
+      "loss": 0.5649,
+      "step": 13885
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0391118297093303e-05,
+      "loss": 0.5823,
+      "step": 13886
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.038994612427257e-05,
+      "loss": 0.6225,
+      "step": 13887
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0388773946085833e-05,
+      "loss": 0.6735,
+      "step": 13888
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0387601762549227e-05,
+      "loss": 0.5727,
+      "step": 13889
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0386429573678878e-05,
+      "loss": 0.5986,
+      "step": 13890
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.038525737949092e-05,
+      "loss": 0.6177,
+      "step": 13891
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0384085180001483e-05,
+      "loss": 0.579,
+      "step": 13892
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0382912975226696e-05,
+      "loss": 0.7107,
+      "step": 13893
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0381740765182687e-05,
+      "loss": 0.6094,
+      "step": 13894
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0380568549885593e-05,
+      "loss": 0.6929,
+      "step": 13895
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0379396329351538e-05,
+      "loss": 0.6316,
+      "step": 13896
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0378224103596662e-05,
+      "loss": 0.6209,
+      "step": 13897
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0377051872637085e-05,
+      "loss": 0.5969,
+      "step": 13898
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0375879636488942e-05,
+      "loss": 0.6084,
+      "step": 13899
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0374707395168368e-05,
+      "loss": 0.622,
+      "step": 13900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.037353514869149e-05,
+      "loss": 0.6188,
+      "step": 13901
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0372362897074441e-05,
+      "loss": 0.6592,
+      "step": 13902
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0371190640333351e-05,
+      "loss": 0.6489,
+      "step": 13903
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0370018378484351e-05,
+      "loss": 0.611,
+      "step": 13904
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0368846111543574e-05,
+      "loss": 0.6724,
+      "step": 13905
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.036767383952715e-05,
+      "loss": 0.6054,
+      "step": 13906
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.036650156245121e-05,
+      "loss": 0.6484,
+      "step": 13907
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0365329280331886e-05,
+      "loss": 0.5598,
+      "step": 13908
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0364156993185312e-05,
+      "loss": 0.624,
+      "step": 13909
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0362984701027617e-05,
+      "loss": 0.6025,
+      "step": 13910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0361812403874934e-05,
+      "loss": 0.5911,
+      "step": 13911
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0360640101743393e-05,
+      "loss": 0.6407,
+      "step": 13912
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0359467794649129e-05,
+      "loss": 0.5719,
+      "step": 13913
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0358295482608269e-05,
+      "loss": 0.6169,
+      "step": 13914
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0357123165636953e-05,
+      "loss": 0.5207,
+      "step": 13915
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0355950843751306e-05,
+      "loss": 0.6635,
+      "step": 13916
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0354778516967463e-05,
+      "loss": 0.5512,
+      "step": 13917
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0353606185301554e-05,
+      "loss": 0.6478,
+      "step": 13918
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0352433848769717e-05,
+      "loss": 0.6536,
+      "step": 13919
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0351261507388077e-05,
+      "loss": 0.6294,
+      "step": 13920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0350089161172771e-05,
+      "loss": 0.6474,
+      "step": 13921
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0348916810139933e-05,
+      "loss": 0.5834,
+      "step": 13922
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.034774445430569e-05,
+      "loss": 0.6497,
+      "step": 13923
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0346572093686181e-05,
+      "loss": 0.5636,
+      "step": 13924
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0345399728297532e-05,
+      "loss": 0.6444,
+      "step": 13925
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0344227358155883e-05,
+      "loss": 0.6385,
+      "step": 13926
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0343054983277358e-05,
+      "loss": 0.6408,
+      "step": 13927
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0341882603678099e-05,
+      "loss": 0.628,
+      "step": 13928
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0340710219374233e-05,
+      "loss": 0.626,
+      "step": 13929
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0339537830381893e-05,
+      "loss": 0.6265,
+      "step": 13930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.033836543671722e-05,
+      "loss": 0.6096,
+      "step": 13931
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0337193038396338e-05,
+      "loss": 0.6049,
+      "step": 13932
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0336020635435384e-05,
+      "loss": 0.6229,
+      "step": 13933
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.033484822785049e-05,
+      "loss": 0.6529,
+      "step": 13934
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.033367581565779e-05,
+      "loss": 0.6411,
+      "step": 13935
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0332503398873418e-05,
+      "loss": 0.6387,
+      "step": 13936
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.033133097751351e-05,
+      "loss": 0.5853,
+      "step": 13937
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0330158551594191e-05,
+      "loss": 0.6476,
+      "step": 13938
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0328986121131604e-05,
+      "loss": 0.6154,
+      "step": 13939
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0327813686141878e-05,
+      "loss": 0.6494,
+      "step": 13940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0326641246641147e-05,
+      "loss": 0.6289,
+      "step": 13941
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0325468802645546e-05,
+      "loss": 0.6621,
+      "step": 13942
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0324296354171209e-05,
+      "loss": 0.6308,
+      "step": 13943
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.032312390123427e-05,
+      "loss": 0.6441,
+      "step": 13944
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.032195144385086e-05,
+      "loss": 0.5899,
+      "step": 13945
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0320778982037117e-05,
+      "loss": 0.6038,
+      "step": 13946
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.031960651580917e-05,
+      "loss": 0.6801,
+      "step": 13947
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.031843404518316e-05,
+      "loss": 0.6052,
+      "step": 13948
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0317261570175216e-05,
+      "loss": 0.563,
+      "step": 13949
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0316089090801474e-05,
+      "loss": 0.6421,
+      "step": 13950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.031491660707807e-05,
+      "loss": 0.6067,
+      "step": 13951
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0313744119021134e-05,
+      "loss": 0.6282,
+      "step": 13952
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0312571626646805e-05,
+      "loss": 0.6481,
+      "step": 13953
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0311399129971215e-05,
+      "loss": 0.6219,
+      "step": 13954
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0310226629010503e-05,
+      "loss": 0.6346,
+      "step": 13955
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0309054123780797e-05,
+      "loss": 0.6819,
+      "step": 13956
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0307881614298234e-05,
+      "loss": 0.5451,
+      "step": 13957
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0306709100578952e-05,
+      "loss": 0.5584,
+      "step": 13958
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0305536582639082e-05,
+      "loss": 0.598,
+      "step": 13959
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.030436406049476e-05,
+      "loss": 0.6123,
+      "step": 13960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.030319153416212e-05,
+      "loss": 0.7359,
+      "step": 13961
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.03020190036573e-05,
+      "loss": 0.6147,
+      "step": 13962
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0300846468996431e-05,
+      "loss": 0.665,
+      "step": 13963
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0299673930195654e-05,
+      "loss": 0.6562,
+      "step": 13964
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0298501387271094e-05,
+      "loss": 0.5931,
+      "step": 13965
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0297328840238896e-05,
+      "loss": 0.6179,
+      "step": 13966
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0296156289115194e-05,
+      "loss": 0.6104,
+      "step": 13967
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0294983733916118e-05,
+      "loss": 0.6599,
+      "step": 13968
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0293811174657809e-05,
+      "loss": 0.6364,
+      "step": 13969
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0292638611356398e-05,
+      "loss": 0.578,
+      "step": 13970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0291466044028024e-05,
+      "loss": 0.5483,
+      "step": 13971
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0290293472688821e-05,
+      "loss": 0.5558,
+      "step": 13972
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0289120897354925e-05,
+      "loss": 0.6989,
+      "step": 13973
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.028794831804247e-05,
+      "loss": 0.6547,
+      "step": 13974
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0286775734767596e-05,
+      "loss": 0.6874,
+      "step": 13975
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0285603147546433e-05,
+      "loss": 0.6355,
+      "step": 13976
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.028443055639512e-05,
+      "loss": 0.5965,
+      "step": 13977
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0283257961329794e-05,
+      "loss": 0.6434,
+      "step": 13978
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0282085362366587e-05,
+      "loss": 0.6684,
+      "step": 13979
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0280912759521641e-05,
+      "loss": 0.5976,
+      "step": 13980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0279740152811087e-05,
+      "loss": 0.5933,
+      "step": 13981
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0278567542251063e-05,
+      "loss": 0.67,
+      "step": 13982
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0277394927857705e-05,
+      "loss": 0.588,
+      "step": 13983
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027622230964715e-05,
+      "loss": 0.5853,
+      "step": 13984
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027504968763553e-05,
+      "loss": 0.5388,
+      "step": 13985
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0273877061838986e-05,
+      "loss": 0.6426,
+      "step": 13986
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0272704432273654e-05,
+      "loss": 0.6051,
+      "step": 13987
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027153179895567e-05,
+      "loss": 0.646,
+      "step": 13988
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027035916190117e-05,
+      "loss": 0.6363,
+      "step": 13989
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0269186521126287e-05,
+      "loss": 0.6446,
+      "step": 13990
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0268013876647165e-05,
+      "loss": 0.608,
+      "step": 13991
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0266841228479933e-05,
+      "loss": 0.5736,
+      "step": 13992
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0265668576640735e-05,
+      "loss": 0.6333,
+      "step": 13993
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.02644959211457e-05,
+      "loss": 0.6579,
+      "step": 13994
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0263323262010972e-05,
+      "loss": 0.5747,
+      "step": 13995
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0262150599252682e-05,
+      "loss": 0.7066,
+      "step": 13996
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.026097793288697e-05,
+      "loss": 0.7069,
+      "step": 13997
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0259805262929974e-05,
+      "loss": 0.5982,
+      "step": 13998
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0258632589397828e-05,
+      "loss": 0.6156,
+      "step": 13999
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.025745991230667e-05,
+      "loss": 0.7272,
+      "step": 14000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0256287231672637e-05,
+      "loss": 0.5158,
+      "step": 14001
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0255114547511868e-05,
+      "loss": 0.6881,
+      "step": 14002
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0253941859840498e-05,
+      "loss": 0.5901,
+      "step": 14003
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0252769168674666e-05,
+      "loss": 0.6383,
+      "step": 14004
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0251596474030506e-05,
+      "loss": 0.5759,
+      "step": 14005
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.025042377592416e-05,
+      "loss": 0.5736,
+      "step": 14006
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024925107437176e-05,
+      "loss": 0.5875,
+      "step": 14007
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0248078369389446e-05,
+      "loss": 0.6424,
+      "step": 14008
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0246905660993359e-05,
+      "loss": 0.6217,
+      "step": 14009
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0245732949199629e-05,
+      "loss": 0.6013,
+      "step": 14010
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0244560234024403e-05,
+      "loss": 0.6773,
+      "step": 14011
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024338751548381e-05,
+      "loss": 0.5269,
+      "step": 14012
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0242214793593993e-05,
+      "loss": 0.6084,
+      "step": 14013
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0241042068371084e-05,
+      "loss": 0.6353,
+      "step": 14014
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0239869339831225e-05,
+      "loss": 0.6134,
+      "step": 14015
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0238696607990554e-05,
+      "loss": 0.5745,
+      "step": 14016
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0237523872865207e-05,
+      "loss": 0.5459,
+      "step": 14017
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0236351134471325e-05,
+      "loss": 0.5922,
+      "step": 14018
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0235178392825043e-05,
+      "loss": 0.6307,
+      "step": 14019
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0234005647942499e-05,
+      "loss": 0.6136,
+      "step": 14020
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0232832899839833e-05,
+      "loss": 0.5852,
+      "step": 14021
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0231660148533183e-05,
+      "loss": 0.6494,
+      "step": 14022
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0230487394038681e-05,
+      "loss": 0.5989,
+      "step": 14023
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0229314636372474e-05,
+      "loss": 0.5941,
+      "step": 14024
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0228141875550695e-05,
+      "loss": 0.5939,
+      "step": 14025
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0226969111589482e-05,
+      "loss": 0.6083,
+      "step": 14026
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0225796344504975e-05,
+      "loss": 0.6379,
+      "step": 14027
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0224623574313312e-05,
+      "loss": 0.608,
+      "step": 14028
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0223450801030631e-05,
+      "loss": 0.5836,
+      "step": 14029
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0222278024673072e-05,
+      "loss": 0.5767,
+      "step": 14030
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0221105245256774e-05,
+      "loss": 0.6214,
+      "step": 14031
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0219932462797868e-05,
+      "loss": 0.5692,
+      "step": 14032
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0218759677312502e-05,
+      "loss": 0.6128,
+      "step": 14033
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0217586888816808e-05,
+      "loss": 0.6467,
+      "step": 14034
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0216414097326933e-05,
+      "loss": 0.6499,
+      "step": 14035
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0215241302859003e-05,
+      "loss": 0.5927,
+      "step": 14036
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.021406850542917e-05,
+      "loss": 0.6262,
+      "step": 14037
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0212895705053559e-05,
+      "loss": 0.6496,
+      "step": 14038
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.021172290174832e-05,
+      "loss": 0.6438,
+      "step": 14039
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0210550095529587e-05,
+      "loss": 0.6038,
+      "step": 14040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.02093772864135e-05,
+      "loss": 0.579,
+      "step": 14041
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0208204474416199e-05,
+      "loss": 0.6222,
+      "step": 14042
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.020703165955382e-05,
+      "loss": 0.5632,
+      "step": 14043
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0205858841842504e-05,
+      "loss": 0.6079,
+      "step": 14044
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0204686021298387e-05,
+      "loss": 0.6251,
+      "step": 14045
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0203513197937614e-05,
+      "loss": 0.6969,
+      "step": 14046
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0202340371776318e-05,
+      "loss": 0.6924,
+      "step": 14047
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.020116754283064e-05,
+      "loss": 0.654,
+      "step": 14048
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0199994711116723e-05,
+      "loss": 0.5608,
+      "step": 14049
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0198821876650702e-05,
+      "loss": 0.5934,
+      "step": 14050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0197649039448718e-05,
+      "loss": 0.5852,
+      "step": 14051
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0196476199526907e-05,
+      "loss": 0.5782,
+      "step": 14052
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0195303356901414e-05,
+      "loss": 0.6078,
+      "step": 14053
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0194130511588371e-05,
+      "loss": 0.6659,
+      "step": 14054
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0192957663603927e-05,
+      "loss": 0.5904,
+      "step": 14055
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0191784812964212e-05,
+      "loss": 0.6015,
+      "step": 14056
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0190611959685372e-05,
+      "loss": 0.6293,
+      "step": 14057
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0189439103783541e-05,
+      "loss": 0.6489,
+      "step": 14058
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0188266245274862e-05,
+      "loss": 0.637,
+      "step": 14059
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0187093384175476e-05,
+      "loss": 0.6274,
+      "step": 14060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.018592052050152e-05,
+      "loss": 0.6566,
+      "step": 14061
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0184747654269136e-05,
+      "loss": 0.6522,
+      "step": 14062
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.018357478549446e-05,
+      "loss": 0.6194,
+      "step": 14063
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0182401914193631e-05,
+      "loss": 0.6364,
+      "step": 14064
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0181229040382797e-05,
+      "loss": 0.6282,
+      "step": 14065
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0180056164078087e-05,
+      "loss": 0.6076,
+      "step": 14066
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.017888328529565e-05,
+      "loss": 0.5665,
+      "step": 14067
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0177710404051619e-05,
+      "loss": 0.6297,
+      "step": 14068
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.017653752036214e-05,
+      "loss": 0.5801,
+      "step": 14069
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0175364634243345e-05,
+      "loss": 0.5657,
+      "step": 14070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0174191745711381e-05,
+      "loss": 0.5613,
+      "step": 14071
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0173018854782386e-05,
+      "loss": 0.6697,
+      "step": 14072
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.01718459614725e-05,
+      "loss": 0.6147,
+      "step": 14073
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.017067306579786e-05,
+      "loss": 0.5599,
+      "step": 14074
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.016950016777461e-05,
+      "loss": 0.6729,
+      "step": 14075
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0168327267418889e-05,
+      "loss": 0.6207,
+      "step": 14076
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0167154364746836e-05,
+      "loss": 0.5715,
+      "step": 14077
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0165981459774595e-05,
+      "loss": 0.6108,
+      "step": 14078
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0164808552518299e-05,
+      "loss": 0.588,
+      "step": 14079
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0163635642994094e-05,
+      "loss": 0.6455,
+      "step": 14080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.016246273121812e-05,
+      "loss": 0.5481,
+      "step": 14081
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0161289817206517e-05,
+      "loss": 0.5818,
+      "step": 14082
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.016011690097542e-05,
+      "loss": 0.5991,
+      "step": 14083
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0158943982540978e-05,
+      "loss": 0.5658,
+      "step": 14084
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0157771061919327e-05,
+      "loss": 0.6316,
+      "step": 14085
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0156598139126603e-05,
+      "loss": 0.6073,
+      "step": 14086
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0155425214178955e-05,
+      "loss": 0.5983,
+      "step": 14087
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.015425228709252e-05,
+      "loss": 0.6249,
+      "step": 14088
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0153079357883435e-05,
+      "loss": 0.6746,
+      "step": 14089
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0151906426567845e-05,
+      "loss": 0.5776,
+      "step": 14090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.015073349316189e-05,
+      "loss": 0.5699,
+      "step": 14091
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0149560557681708e-05,
+      "loss": 0.6406,
+      "step": 14092
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0148387620143445e-05,
+      "loss": 0.6244,
+      "step": 14093
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0147214680563233e-05,
+      "loss": 0.6015,
+      "step": 14094
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.014604173895722e-05,
+      "loss": 0.6387,
+      "step": 14095
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0144868795341544e-05,
+      "loss": 0.6291,
+      "step": 14096
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0143695849732347e-05,
+      "loss": 0.6009,
+      "step": 14097
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0142522902145768e-05,
+      "loss": 0.5791,
+      "step": 14098
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0141349952597948e-05,
+      "loss": 0.6098,
+      "step": 14099
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0140177001105031e-05,
+      "loss": 0.5707,
+      "step": 14100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0139004047683152e-05,
+      "loss": 0.64,
+      "step": 14101
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0137831092348458e-05,
+      "loss": 0.6999,
+      "step": 14102
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0136658135117086e-05,
+      "loss": 0.606,
+      "step": 14103
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0135485176005178e-05,
+      "loss": 0.5645,
+      "step": 14104
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0134312215028875e-05,
+      "loss": 0.6269,
+      "step": 14105
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0133139252204315e-05,
+      "loss": 0.6971,
+      "step": 14106
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0131966287547646e-05,
+      "loss": 0.6236,
+      "step": 14107
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0130793321075003e-05,
+      "loss": 0.5702,
+      "step": 14108
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.012962035280253e-05,
+      "loss": 0.5783,
+      "step": 14109
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0128447382746365e-05,
+      "loss": 0.6356,
+      "step": 14110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0127274410922653e-05,
+      "loss": 0.6112,
+      "step": 14111
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0126101437347533e-05,
+      "loss": 0.582,
+      "step": 14112
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0124928462037149e-05,
+      "loss": 0.5763,
+      "step": 14113
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0123755485007637e-05,
+      "loss": 0.6038,
+      "step": 14114
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0122582506275141e-05,
+      "loss": 0.5549,
+      "step": 14115
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0121409525855798e-05,
+      "loss": 0.5769,
+      "step": 14116
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0120236543765759e-05,
+      "loss": 0.6243,
+      "step": 14117
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0119063560021156e-05,
+      "loss": 0.6571,
+      "step": 14118
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0117890574638137e-05,
+      "loss": 0.6303,
+      "step": 14119
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0116717587632838e-05,
+      "loss": 0.6942,
+      "step": 14120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.01155445990214e-05,
+      "loss": 0.5831,
+      "step": 14121
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0114371608819972e-05,
+      "loss": 0.6249,
+      "step": 14122
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0113198617044686e-05,
+      "loss": 0.6062,
+      "step": 14123
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0112025623711692e-05,
+      "loss": 0.6053,
+      "step": 14124
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0110852628837122e-05,
+      "loss": 0.6319,
+      "step": 14125
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0109679632437123e-05,
+      "loss": 0.6126,
+      "step": 14126
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0108506634527839e-05,
+      "loss": 0.6449,
+      "step": 14127
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0107333635125407e-05,
+      "loss": 0.6765,
+      "step": 14128
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0106160634245972e-05,
+      "loss": 0.6188,
+      "step": 14129
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010498763190567e-05,
+      "loss": 0.6339,
+      "step": 14130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0103814628120648e-05,
+      "loss": 0.6515,
+      "step": 14131
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0102641622907043e-05,
+      "loss": 0.655,
+      "step": 14132
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0101468616281003e-05,
+      "loss": 0.6529,
+      "step": 14133
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010029560825866e-05,
+      "loss": 0.7021,
+      "step": 14134
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0099122598856168e-05,
+      "loss": 0.6266,
+      "step": 14135
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0097949588089657e-05,
+      "loss": 0.596,
+      "step": 14136
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0096776575975276e-05,
+      "loss": 0.6074,
+      "step": 14137
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0095603562529163e-05,
+      "loss": 0.5667,
+      "step": 14138
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0094430547767462e-05,
+      "loss": 0.6381,
+      "step": 14139
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0093257531706313e-05,
+      "loss": 0.616,
+      "step": 14140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0092084514361857e-05,
+      "loss": 0.5998,
+      "step": 14141
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0090911495750243e-05,
+      "loss": 0.6017,
+      "step": 14142
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00897384758876e-05,
+      "loss": 0.6397,
+      "step": 14143
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.008856545479008e-05,
+      "loss": 0.5824,
+      "step": 14144
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.008739243247382e-05,
+      "loss": 0.5488,
+      "step": 14145
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0086219408954966e-05,
+      "loss": 0.5642,
+      "step": 14146
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0085046384249654e-05,
+      "loss": 0.642,
+      "step": 14147
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.008387335837403e-05,
+      "loss": 0.628,
+      "step": 14148
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0082700331344238e-05,
+      "loss": 0.6417,
+      "step": 14149
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0081527303176413e-05,
+      "loss": 0.5939,
+      "step": 14150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0080354273886701e-05,
+      "loss": 0.5378,
+      "step": 14151
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0079181243491245e-05,
+      "loss": 0.6288,
+      "step": 14152
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0078008212006185e-05,
+      "loss": 0.6021,
+      "step": 14153
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0076835179447664e-05,
+      "loss": 0.6197,
+      "step": 14154
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0075662145831823e-05,
+      "loss": 0.6437,
+      "step": 14155
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0074489111174804e-05,
+      "loss": 0.6077,
+      "step": 14156
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0073316075492748e-05,
+      "loss": 0.594,
+      "step": 14157
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00721430388018e-05,
+      "loss": 0.5976,
+      "step": 14158
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00709700011181e-05,
+      "loss": 0.6441,
+      "step": 14159
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0069796962457792e-05,
+      "loss": 0.6586,
+      "step": 14160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0068623922837016e-05,
+      "loss": 0.5504,
+      "step": 14161
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0067450882271915e-05,
+      "loss": 0.6065,
+      "step": 14162
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0066277840778626e-05,
+      "loss": 0.6391,
+      "step": 14163
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0065104798373302e-05,
+      "loss": 0.5913,
+      "step": 14164
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0063931755072074e-05,
+      "loss": 0.6887,
+      "step": 14165
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0062758710891091e-05,
+      "loss": 0.5548,
+      "step": 14166
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0061585665846494e-05,
+      "loss": 0.6197,
+      "step": 14167
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0060412619954423e-05,
+      "loss": 0.636,
+      "step": 14168
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0059239573231021e-05,
+      "loss": 0.5548,
+      "step": 14169
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.005806652569243e-05,
+      "loss": 0.6151,
+      "step": 14170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0056893477354795e-05,
+      "loss": 0.5874,
+      "step": 14171
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0055720428234253e-05,
+      "loss": 0.6943,
+      "step": 14172
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0054547378346951e-05,
+      "loss": 0.5823,
+      "step": 14173
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0053374327709027e-05,
+      "loss": 0.6366,
+      "step": 14174
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0052201276336624e-05,
+      "loss": 0.5685,
+      "step": 14175
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0051028224245888e-05,
+      "loss": 0.5588,
+      "step": 14176
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004985517145296e-05,
+      "loss": 0.6102,
+      "step": 14177
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004868211797398e-05,
+      "loss": 0.5795,
+      "step": 14178
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004750906382509e-05,
+      "loss": 0.5838,
+      "step": 14179
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0046336009022435e-05,
+      "loss": 0.5927,
+      "step": 14180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0045162953582154e-05,
+      "loss": 0.6676,
+      "step": 14181
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0043989897520394e-05,
+      "loss": 0.5935,
+      "step": 14182
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0042816840853293e-05,
+      "loss": 0.5918,
+      "step": 14183
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0041643783596993e-05,
+      "loss": 0.5919,
+      "step": 14184
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004047072576764e-05,
+      "loss": 0.6619,
+      "step": 14185
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0039297667381373e-05,
+      "loss": 0.663,
+      "step": 14186
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0038124608454337e-05,
+      "loss": 0.6524,
+      "step": 14187
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0036951549002672e-05,
+      "loss": 0.6279,
+      "step": 14188
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0035778489042522e-05,
+      "loss": 0.606,
+      "step": 14189
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0034605428590027e-05,
+      "loss": 0.5925,
+      "step": 14190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0033432367661333e-05,
+      "loss": 0.6633,
+      "step": 14191
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0032259306272579e-05,
+      "loss": 0.5507,
+      "step": 14192
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.003108624443991e-05,
+      "loss": 0.6649,
+      "step": 14193
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0029913182179463e-05,
+      "loss": 0.6697,
+      "step": 14194
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0028740119507386e-05,
+      "loss": 0.7169,
+      "step": 14195
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0027567056439821e-05,
+      "loss": 0.633,
+      "step": 14196
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002639399299291e-05,
+      "loss": 0.6281,
+      "step": 14197
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0025220929182793e-05,
+      "loss": 0.6085,
+      "step": 14198
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0024047865025612e-05,
+      "loss": 0.6318,
+      "step": 14199
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0022874800537514e-05,
+      "loss": 0.6674,
+      "step": 14200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0021701735734636e-05,
+      "loss": 0.6304,
+      "step": 14201
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0020528670633127e-05,
+      "loss": 0.6204,
+      "step": 14202
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0019355605249123e-05,
+      "loss": 0.6021,
+      "step": 14203
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.001818253959877e-05,
+      "loss": 0.6398,
+      "step": 14204
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0017009473698207e-05,
+      "loss": 0.6053,
+      "step": 14205
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0015836407563579e-05,
+      "loss": 0.6559,
+      "step": 14206
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.001466334121103e-05,
+      "loss": 0.6006,
+      "step": 14207
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00134902746567e-05,
+      "loss": 0.5941,
+      "step": 14208
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0012317207916733e-05,
+      "loss": 0.5814,
+      "step": 14209
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0011144141007269e-05,
+      "loss": 0.6846,
+      "step": 14210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0009971073944453e-05,
+      "loss": 0.6422,
+      "step": 14211
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0008798006744425e-05,
+      "loss": 0.6039,
+      "step": 14212
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0007624939423328e-05,
+      "loss": 0.5973,
+      "step": 14213
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0006451871997306e-05,
+      "loss": 0.5249,
+      "step": 14214
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0005278804482502e-05,
+      "loss": 0.5798,
+      "step": 14215
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0004105736895056e-05,
+      "loss": 0.621,
+      "step": 14216
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0002932669251109e-05,
+      "loss": 0.5802,
+      "step": 14217
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.000175960156681e-05,
+      "loss": 0.6934,
+      "step": 14218
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0000586533858294e-05,
+      "loss": 0.5654,
+      "step": 14219
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.999413466141711e-06,
+      "loss": 0.6177,
+      "step": 14220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.998240398433195e-06,
+      "loss": 0.6234,
+      "step": 14221
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.997067330748894e-06,
+      "loss": 0.5556,
+      "step": 14222
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.995894263104951e-06,
+      "loss": 0.6578,
+      "step": 14223
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.994721195517503e-06,
+      "loss": 0.6256,
+      "step": 14224
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.993548128002699e-06,
+      "loss": 0.5933,
+      "step": 14225
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.992375060576675e-06,
+      "loss": 0.6919,
+      "step": 14226
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.991201993255582e-06,
+      "loss": 0.6717,
+      "step": 14227
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.990028926055552e-06,
+      "loss": 0.5793,
+      "step": 14228
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.988855858992736e-06,
+      "loss": 0.675,
+      "step": 14229
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.987682792083272e-06,
+      "loss": 0.5919,
+      "step": 14230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.986509725343303e-06,
+      "loss": 0.594,
+      "step": 14231
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.985336658788973e-06,
+      "loss": 0.6178,
+      "step": 14232
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.984163592436423e-06,
+      "loss": 0.5883,
+      "step": 14233
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.982990526301798e-06,
+      "loss": 0.6167,
+      "step": 14234
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.981817460401235e-06,
+      "loss": 0.5859,
+      "step": 14235
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.980644394750882e-06,
+      "loss": 0.5567,
+      "step": 14236
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.979471329366878e-06,
+      "loss": 0.5946,
+      "step": 14237
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.978298264265367e-06,
+      "loss": 0.6105,
+      "step": 14238
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.977125199462491e-06,
+      "loss": 0.6509,
+      "step": 14239
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.975952134974393e-06,
+      "loss": 0.6404,
+      "step": 14240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.974779070817213e-06,
+      "loss": 0.5837,
+      "step": 14241
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.973606007007096e-06,
+      "loss": 0.6222,
+      "step": 14242
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.972432943560184e-06,
+      "loss": 0.5415,
+      "step": 14243
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.971259880492617e-06,
+      "loss": 0.5684,
+      "step": 14244
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.970086817820542e-06,
+      "loss": 0.5342,
+      "step": 14245
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.968913755560097e-06,
+      "loss": 0.7033,
+      "step": 14246
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.967740693727428e-06,
+      "loss": 0.5962,
+      "step": 14247
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.966567632338672e-06,
+      "loss": 0.687,
+      "step": 14248
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.96539457140998e-06,
+      "loss": 0.5822,
+      "step": 14249
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.964221510957483e-06,
+      "loss": 0.694,
+      "step": 14250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.963048450997333e-06,
+      "loss": 0.5881,
+      "step": 14251
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.961875391545668e-06,
+      "loss": 0.5993,
+      "step": 14252
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.96070233261863e-06,
+      "loss": 0.6584,
+      "step": 14253
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.959529274232364e-06,
+      "loss": 0.6004,
+      "step": 14254
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.95835621640301e-06,
+      "loss": 0.5851,
+      "step": 14255
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.957183159146713e-06,
+      "loss": 0.6041,
+      "step": 14256
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.956010102479611e-06,
+      "loss": 0.5829,
+      "step": 14257
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.95483704641785e-06,
+      "loss": 0.6842,
+      "step": 14258
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.953663990977568e-06,
+      "loss": 0.6262,
+      "step": 14259
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.952490936174914e-06,
+      "loss": 0.5796,
+      "step": 14260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.951317882026025e-06,
+      "loss": 0.609,
+      "step": 14261
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.950144828547043e-06,
+      "loss": 0.5668,
+      "step": 14262
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.948971775754115e-06,
+      "loss": 0.7045,
+      "step": 14263
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.947798723663377e-06,
+      "loss": 0.59,
+      "step": 14264
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.94662567229098e-06,
+      "loss": 0.629,
+      "step": 14265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.945452621653054e-06,
+      "loss": 0.5831,
+      "step": 14266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.944279571765752e-06,
+      "loss": 0.5862,
+      "step": 14267
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.94310652264521e-06,
+      "loss": 0.625,
+      "step": 14268
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.941933474307575e-06,
+      "loss": 0.6384,
+      "step": 14269
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.940760426768982e-06,
+      "loss": 0.5709,
+      "step": 14270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.939587380045582e-06,
+      "loss": 0.6126,
+      "step": 14271
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.93841433415351e-06,
+      "loss": 0.5673,
+      "step": 14272
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.93724128910891e-06,
+      "loss": 0.6296,
+      "step": 14273
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.936068244927929e-06,
+      "loss": 0.5833,
+      "step": 14274
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.934895201626701e-06,
+      "loss": 0.6506,
+      "step": 14275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.933722159221375e-06,
+      "loss": 0.6411,
+      "step": 14276
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.93254911772809e-06,
+      "loss": 0.607,
+      "step": 14277
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.93137607716299e-06,
+      "loss": 0.6068,
+      "step": 14278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.930203037542211e-06,
+      "loss": 0.5775,
+      "step": 14279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.929029998881903e-06,
+      "loss": 0.5725,
+      "step": 14280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.927856961198203e-06,
+      "loss": 0.5629,
+      "step": 14281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.926683924507256e-06,
+      "loss": 0.6032,
+      "step": 14282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9255108888252e-06,
+      "loss": 0.6687,
+      "step": 14283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.92433785416818e-06,
+      "loss": 0.6191,
+      "step": 14284
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.923164820552341e-06,
+      "loss": 0.6434,
+      "step": 14285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.921991787993817e-06,
+      "loss": 0.5946,
+      "step": 14286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.920818756508758e-06,
+      "loss": 0.5748,
+      "step": 14287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9196457261133e-06,
+      "loss": 0.5882,
+      "step": 14288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91847269682359e-06,
+      "loss": 0.6253,
+      "step": 14289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.917299668655767e-06,
+      "loss": 0.5328,
+      "step": 14290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.916126641625973e-06,
+      "loss": 0.5954,
+      "step": 14291
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.914953615750349e-06,
+      "loss": 0.6507,
+      "step": 14292
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.913780591045037e-06,
+      "loss": 0.6081,
+      "step": 14293
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.912607567526183e-06,
+      "loss": 0.5876,
+      "step": 14294
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.911434545209922e-06,
+      "loss": 0.5885,
+      "step": 14295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.910261524112404e-06,
+      "loss": 0.5946,
+      "step": 14296
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.909088504249762e-06,
+      "loss": 0.6685,
+      "step": 14297
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.907915485638144e-06,
+      "loss": 0.6436,
+      "step": 14298
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.906742468293689e-06,
+      "loss": 0.6175,
+      "step": 14299
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.905569452232541e-06,
+      "loss": 0.6288,
+      "step": 14300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.904396437470839e-06,
+      "loss": 0.6192,
+      "step": 14301
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.903223424024729e-06,
+      "loss": 0.6154,
+      "step": 14302
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.902050411910346e-06,
+      "loss": 0.6156,
+      "step": 14303
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.900877401143834e-06,
+      "loss": 0.556,
+      "step": 14304
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89970439174134e-06,
+      "loss": 0.6193,
+      "step": 14305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.898531383719e-06,
+      "loss": 0.6231,
+      "step": 14306
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.897358377092959e-06,
+      "loss": 0.6004,
+      "step": 14307
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.896185371879354e-06,
+      "loss": 0.5878,
+      "step": 14308
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.895012368094332e-06,
+      "loss": 0.5932,
+      "step": 14309
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.893839365754031e-06,
+      "loss": 0.6553,
+      "step": 14310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.892666364874595e-06,
+      "loss": 0.6316,
+      "step": 14311
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.891493365472163e-06,
+      "loss": 0.6305,
+      "step": 14312
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.890320367562879e-06,
+      "loss": 0.6796,
+      "step": 14313
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.88914737116288e-06,
+      "loss": 0.6252,
+      "step": 14314
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.887974376288312e-06,
+      "loss": 0.6196,
+      "step": 14315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.886801382955316e-06,
+      "loss": 0.5991,
+      "step": 14316
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.885628391180031e-06,
+      "loss": 0.684,
+      "step": 14317
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.884455400978601e-06,
+      "loss": 0.5613,
+      "step": 14318
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.883282412367165e-06,
+      "loss": 0.6383,
+      "step": 14319
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.882109425361868e-06,
+      "loss": 0.5524,
+      "step": 14320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.880936439978846e-06,
+      "loss": 0.6392,
+      "step": 14321
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.879763456234246e-06,
+      "loss": 0.6272,
+      "step": 14322
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.878590474144204e-06,
+      "loss": 0.5233,
+      "step": 14323
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.877417493724864e-06,
+      "loss": 0.6178,
+      "step": 14324
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.876244514992368e-06,
+      "loss": 0.6308,
+      "step": 14325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.875071537962855e-06,
+      "loss": 0.6437,
+      "step": 14326
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.87389856265247e-06,
+      "loss": 0.6007,
+      "step": 14327
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.872725589077348e-06,
+      "loss": 0.6277,
+      "step": 14328
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.871552617253636e-06,
+      "loss": 0.5585,
+      "step": 14329
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.870379647197472e-06,
+      "loss": 0.6645,
+      "step": 14330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.869206678925e-06,
+      "loss": 0.5918,
+      "step": 14331
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.868033712452356e-06,
+      "loss": 0.6671,
+      "step": 14332
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.866860747795686e-06,
+      "loss": 0.5558,
+      "step": 14333
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.865687784971128e-06,
+      "loss": 0.6373,
+      "step": 14334
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.864514823994824e-06,
+      "loss": 0.6422,
+      "step": 14335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.863341864882918e-06,
+      "loss": 0.6133,
+      "step": 14336
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.862168907651546e-06,
+      "loss": 0.6258,
+      "step": 14337
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.860995952316851e-06,
+      "loss": 0.6268,
+      "step": 14338
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.859822998894972e-06,
+      "loss": 0.6461,
+      "step": 14339
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.858650047402053e-06,
+      "loss": 0.6357,
+      "step": 14340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.857477097854234e-06,
+      "loss": 0.605,
+      "step": 14341
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.856304150267656e-06,
+      "loss": 0.6184,
+      "step": 14342
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.855131204658457e-06,
+      "loss": 0.6999,
+      "step": 14343
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.853958261042784e-06,
+      "loss": 0.556,
+      "step": 14344
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.852785319436769e-06,
+      "loss": 0.5873,
+      "step": 14345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.851612379856559e-06,
+      "loss": 0.5457,
+      "step": 14346
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.850439442318294e-06,
+      "loss": 0.6292,
+      "step": 14347
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.849266506838113e-06,
+      "loss": 0.562,
+      "step": 14348
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.848093573432158e-06,
+      "loss": 0.6084,
+      "step": 14349
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.846920642116567e-06,
+      "loss": 0.6698,
+      "step": 14350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.845747712907484e-06,
+      "loss": 0.54,
+      "step": 14351
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.844574785821047e-06,
+      "loss": 0.5546,
+      "step": 14352
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8434018608734e-06,
+      "loss": 0.6521,
+      "step": 14353
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.842228938080678e-06,
+      "loss": 0.6522,
+      "step": 14354
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.841056017459024e-06,
+      "loss": 0.6453,
+      "step": 14355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.839883099024581e-06,
+      "loss": 0.5945,
+      "step": 14356
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.838710182793488e-06,
+      "loss": 0.6328,
+      "step": 14357
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.837537268781884e-06,
+      "loss": 0.6205,
+      "step": 14358
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.836364357005908e-06,
+      "loss": 0.5598,
+      "step": 14359
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.835191447481705e-06,
+      "loss": 0.6582,
+      "step": 14360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.83401854022541e-06,
+      "loss": 0.5854,
+      "step": 14361
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.832845635253167e-06,
+      "loss": 0.5627,
+      "step": 14362
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.831672732581115e-06,
+      "loss": 0.6082,
+      "step": 14363
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.830499832225394e-06,
+      "loss": 0.6404,
+      "step": 14364
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.829326934202143e-06,
+      "loss": 0.5551,
+      "step": 14365
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.828154038527503e-06,
+      "loss": 0.5865,
+      "step": 14366
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.826981145217618e-06,
+      "loss": 0.679,
+      "step": 14367
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.825808254288622e-06,
+      "loss": 0.6438,
+      "step": 14368
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.824635365756658e-06,
+      "loss": 0.5716,
+      "step": 14369
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.823462479637864e-06,
+      "loss": 0.6509,
+      "step": 14370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.822289595948384e-06,
+      "loss": 0.6401,
+      "step": 14371
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.821116714704353e-06,
+      "loss": 0.5935,
+      "step": 14372
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.819943835921916e-06,
+      "loss": 0.628,
+      "step": 14373
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.818770959617207e-06,
+      "loss": 0.6443,
+      "step": 14374
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.81759808580637e-06,
+      "loss": 0.5687,
+      "step": 14375
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.816425214505542e-06,
+      "loss": 0.6694,
+      "step": 14376
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.815252345730866e-06,
+      "loss": 0.601,
+      "step": 14377
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.814079479498481e-06,
+      "loss": 0.6409,
+      "step": 14378
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.812906615824524e-06,
+      "loss": 0.6302,
+      "step": 14379
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.811733754725138e-06,
+      "loss": 0.6435,
+      "step": 14380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.810560896216459e-06,
+      "loss": 0.59,
+      "step": 14381
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.809388040314632e-06,
+      "loss": 0.597,
+      "step": 14382
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80821518703579e-06,
+      "loss": 0.5983,
+      "step": 14383
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.807042336396077e-06,
+      "loss": 0.6303,
+      "step": 14384
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.805869488411627e-06,
+      "loss": 0.6916,
+      "step": 14385
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.804696643098586e-06,
+      "loss": 0.6378,
+      "step": 14386
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.803523800473094e-06,
+      "loss": 0.6912,
+      "step": 14387
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.802350960551284e-06,
+      "loss": 0.5935,
+      "step": 14388
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.801178123349298e-06,
+      "loss": 0.5549,
+      "step": 14389
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.800005288883277e-06,
+      "loss": 0.5718,
+      "step": 14390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79883245716936e-06,
+      "loss": 0.6034,
+      "step": 14391
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.797659628223682e-06,
+      "loss": 0.5478,
+      "step": 14392
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.796486802062388e-06,
+      "loss": 0.5623,
+      "step": 14393
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.795313978701613e-06,
+      "loss": 0.6472,
+      "step": 14394
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7941411581575e-06,
+      "loss": 0.6589,
+      "step": 14395
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.792968340446182e-06,
+      "loss": 0.6875,
+      "step": 14396
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.791795525583803e-06,
+      "loss": 0.5506,
+      "step": 14397
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7906227135865e-06,
+      "loss": 0.5523,
+      "step": 14398
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.789449904470413e-06,
+      "loss": 0.5899,
+      "step": 14399
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.788277098251681e-06,
+      "loss": 0.5917,
+      "step": 14400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.787104294946441e-06,
+      "loss": 0.6791,
+      "step": 14401
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.785931494570836e-06,
+      "loss": 0.6249,
+      "step": 14402
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.784758697140997e-06,
+      "loss": 0.598,
+      "step": 14403
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.783585902673072e-06,
+      "loss": 0.5814,
+      "step": 14404
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.782413111183192e-06,
+      "loss": 0.625,
+      "step": 14405
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7812403226875e-06,
+      "loss": 0.5455,
+      "step": 14406
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.780067537202132e-06,
+      "loss": 0.6335,
+      "step": 14407
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.778894754743228e-06,
+      "loss": 0.6468,
+      "step": 14408
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.777721975326928e-06,
+      "loss": 0.5985,
+      "step": 14409
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.776549198969369e-06,
+      "loss": 0.5864,
+      "step": 14410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.77537642568669e-06,
+      "loss": 0.5848,
+      "step": 14411
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.774203655495025e-06,
+      "loss": 0.5643,
+      "step": 14412
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.77303088841052e-06,
+      "loss": 0.6462,
+      "step": 14413
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.771858124449307e-06,
+      "loss": 0.5609,
+      "step": 14414
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.77068536362753e-06,
+      "loss": 0.6194,
+      "step": 14415
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.76951260596132e-06,
+      "loss": 0.6059,
+      "step": 14416
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.768339851466818e-06,
+      "loss": 0.5555,
+      "step": 14417
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.767167100160167e-06,
+      "loss": 0.5717,
+      "step": 14418
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7659943520575e-06,
+      "loss": 0.6293,
+      "step": 14419
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.764821607174957e-06,
+      "loss": 0.6651,
+      "step": 14420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.763648865528674e-06,
+      "loss": 0.5986,
+      "step": 14421
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.762476127134793e-06,
+      "loss": 0.606,
+      "step": 14422
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.761303392009446e-06,
+      "loss": 0.5932,
+      "step": 14423
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.760130660168777e-06,
+      "loss": 0.5902,
+      "step": 14424
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.758957931628918e-06,
+      "loss": 0.6005,
+      "step": 14425
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.757785206406012e-06,
+      "loss": 0.6127,
+      "step": 14426
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.756612484516192e-06,
+      "loss": 0.6429,
+      "step": 14427
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.755439765975599e-06,
+      "loss": 0.6304,
+      "step": 14428
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.75426705080037e-06,
+      "loss": 0.626,
+      "step": 14429
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.753094339006641e-06,
+      "loss": 0.6641,
+      "step": 14430
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.751921630610557e-06,
+      "loss": 0.6051,
+      "step": 14431
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.750748925628244e-06,
+      "loss": 0.6682,
+      "step": 14432
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.749576224075846e-06,
+      "loss": 0.6768,
+      "step": 14433
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7484035259695e-06,
+      "loss": 0.5954,
+      "step": 14434
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.747230831325339e-06,
+      "loss": 0.5939,
+      "step": 14435
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.746058140159507e-06,
+      "loss": 0.584,
+      "step": 14436
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.744885452488137e-06,
+      "loss": 0.6269,
+      "step": 14437
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.743712768327368e-06,
+      "loss": 0.6374,
+      "step": 14438
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.742540087693334e-06,
+      "loss": 0.6242,
+      "step": 14439
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.741367410602177e-06,
+      "loss": 0.649,
+      "step": 14440
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.740194737070031e-06,
+      "loss": 0.6051,
+      "step": 14441
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.739022067113032e-06,
+      "loss": 0.5486,
+      "step": 14442
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.737849400747324e-06,
+      "loss": 0.6154,
+      "step": 14443
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.736676737989033e-06,
+      "loss": 0.6185,
+      "step": 14444
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.735504078854305e-06,
+      "loss": 0.6543,
+      "step": 14445
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.734331423359272e-06,
+      "loss": 0.5712,
+      "step": 14446
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.733158771520072e-06,
+      "loss": 0.5824,
+      "step": 14447
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.731986123352842e-06,
+      "loss": 0.6166,
+      "step": 14448
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.730813478873716e-06,
+      "loss": 0.5335,
+      "step": 14449
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.729640838098835e-06,
+      "loss": 0.5994,
+      "step": 14450
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.728468201044335e-06,
+      "loss": 0.6133,
+      "step": 14451
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72729556772635e-06,
+      "loss": 0.6454,
+      "step": 14452
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.726122938161017e-06,
+      "loss": 0.6253,
+      "step": 14453
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.724950312364474e-06,
+      "loss": 0.6292,
+      "step": 14454
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.723777690352855e-06,
+      "loss": 0.663,
+      "step": 14455
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7226050721423e-06,
+      "loss": 0.669,
+      "step": 14456
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.721432457748942e-06,
+      "loss": 0.5887,
+      "step": 14457
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.720259847188917e-06,
+      "loss": 0.5682,
+      "step": 14458
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.719087240478364e-06,
+      "loss": 0.6116,
+      "step": 14459
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.717914637633415e-06,
+      "loss": 0.6101,
+      "step": 14460
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.716742038670211e-06,
+      "loss": 0.6225,
+      "step": 14461
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.715569443604882e-06,
+      "loss": 0.6618,
+      "step": 14462
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.714396852453572e-06,
+      "loss": 0.5854,
+      "step": 14463
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.713224265232409e-06,
+      "loss": 0.5721,
+      "step": 14464
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.712051681957534e-06,
+      "loss": 0.6531,
+      "step": 14465
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.71087910264508e-06,
+      "loss": 0.6998,
+      "step": 14466
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.709706527311184e-06,
+      "loss": 0.5589,
+      "step": 14467
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.708533955971981e-06,
+      "loss": 0.6064,
+      "step": 14468
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.707361388643607e-06,
+      "loss": 0.5448,
+      "step": 14469
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.706188825342196e-06,
+      "loss": 0.6101,
+      "step": 14470
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.705016266083886e-06,
+      "loss": 0.6241,
+      "step": 14471
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.703843710884811e-06,
+      "loss": 0.5648,
+      "step": 14472
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.702671159761107e-06,
+      "loss": 0.63,
+      "step": 14473
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70149861272891e-06,
+      "loss": 0.6641,
+      "step": 14474
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.700326069804353e-06,
+      "loss": 0.6255,
+      "step": 14475
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.699153531003574e-06,
+      "loss": 0.6048,
+      "step": 14476
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.697980996342706e-06,
+      "loss": 0.611,
+      "step": 14477
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.696808465837886e-06,
+      "loss": 0.6401,
+      "step": 14478
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.695635939505245e-06,
+      "loss": 0.6527,
+      "step": 14479
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.694463417360921e-06,
+      "loss": 0.6329,
+      "step": 14480
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.693290899421053e-06,
+      "loss": 0.5873,
+      "step": 14481
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.692118385701768e-06,
+      "loss": 0.66,
+      "step": 14482
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.690945876219206e-06,
+      "loss": 0.5693,
+      "step": 14483
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.6897733709895e-06,
+      "loss": 0.5851,
+      "step": 14484
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.688600870028786e-06,
+      "loss": 0.6228,
+      "step": 14485
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.687428373353197e-06,
+      "loss": 0.6785,
+      "step": 14486
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68625588097887e-06,
+      "loss": 0.5382,
+      "step": 14487
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.685083392921933e-06,
+      "loss": 0.6438,
+      "step": 14488
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68391090919853e-06,
+      "loss": 0.6416,
+      "step": 14489
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.682738429824787e-06,
+      "loss": 0.5916,
+      "step": 14490
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.681565954816843e-06,
+      "loss": 0.6407,
+      "step": 14491
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.680393484190833e-06,
+      "loss": 0.6164,
+      "step": 14492
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.679221017962888e-06,
+      "loss": 0.6117,
+      "step": 14493
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.678048556149144e-06,
+      "loss": 0.5667,
+      "step": 14494
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.676876098765734e-06,
+      "loss": 0.6033,
+      "step": 14495
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.675703645828795e-06,
+      "loss": 0.6153,
+      "step": 14496
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.674531197354457e-06,
+      "loss": 0.595,
+      "step": 14497
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.673358753358857e-06,
+      "loss": 0.6217,
+      "step": 14498
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.672186313858125e-06,
+      "loss": 0.5754,
+      "step": 14499
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.671013878868401e-06,
+      "loss": 0.5807,
+      "step": 14500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.66984144840581e-06,
+      "loss": 0.5785,
+      "step": 14501
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.668669022486495e-06,
+      "loss": 0.5962,
+      "step": 14502
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.667496601126585e-06,
+      "loss": 0.653,
+      "step": 14503
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.666324184342211e-06,
+      "loss": 0.6289,
+      "step": 14504
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.665151772149513e-06,
+      "loss": 0.6032,
+      "step": 14505
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.663979364564619e-06,
+      "loss": 0.6382,
+      "step": 14506
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.662806961603665e-06,
+      "loss": 0.5924,
+      "step": 14507
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.661634563282782e-06,
+      "loss": 0.6727,
+      "step": 14508
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.660462169618108e-06,
+      "loss": 0.6132,
+      "step": 14509
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.659289780625768e-06,
+      "loss": 0.6411,
+      "step": 14510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.658117396321903e-06,
+      "loss": 0.6535,
+      "step": 14511
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.656945016722645e-06,
+      "loss": 0.6078,
+      "step": 14512
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.655772641844122e-06,
+      "loss": 0.5703,
+      "step": 14513
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.654600271702471e-06,
+      "loss": 0.555,
+      "step": 14514
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.653427906313822e-06,
+      "loss": 0.58,
+      "step": 14515
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.652255545694313e-06,
+      "loss": 0.6361,
+      "step": 14516
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.65108318986007e-06,
+      "loss": 0.598,
+      "step": 14517
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.64991083882723e-06,
+      "loss": 0.5749,
+      "step": 14518
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.648738492611924e-06,
+      "loss": 0.668,
+      "step": 14519
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.647566151230288e-06,
+      "loss": 0.5208,
+      "step": 14520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.646393814698448e-06,
+      "loss": 0.5588,
+      "step": 14521
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.645221483032538e-06,
+      "loss": 0.5971,
+      "step": 14522
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.644049156248697e-06,
+      "loss": 0.5417,
+      "step": 14523
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.64287683436305e-06,
+      "loss": 0.5785,
+      "step": 14524
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.641704517391733e-06,
+      "loss": 0.6408,
+      "step": 14525
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.640532205350874e-06,
+      "loss": 0.584,
+      "step": 14526
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.63935989825661e-06,
+      "loss": 0.5979,
+      "step": 14527
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.63818759612507e-06,
+      "loss": 0.6053,
+      "step": 14528
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.637015298972388e-06,
+      "loss": 0.6437,
+      "step": 14529
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.635843006814691e-06,
+      "loss": 0.5774,
+      "step": 14530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.634670719668117e-06,
+      "loss": 0.6384,
+      "step": 14531
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.633498437548794e-06,
+      "loss": 0.5338,
+      "step": 14532
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.632326160472854e-06,
+      "loss": 0.5734,
+      "step": 14533
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.631153888456431e-06,
+      "loss": 0.6828,
+      "step": 14534
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.629981621515652e-06,
+      "loss": 0.5554,
+      "step": 14535
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.628809359666654e-06,
+      "loss": 0.6234,
+      "step": 14536
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62763710292556e-06,
+      "loss": 0.6558,
+      "step": 14537
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.626464851308514e-06,
+      "loss": 0.5758,
+      "step": 14538
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.625292604831634e-06,
+      "loss": 0.6439,
+      "step": 14539
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62412036351106e-06,
+      "loss": 0.6018,
+      "step": 14540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.622948127362918e-06,
+      "loss": 0.6356,
+      "step": 14541
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.621775896403343e-06,
+      "loss": 0.6189,
+      "step": 14542
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.620603670648464e-06,
+      "loss": 0.6424,
+      "step": 14543
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.61943145011441e-06,
+      "loss": 0.6467,
+      "step": 14544
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.618259234817315e-06,
+      "loss": 0.5619,
+      "step": 14545
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.617087024773309e-06,
+      "loss": 0.609,
+      "step": 14546
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.615914819998522e-06,
+      "loss": 0.588,
+      "step": 14547
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.614742620509083e-06,
+      "loss": 0.6358,
+      "step": 14548
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.613570426321125e-06,
+      "loss": 0.6688,
+      "step": 14549
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.612398237450776e-06,
+      "loss": 0.6435,
+      "step": 14550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.61122605391417e-06,
+      "loss": 0.5773,
+      "step": 14551
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.610053875727435e-06,
+      "loss": 0.6175,
+      "step": 14552
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.608881702906698e-06,
+      "loss": 0.6494,
+      "step": 14553
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.607709535468097e-06,
+      "loss": 0.6157,
+      "step": 14554
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.606537373427754e-06,
+      "loss": 0.5767,
+      "step": 14555
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.605365216801805e-06,
+      "loss": 0.5847,
+      "step": 14556
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.604193065606376e-06,
+      "loss": 0.6508,
+      "step": 14557
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.6030209198576e-06,
+      "loss": 0.6345,
+      "step": 14558
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.601848779571603e-06,
+      "loss": 0.6222,
+      "step": 14559
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.600676644764517e-06,
+      "loss": 0.6351,
+      "step": 14560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.59950451545247e-06,
+      "loss": 0.5919,
+      "step": 14561
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.598332391651596e-06,
+      "loss": 0.6377,
+      "step": 14562
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.597160273378018e-06,
+      "loss": 0.5608,
+      "step": 14563
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.595988160647869e-06,
+      "loss": 0.6768,
+      "step": 14564
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.59481605347728e-06,
+      "loss": 0.6059,
+      "step": 14565
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.593643951882377e-06,
+      "loss": 0.6206,
+      "step": 14566
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.59247185587929e-06,
+      "loss": 0.642,
+      "step": 14567
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.591299765484147e-06,
+      "loss": 0.601,
+      "step": 14568
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.590127680713083e-06,
+      "loss": 0.5995,
+      "step": 14569
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.588955601582218e-06,
+      "loss": 0.5807,
+      "step": 14570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.587783528107687e-06,
+      "loss": 0.5904,
+      "step": 14571
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.586611460305614e-06,
+      "loss": 0.6,
+      "step": 14572
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.585439398192133e-06,
+      "loss": 0.5614,
+      "step": 14573
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.584267341783372e-06,
+      "loss": 0.6178,
+      "step": 14574
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.583095291095454e-06,
+      "loss": 0.6181,
+      "step": 14575
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.581923246144514e-06,
+      "loss": 0.5778,
+      "step": 14576
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.580751206946675e-06,
+      "loss": 0.617,
+      "step": 14577
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.57957917351807e-06,
+      "loss": 0.686,
+      "step": 14578
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.578407145874824e-06,
+      "loss": 0.5548,
+      "step": 14579
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.577235124033065e-06,
+      "loss": 0.6502,
+      "step": 14580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.576063108008924e-06,
+      "loss": 0.5865,
+      "step": 14581
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.574891097818527e-06,
+      "loss": 0.6318,
+      "step": 14582
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.573719093478e-06,
+      "loss": 0.5864,
+      "step": 14583
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.572547095003473e-06,
+      "loss": 0.6203,
+      "step": 14584
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.571375102411075e-06,
+      "loss": 0.638,
+      "step": 14585
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.570203115716932e-06,
+      "loss": 0.6148,
+      "step": 14586
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.569031134937172e-06,
+      "loss": 0.6055,
+      "step": 14587
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.567859160087922e-06,
+      "loss": 0.5598,
+      "step": 14588
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.566687191185311e-06,
+      "loss": 0.6411,
+      "step": 14589
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.565515228245462e-06,
+      "loss": 0.6699,
+      "step": 14590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.564343271284509e-06,
+      "loss": 0.6318,
+      "step": 14591
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.563171320318571e-06,
+      "loss": 0.644,
+      "step": 14592
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.561999375363784e-06,
+      "loss": 0.5719,
+      "step": 14593
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.560827436436266e-06,
+      "loss": 0.5817,
+      "step": 14594
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.55965550355215e-06,
+      "loss": 0.6253,
+      "step": 14595
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.558483576727563e-06,
+      "loss": 0.6151,
+      "step": 14596
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.557311655978628e-06,
+      "loss": 0.5983,
+      "step": 14597
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.556139741321475e-06,
+      "loss": 0.6163,
+      "step": 14598
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.554967832772228e-06,
+      "loss": 0.5722,
+      "step": 14599
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.553795930347016e-06,
+      "loss": 0.6215,
+      "step": 14600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.552624034061963e-06,
+      "loss": 0.5479,
+      "step": 14601
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.551452143933197e-06,
+      "loss": 0.6308,
+      "step": 14602
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.550280259976844e-06,
+      "loss": 0.5775,
+      "step": 14603
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.549108382209026e-06,
+      "loss": 0.639,
+      "step": 14604
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.547936510645878e-06,
+      "loss": 0.6507,
+      "step": 14605
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.546764645303518e-06,
+      "loss": 0.6595,
+      "step": 14606
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.545592786198076e-06,
+      "loss": 0.6087,
+      "step": 14607
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.544420933345675e-06,
+      "loss": 0.6417,
+      "step": 14608
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.543249086762443e-06,
+      "loss": 0.5747,
+      "step": 14609
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.542077246464506e-06,
+      "loss": 0.6119,
+      "step": 14610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.540905412467988e-06,
+      "loss": 0.5842,
+      "step": 14611
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.539733584789014e-06,
+      "loss": 0.6394,
+      "step": 14612
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.538561763443711e-06,
+      "loss": 0.5814,
+      "step": 14613
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.537389948448203e-06,
+      "loss": 0.5434,
+      "step": 14614
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.536218139818615e-06,
+      "loss": 0.589,
+      "step": 14615
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.535046337571075e-06,
+      "loss": 0.5514,
+      "step": 14616
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.533874541721704e-06,
+      "loss": 0.6037,
+      "step": 14617
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.532702752286632e-06,
+      "loss": 0.5959,
+      "step": 14618
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.531530969281977e-06,
+      "loss": 0.6582,
+      "step": 14619
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.53035919272387e-06,
+      "loss": 0.6268,
+      "step": 14620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.529187422628432e-06,
+      "loss": 0.5501,
+      "step": 14621
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52801565901179e-06,
+      "loss": 0.6496,
+      "step": 14622
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.526843901890065e-06,
+      "loss": 0.5698,
+      "step": 14623
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.525672151279386e-06,
+      "loss": 0.6676,
+      "step": 14624
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.524500407195873e-06,
+      "loss": 0.6388,
+      "step": 14625
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.523328669655653e-06,
+      "loss": 0.6389,
+      "step": 14626
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.522156938674851e-06,
+      "loss": 0.6342,
+      "step": 14627
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.520985214269586e-06,
+      "loss": 0.5989,
+      "step": 14628
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.519813496455989e-06,
+      "loss": 0.6905,
+      "step": 14629
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.518641785250177e-06,
+      "loss": 0.6087,
+      "step": 14630
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.51747008066828e-06,
+      "loss": 0.6425,
+      "step": 14631
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.516298382726415e-06,
+      "loss": 0.6428,
+      "step": 14632
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.515126691440714e-06,
+      "loss": 0.5543,
+      "step": 14633
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.513955006827292e-06,
+      "loss": 0.5625,
+      "step": 14634
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.512783328902275e-06,
+      "loss": 0.6559,
+      "step": 14635
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.511611657681792e-06,
+      "loss": 0.6228,
+      "step": 14636
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.510439993181958e-06,
+      "loss": 0.6539,
+      "step": 14637
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.509268335418902e-06,
+      "loss": 0.6185,
+      "step": 14638
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.508096684408744e-06,
+      "loss": 0.6548,
+      "step": 14639
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.506925040167609e-06,
+      "loss": 0.5605,
+      "step": 14640
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.505753402711618e-06,
+      "loss": 0.5921,
+      "step": 14641
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.504581772056898e-06,
+      "loss": 0.5468,
+      "step": 14642
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.503410148219565e-06,
+      "loss": 0.5712,
+      "step": 14643
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.502238531215743e-06,
+      "loss": 0.5633,
+      "step": 14644
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.50106692106156e-06,
+      "loss": 0.5918,
+      "step": 14645
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49989531777313e-06,
+      "loss": 0.5286,
+      "step": 14646
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49872372136658e-06,
+      "loss": 0.5782,
+      "step": 14647
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.497552131858035e-06,
+      "loss": 0.6043,
+      "step": 14648
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.496380549263613e-06,
+      "loss": 0.6091,
+      "step": 14649
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.495208973599438e-06,
+      "loss": 0.5704,
+      "step": 14650
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.494037404881629e-06,
+      "loss": 0.5719,
+      "step": 14651
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.492865843126312e-06,
+      "loss": 0.6058,
+      "step": 14652
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.491694288349605e-06,
+      "loss": 0.5792,
+      "step": 14653
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.490522740567633e-06,
+      "loss": 0.6595,
+      "step": 14654
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.489351199796514e-06,
+      "loss": 0.5557,
+      "step": 14655
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.488179666052374e-06,
+      "loss": 0.6913,
+      "step": 14656
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.487008139351327e-06,
+      "loss": 0.6789,
+      "step": 14657
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.485836619709501e-06,
+      "loss": 0.6761,
+      "step": 14658
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.484665107143017e-06,
+      "loss": 0.682,
+      "step": 14659
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.48349360166799e-06,
+      "loss": 0.5941,
+      "step": 14660
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.48232210330055e-06,
+      "loss": 0.5931,
+      "step": 14661
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.481150612056807e-06,
+      "loss": 0.6211,
+      "step": 14662
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47997912795289e-06,
+      "loss": 0.5969,
+      "step": 14663
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.478807651004918e-06,
+      "loss": 0.6647,
+      "step": 14664
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47763618122901e-06,
+      "loss": 0.5526,
+      "step": 14665
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.476464718641287e-06,
+      "loss": 0.6215,
+      "step": 14666
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47529326325787e-06,
+      "loss": 0.7061,
+      "step": 14667
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.474121815094878e-06,
+      "loss": 0.6624,
+      "step": 14668
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47295037416843e-06,
+      "loss": 0.6727,
+      "step": 14669
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.471778940494652e-06,
+      "loss": 0.6142,
+      "step": 14670
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.470607514089657e-06,
+      "loss": 0.6237,
+      "step": 14671
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46943609496957e-06,
+      "loss": 0.6194,
+      "step": 14672
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.468264683150507e-06,
+      "loss": 0.6515,
+      "step": 14673
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46709327864859e-06,
+      "loss": 0.6235,
+      "step": 14674
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.465921881479935e-06,
+      "loss": 0.5785,
+      "step": 14675
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.464750491660668e-06,
+      "loss": 0.5612,
+      "step": 14676
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.463579109206904e-06,
+      "loss": 0.5568,
+      "step": 14677
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46240773413476e-06,
+      "loss": 0.6029,
+      "step": 14678
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.461236366460362e-06,
+      "loss": 0.637,
+      "step": 14679
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46006500619982e-06,
+      "loss": 0.6102,
+      "step": 14680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.458893653369261e-06,
+      "loss": 0.62,
+      "step": 14681
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457722307984801e-06,
+      "loss": 0.6595,
+      "step": 14682
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.456550970062559e-06,
+      "loss": 0.5571,
+      "step": 14683
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.455379639618651e-06,
+      "loss": 0.5575,
+      "step": 14684
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4542083166692e-06,
+      "loss": 0.6166,
+      "step": 14685
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45303700123032e-06,
+      "loss": 0.5816,
+      "step": 14686
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.451865693318134e-06,
+      "loss": 0.6414,
+      "step": 14687
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.450694392948755e-06,
+      "loss": 0.6822,
+      "step": 14688
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.449523100138305e-06,
+      "loss": 0.591,
+      "step": 14689
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4483518149029e-06,
+      "loss": 0.5898,
+      "step": 14690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44718053725866e-06,
+      "loss": 0.5669,
+      "step": 14691
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.446009267221702e-06,
+      "loss": 0.5421,
+      "step": 14692
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44483800480814e-06,
+      "loss": 0.5574,
+      "step": 14693
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4436667500341e-06,
+      "loss": 0.6125,
+      "step": 14694
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44249550291569e-06,
+      "loss": 0.5945,
+      "step": 14695
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.441324263469035e-06,
+      "loss": 0.609,
+      "step": 14696
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.440153031710247e-06,
+      "loss": 0.6003,
+      "step": 14697
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.438981807655446e-06,
+      "loss": 0.5238,
+      "step": 14698
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437810591320748e-06,
+      "loss": 0.5801,
+      "step": 14699
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.436639382722268e-06,
+      "loss": 0.5971,
+      "step": 14700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.435468181876128e-06,
+      "loss": 0.5892,
+      "step": 14701
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.434296988798442e-06,
+      "loss": 0.6186,
+      "step": 14702
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.433125803505325e-06,
+      "loss": 0.6565,
+      "step": 14703
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.431954626012896e-06,
+      "loss": 0.6105,
+      "step": 14704
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.43078345633727e-06,
+      "loss": 0.6254,
+      "step": 14705
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.429612294494563e-06,
+      "loss": 0.6221,
+      "step": 14706
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.428441140500895e-06,
+      "loss": 0.616,
+      "step": 14707
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.427269994372376e-06,
+      "loss": 0.6192,
+      "step": 14708
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.426098856125125e-06,
+      "loss": 0.5418,
+      "step": 14709
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.42492772577526e-06,
+      "loss": 0.5854,
+      "step": 14710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.423756603338894e-06,
+      "loss": 0.6679,
+      "step": 14711
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.422585488832144e-06,
+      "loss": 0.6519,
+      "step": 14712
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.421414382271123e-06,
+      "loss": 0.6152,
+      "step": 14713
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.420243283671952e-06,
+      "loss": 0.6686,
+      "step": 14714
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41907219305074e-06,
+      "loss": 0.6421,
+      "step": 14715
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.417901110423605e-06,
+      "loss": 0.6505,
+      "step": 14716
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.416730035806664e-06,
+      "loss": 0.6051,
+      "step": 14717
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41555896921603e-06,
+      "loss": 0.641,
+      "step": 14718
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.414387910667816e-06,
+      "loss": 0.5852,
+      "step": 14719
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41321686017814e-06,
+      "loss": 0.6628,
+      "step": 14720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.412045817763116e-06,
+      "loss": 0.5667,
+      "step": 14721
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.410874783438858e-06,
+      "loss": 0.6041,
+      "step": 14722
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.409703757221482e-06,
+      "loss": 0.6115,
+      "step": 14723
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.408532739127098e-06,
+      "loss": 0.5769,
+      "step": 14724
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.407361729171827e-06,
+      "loss": 0.5898,
+      "step": 14725
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.406190727371775e-06,
+      "loss": 0.5712,
+      "step": 14726
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.405019733743066e-06,
+      "loss": 0.6709,
+      "step": 14727
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.403848748301802e-06,
+      "loss": 0.6363,
+      "step": 14728
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.402677771064108e-06,
+      "loss": 0.6234,
+      "step": 14729
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.40150680204609e-06,
+      "loss": 0.5729,
+      "step": 14730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.400335841263863e-06,
+      "loss": 0.5946,
+      "step": 14731
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.399164888733544e-06,
+      "loss": 0.6197,
+      "step": 14732
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.397993944471245e-06,
+      "loss": 0.6715,
+      "step": 14733
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.396823008493077e-06,
+      "loss": 0.5825,
+      "step": 14734
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.395652080815154e-06,
+      "loss": 0.6023,
+      "step": 14735
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.39448116145359e-06,
+      "loss": 0.626,
+      "step": 14736
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.393310250424497e-06,
+      "loss": 0.615,
+      "step": 14737
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.392139347743989e-06,
+      "loss": 0.5968,
+      "step": 14738
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.390968453428176e-06,
+      "loss": 0.5939,
+      "step": 14739
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.389797567493172e-06,
+      "loss": 0.6491,
+      "step": 14740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.388626689955091e-06,
+      "loss": 0.6332,
+      "step": 14741
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.387455820830044e-06,
+      "loss": 0.6113,
+      "step": 14742
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.386284960134143e-06,
+      "loss": 0.5816,
+      "step": 14743
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.3851141078835e-06,
+      "loss": 0.6115,
+      "step": 14744
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.383943264094227e-06,
+      "loss": 0.5978,
+      "step": 14745
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.382772428782436e-06,
+      "loss": 0.6301,
+      "step": 14746
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.38160160196424e-06,
+      "loss": 0.6104,
+      "step": 14747
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.380430783655748e-06,
+      "loss": 0.5763,
+      "step": 14748
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.379259973873075e-06,
+      "loss": 0.6215,
+      "step": 14749
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.378089172632325e-06,
+      "loss": 0.6061,
+      "step": 14750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.376918379949618e-06,
+      "loss": 0.6416,
+      "step": 14751
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.375747595841062e-06,
+      "loss": 0.6052,
+      "step": 14752
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.374576820322765e-06,
+      "loss": 0.6149,
+      "step": 14753
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.373406053410844e-06,
+      "loss": 0.6318,
+      "step": 14754
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.372235295121404e-06,
+      "loss": 0.6121,
+      "step": 14755
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37106454547056e-06,
+      "loss": 0.6405,
+      "step": 14756
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.369893804474416e-06,
+      "loss": 0.6267,
+      "step": 14757
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.368723072149093e-06,
+      "loss": 0.676,
+      "step": 14758
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.367552348510692e-06,
+      "loss": 0.5581,
+      "step": 14759
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.366381633575327e-06,
+      "loss": 0.652,
+      "step": 14760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.365210927359105e-06,
+      "loss": 0.6252,
+      "step": 14761
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.36404022987814e-06,
+      "loss": 0.5771,
+      "step": 14762
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.362869541148542e-06,
+      "loss": 0.6216,
+      "step": 14763
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.361698861186417e-06,
+      "loss": 0.5871,
+      "step": 14764
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.360528190007879e-06,
+      "loss": 0.6107,
+      "step": 14765
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.359357527629032e-06,
+      "loss": 0.659,
+      "step": 14766
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.35818687406599e-06,
+      "loss": 0.6255,
+      "step": 14767
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.357016229334859e-06,
+      "loss": 0.6127,
+      "step": 14768
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.355845593451751e-06,
+      "loss": 0.6542,
+      "step": 14769
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.354674966432772e-06,
+      "loss": 0.5767,
+      "step": 14770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.353504348294034e-06,
+      "loss": 0.6953,
+      "step": 14771
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.352333739051646e-06,
+      "loss": 0.6076,
+      "step": 14772
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.351163138721711e-06,
+      "loss": 0.6573,
+      "step": 14773
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.349992547320344e-06,
+      "loss": 0.5982,
+      "step": 14774
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34882196486365e-06,
+      "loss": 0.5543,
+      "step": 14775
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.347651391367738e-06,
+      "loss": 0.6218,
+      "step": 14776
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.346480826848714e-06,
+      "loss": 0.5746,
+      "step": 14777
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.345310271322692e-06,
+      "loss": 0.5919,
+      "step": 14778
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.344139724805774e-06,
+      "loss": 0.6336,
+      "step": 14779
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34296918731407e-06,
+      "loss": 0.5749,
+      "step": 14780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.341798658863686e-06,
+      "loss": 0.5574,
+      "step": 14781
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34062813947073e-06,
+      "loss": 0.6099,
+      "step": 14782
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.339457629151314e-06,
+      "loss": 0.6546,
+      "step": 14783
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.338287127921538e-06,
+      "loss": 0.5317,
+      "step": 14784
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.337116635797516e-06,
+      "loss": 0.617,
+      "step": 14785
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.335946152795347e-06,
+      "loss": 0.6189,
+      "step": 14786
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.334775678931147e-06,
+      "loss": 0.6127,
+      "step": 14787
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.333605214221014e-06,
+      "loss": 0.5687,
+      "step": 14788
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.332434758681062e-06,
+      "loss": 0.6432,
+      "step": 14789
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.331264312327394e-06,
+      "loss": 0.5616,
+      "step": 14790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.330093875176113e-06,
+      "loss": 0.5671,
+      "step": 14791
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.328923447243333e-06,
+      "loss": 0.6509,
+      "step": 14792
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.327753028545155e-06,
+      "loss": 0.6198,
+      "step": 14793
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.326582619097686e-06,
+      "loss": 0.646,
+      "step": 14794
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.32541221891703e-06,
+      "loss": 0.5817,
+      "step": 14795
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.324241828019299e-06,
+      "loss": 0.575,
+      "step": 14796
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.32307144642059e-06,
+      "loss": 0.6255,
+      "step": 14797
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.321901074137013e-06,
+      "loss": 0.5928,
+      "step": 14798
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.320730711184674e-06,
+      "loss": 0.5998,
+      "step": 14799
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.319560357579679e-06,
+      "loss": 0.5964,
+      "step": 14800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31839001333813e-06,
+      "loss": 0.5984,
+      "step": 14801
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.317219678476132e-06,
+      "loss": 0.6179,
+      "step": 14802
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.316049353009793e-06,
+      "loss": 0.6908,
+      "step": 14803
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.314879036955213e-06,
+      "loss": 0.6367,
+      "step": 14804
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.313708730328503e-06,
+      "loss": 0.6147,
+      "step": 14805
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.312538433145763e-06,
+      "loss": 0.6024,
+      "step": 14806
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.311368145423098e-06,
+      "loss": 0.5522,
+      "step": 14807
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.310197867176612e-06,
+      "loss": 0.5778,
+      "step": 14808
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.30902759842241e-06,
+      "loss": 0.6333,
+      "step": 14809
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.307857339176593e-06,
+      "loss": 0.6081,
+      "step": 14810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.306687089455271e-06,
+      "loss": 0.5822,
+      "step": 14811
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.305516849274542e-06,
+      "loss": 0.6377,
+      "step": 14812
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.30434661865051e-06,
+      "loss": 0.6705,
+      "step": 14813
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.303176397599283e-06,
+      "loss": 0.7087,
+      "step": 14814
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.302006186136959e-06,
+      "loss": 0.6194,
+      "step": 14815
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.300835984279644e-06,
+      "loss": 0.6819,
+      "step": 14816
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.299665792043437e-06,
+      "loss": 0.5946,
+      "step": 14817
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.29849560944445e-06,
+      "loss": 0.5415,
+      "step": 14818
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.297325436498775e-06,
+      "loss": 0.6785,
+      "step": 14819
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.296155273222521e-06,
+      "loss": 0.6051,
+      "step": 14820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.294985119631788e-06,
+      "loss": 0.64,
+      "step": 14821
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.293814975742679e-06,
+      "loss": 0.611,
+      "step": 14822
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2926448415713e-06,
+      "loss": 0.5649,
+      "step": 14823
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.291474717133744e-06,
+      "loss": 0.5537,
+      "step": 14824
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.290304602446123e-06,
+      "loss": 0.6274,
+      "step": 14825
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.289134497524533e-06,
+      "loss": 0.6271,
+      "step": 14826
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.287964402385077e-06,
+      "loss": 0.6093,
+      "step": 14827
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.286794317043855e-06,
+      "loss": 0.6568,
+      "step": 14828
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.285624241516973e-06,
+      "loss": 0.5503,
+      "step": 14829
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.284454175820526e-06,
+      "loss": 0.6117,
+      "step": 14830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.28328411997062e-06,
+      "loss": 0.614,
+      "step": 14831
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.282114073983353e-06,
+      "loss": 0.6137,
+      "step": 14832
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.280944037874826e-06,
+      "loss": 0.6086,
+      "step": 14833
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.279774011661144e-06,
+      "loss": 0.5855,
+      "step": 14834
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.278603995358402e-06,
+      "loss": 0.6119,
+      "step": 14835
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.277433988982705e-06,
+      "loss": 0.6249,
+      "step": 14836
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.27626399255015e-06,
+      "loss": 0.6097,
+      "step": 14837
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.27509400607684e-06,
+      "loss": 0.6085,
+      "step": 14838
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.273924029578872e-06,
+      "loss": 0.636,
+      "step": 14839
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.272754063072348e-06,
+      "loss": 0.5724,
+      "step": 14840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.271584106573364e-06,
+      "loss": 0.5932,
+      "step": 14841
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.270414160098028e-06,
+      "loss": 0.6387,
+      "step": 14842
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.26924422366243e-06,
+      "loss": 0.5638,
+      "step": 14843
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.268074297282675e-06,
+      "loss": 0.5938,
+      "step": 14844
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.26690438097486e-06,
+      "loss": 0.6453,
+      "step": 14845
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.265734474755085e-06,
+      "loss": 0.6031,
+      "step": 14846
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.26456457863945e-06,
+      "loss": 0.5439,
+      "step": 14847
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.263394692644052e-06,
+      "loss": 0.6362,
+      "step": 14848
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.262224816784992e-06,
+      "loss": 0.6518,
+      "step": 14849
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.261054951078364e-06,
+      "loss": 0.6061,
+      "step": 14850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.259885095540271e-06,
+      "loss": 0.5774,
+      "step": 14851
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25871525018681e-06,
+      "loss": 0.6546,
+      "step": 14852
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25754541503408e-06,
+      "loss": 0.6113,
+      "step": 14853
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.256375590098176e-06,
+      "loss": 0.5921,
+      "step": 14854
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.255205775395194e-06,
+      "loss": 0.6927,
+      "step": 14855
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.254035970941237e-06,
+      "loss": 0.5744,
+      "step": 14856
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2528661767524e-06,
+      "loss": 0.617,
+      "step": 14857
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.251696392844781e-06,
+      "loss": 0.625,
+      "step": 14858
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.250526619234478e-06,
+      "loss": 0.633,
+      "step": 14859
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.249356855937587e-06,
+      "loss": 0.6705,
+      "step": 14860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.248187102970206e-06,
+      "loss": 0.6202,
+      "step": 14861
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.247017360348429e-06,
+      "loss": 0.5653,
+      "step": 14862
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.245847628088358e-06,
+      "loss": 0.5544,
+      "step": 14863
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.244677906206083e-06,
+      "loss": 0.5898,
+      "step": 14864
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.243508194717703e-06,
+      "loss": 0.5735,
+      "step": 14865
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.24233849363932e-06,
+      "loss": 0.5762,
+      "step": 14866
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.24116880298702e-06,
+      "loss": 0.6319,
+      "step": 14867
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.239999122776906e-06,
+      "loss": 0.677,
+      "step": 14868
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.238829453025072e-06,
+      "loss": 0.5619,
+      "step": 14869
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.237659793747615e-06,
+      "loss": 0.6726,
+      "step": 14870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.236490144960625e-06,
+      "loss": 0.6004,
+      "step": 14871
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.235320506680206e-06,
+      "loss": 0.6523,
+      "step": 14872
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.234150878922445e-06,
+      "loss": 0.6324,
+      "step": 14873
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.232981261703442e-06,
+      "loss": 0.5699,
+      "step": 14874
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.231811655039293e-06,
+      "loss": 0.6129,
+      "step": 14875
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.230642058946086e-06,
+      "loss": 0.639,
+      "step": 14876
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.229472473439925e-06,
+      "loss": 0.5678,
+      "step": 14877
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.228302898536897e-06,
+      "loss": 0.5781,
+      "step": 14878
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.227133334253101e-06,
+      "loss": 0.6524,
+      "step": 14879
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.225963780604627e-06,
+      "loss": 0.6282,
+      "step": 14880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.224794237607575e-06,
+      "loss": 0.5654,
+      "step": 14881
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.223624705278033e-06,
+      "loss": 0.593,
+      "step": 14882
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.222455183632098e-06,
+      "loss": 0.6828,
+      "step": 14883
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.221285672685862e-06,
+      "loss": 0.5907,
+      "step": 14884
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.220116172455422e-06,
+      "loss": 0.6189,
+      "step": 14885
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.218946682956865e-06,
+      "loss": 0.6409,
+      "step": 14886
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.21777720420629e-06,
+      "loss": 0.6212,
+      "step": 14887
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.216607736219787e-06,
+      "loss": 0.6524,
+      "step": 14888
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.215438279013451e-06,
+      "loss": 0.595,
+      "step": 14889
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.214268832603375e-06,
+      "loss": 0.6162,
+      "step": 14890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.213099397005647e-06,
+      "loss": 0.5907,
+      "step": 14891
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.211929972236365e-06,
+      "loss": 0.6178,
+      "step": 14892
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.210760558311619e-06,
+      "loss": 0.5612,
+      "step": 14893
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2095911552475e-06,
+      "loss": 0.5889,
+      "step": 14894
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2084217630601e-06,
+      "loss": 0.5444,
+      "step": 14895
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.207252381765513e-06,
+      "loss": 0.6088,
+      "step": 14896
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.206083011379833e-06,
+      "loss": 0.6445,
+      "step": 14897
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.204913651919144e-06,
+      "loss": 0.5731,
+      "step": 14898
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.203744303399546e-06,
+      "loss": 0.5573,
+      "step": 14899
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.202574965837122e-06,
+      "loss": 0.6197,
+      "step": 14900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.20140563924797e-06,
+      "loss": 0.5788,
+      "step": 14901
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.200236323648176e-06,
+      "loss": 0.6298,
+      "step": 14902
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.199067019053836e-06,
+      "loss": 0.6842,
+      "step": 14903
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.197897725481035e-06,
+      "loss": 0.6358,
+      "step": 14904
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.196728442945867e-06,
+      "loss": 0.6252,
+      "step": 14905
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19555917146442e-06,
+      "loss": 0.6012,
+      "step": 14906
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.194389911052786e-06,
+      "loss": 0.5362,
+      "step": 14907
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.193220661727056e-06,
+      "loss": 0.6323,
+      "step": 14908
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.192051423503316e-06,
+      "loss": 0.5879,
+      "step": 14909
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.190882196397663e-06,
+      "loss": 0.5325,
+      "step": 14910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.189712980426178e-06,
+      "loss": 0.7277,
+      "step": 14911
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.188543775604957e-06,
+      "loss": 0.6092,
+      "step": 14912
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.187374581950083e-06,
+      "loss": 0.5667,
+      "step": 14913
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.186205399477654e-06,
+      "loss": 0.6335,
+      "step": 14914
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.185036228203749e-06,
+      "loss": 0.5682,
+      "step": 14915
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.183867068144465e-06,
+      "loss": 0.6398,
+      "step": 14916
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.182697919315884e-06,
+      "loss": 0.6107,
+      "step": 14917
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.1815287817341e-06,
+      "loss": 0.5892,
+      "step": 14918
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.180359655415198e-06,
+      "loss": 0.5806,
+      "step": 14919
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.179190540375268e-06,
+      "loss": 0.5841,
+      "step": 14920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.178021436630397e-06,
+      "loss": 0.5514,
+      "step": 14921
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.176852344196674e-06,
+      "loss": 0.6516,
+      "step": 14922
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.175683263090188e-06,
+      "loss": 0.5604,
+      "step": 14923
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.17451419332702e-06,
+      "loss": 0.6461,
+      "step": 14924
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.173345134923265e-06,
+      "loss": 0.5669,
+      "step": 14925
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.172176087895006e-06,
+      "loss": 0.5949,
+      "step": 14926
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.171007052258331e-06,
+      "loss": 0.6645,
+      "step": 14927
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16983802802933e-06,
+      "loss": 0.6453,
+      "step": 14928
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.168669015224085e-06,
+      "loss": 0.612,
+      "step": 14929
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.167500013858685e-06,
+      "loss": 0.6252,
+      "step": 14930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.166331023949217e-06,
+      "loss": 0.6044,
+      "step": 14931
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.165162045511766e-06,
+      "loss": 0.6092,
+      "step": 14932
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16399307856242e-06,
+      "loss": 0.642,
+      "step": 14933
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.162824123117263e-06,
+      "loss": 0.6463,
+      "step": 14934
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16165517919238e-06,
+      "loss": 0.5699,
+      "step": 14935
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16048624680386e-06,
+      "loss": 0.5874,
+      "step": 14936
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.159317325967787e-06,
+      "loss": 0.6281,
+      "step": 14937
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.158148416700246e-06,
+      "loss": 0.5156,
+      "step": 14938
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.156979519017324e-06,
+      "loss": 0.598,
+      "step": 14939
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.155810632935102e-06,
+      "loss": 0.6572,
+      "step": 14940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15464175846967e-06,
+      "loss": 0.6329,
+      "step": 14941
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.153472895637107e-06,
+      "loss": 0.5633,
+      "step": 14942
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.152304044453505e-06,
+      "loss": 0.6793,
+      "step": 14943
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.151135204934941e-06,
+      "loss": 0.5768,
+      "step": 14944
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.149966377097504e-06,
+      "loss": 0.5494,
+      "step": 14945
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.148797560957276e-06,
+      "loss": 0.6356,
+      "step": 14946
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.147628756530342e-06,
+      "loss": 0.6728,
+      "step": 14947
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.146459963832783e-06,
+      "loss": 0.5815,
+      "step": 14948
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.145291182880685e-06,
+      "loss": 0.6438,
+      "step": 14949
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.144122413690135e-06,
+      "loss": 0.6539,
+      "step": 14950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.14295365627721e-06,
+      "loss": 0.6363,
+      "step": 14951
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.141784910657997e-06,
+      "loss": 0.5997,
+      "step": 14952
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.140616176848576e-06,
+      "loss": 0.539,
+      "step": 14953
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.139447454865034e-06,
+      "loss": 0.578,
+      "step": 14954
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.138278744723448e-06,
+      "loss": 0.5951,
+      "step": 14955
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.137110046439906e-06,
+      "loss": 0.6274,
+      "step": 14956
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.135941360030487e-06,
+      "loss": 0.6379,
+      "step": 14957
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.134772685511273e-06,
+      "loss": 0.6526,
+      "step": 14958
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.133604022898351e-06,
+      "loss": 0.585,
+      "step": 14959
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.132435372207795e-06,
+      "loss": 0.6572,
+      "step": 14960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.131266733455692e-06,
+      "loss": 0.5671,
+      "step": 14961
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.130098106658121e-06,
+      "loss": 0.612,
+      "step": 14962
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.128929491831167e-06,
+      "loss": 0.6181,
+      "step": 14963
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.127760888990907e-06,
+      "loss": 0.6296,
+      "step": 14964
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.126592298153425e-06,
+      "loss": 0.5903,
+      "step": 14965
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.125423719334797e-06,
+      "loss": 0.679,
+      "step": 14966
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.124255152551112e-06,
+      "loss": 0.5572,
+      "step": 14967
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.123086597818442e-06,
+      "loss": 0.677,
+      "step": 14968
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.121918055152869e-06,
+      "loss": 0.5613,
+      "step": 14969
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.12074952457048e-06,
+      "loss": 0.556,
+      "step": 14970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.119581006087348e-06,
+      "loss": 0.5914,
+      "step": 14971
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.118412499719556e-06,
+      "loss": 0.5888,
+      "step": 14972
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11724400548318e-06,
+      "loss": 0.5942,
+      "step": 14973
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.116075523394305e-06,
+      "loss": 0.6086,
+      "step": 14974
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.114907053469005e-06,
+      "loss": 0.6255,
+      "step": 14975
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.113738595723364e-06,
+      "loss": 0.6198,
+      "step": 14976
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.112570150173457e-06,
+      "loss": 0.5651,
+      "step": 14977
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.111401716835365e-06,
+      "loss": 0.6345,
+      "step": 14978
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.110233295725167e-06,
+      "loss": 0.6088,
+      "step": 14979
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.109064886858938e-06,
+      "loss": 0.5891,
+      "step": 14980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.107896490252762e-06,
+      "loss": 0.6612,
+      "step": 14981
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10672810592271e-06,
+      "loss": 0.6354,
+      "step": 14982
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10555973388487e-06,
+      "loss": 0.6535,
+      "step": 14983
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.104391374155311e-06,
+      "loss": 0.5902,
+      "step": 14984
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.103223026750114e-06,
+      "loss": 0.6065,
+      "step": 14985
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.102054691685354e-06,
+      "loss": 0.6727,
+      "step": 14986
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.100886368977114e-06,
+      "loss": 0.5905,
+      "step": 14987
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.099718058641466e-06,
+      "loss": 0.649,
+      "step": 14988
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.098549760694488e-06,
+      "loss": 0.6316,
+      "step": 14989
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.097381475152258e-06,
+      "loss": 0.4864,
+      "step": 14990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.096213202030851e-06,
+      "loss": 0.5743,
+      "step": 14991
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.095044941346346e-06,
+      "loss": 0.6623,
+      "step": 14992
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.093876693114817e-06,
+      "loss": 0.5858,
+      "step": 14993
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.092708457352342e-06,
+      "loss": 0.6145,
+      "step": 14994
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.091540234074994e-06,
+      "loss": 0.6554,
+      "step": 14995
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.090372023298853e-06,
+      "loss": 0.6279,
+      "step": 14996
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.08920382503999e-06,
+      "loss": 0.6146,
+      "step": 14997
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.088035639314484e-06,
+      "loss": 0.6382,
+      "step": 14998
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.086867466138407e-06,
+      "loss": 0.5944,
+      "step": 14999
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.085699305527836e-06,
+      "loss": 0.6275,
+      "step": 15000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.084531157498848e-06,
+      "loss": 0.6219,
+      "step": 15001
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.083363022067515e-06,
+      "loss": 0.6175,
+      "step": 15002
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.082194899249914e-06,
+      "loss": 0.6451,
+      "step": 15003
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.081026789062113e-06,
+      "loss": 0.5906,
+      "step": 15004
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.079858691520196e-06,
+      "loss": 0.6447,
+      "step": 15005
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.078690606640229e-06,
+      "loss": 0.6594,
+      "step": 15006
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.07752253443829e-06,
+      "loss": 0.6319,
+      "step": 15007
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.07635447493045e-06,
+      "loss": 0.5789,
+      "step": 15008
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.075186428132785e-06,
+      "loss": 0.6925,
+      "step": 15009
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.074018394061368e-06,
+      "loss": 0.6732,
+      "step": 15010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.072850372732268e-06,
+      "loss": 0.6268,
+      "step": 15011
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.071682364161567e-06,
+      "loss": 0.5694,
+      "step": 15012
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.070514368365329e-06,
+      "loss": 0.5905,
+      "step": 15013
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.069346385359632e-06,
+      "loss": 0.6771,
+      "step": 15014
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.068178415160543e-06,
+      "loss": 0.5894,
+      "step": 15015
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.067010457784142e-06,
+      "loss": 0.5383,
+      "step": 15016
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.065842513246493e-06,
+      "loss": 0.5936,
+      "step": 15017
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.064674581563676e-06,
+      "loss": 0.5937,
+      "step": 15018
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.063506662751755e-06,
+      "loss": 0.5865,
+      "step": 15019
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.062338756826805e-06,
+      "loss": 0.63,
+      "step": 15020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.061170863804901e-06,
+      "loss": 0.5976,
+      "step": 15021
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.060002983702108e-06,
+      "loss": 0.6143,
+      "step": 15022
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.058835116534501e-06,
+      "loss": 0.628,
+      "step": 15023
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.05766726231815e-06,
+      "loss": 0.6056,
+      "step": 15024
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.056499421069125e-06,
+      "loss": 0.5711,
+      "step": 15025
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.055331592803497e-06,
+      "loss": 0.7069,
+      "step": 15026
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.054163777537337e-06,
+      "loss": 0.6923,
+      "step": 15027
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.052995975286713e-06,
+      "loss": 0.5499,
+      "step": 15028
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.051828186067699e-06,
+      "loss": 0.5887,
+      "step": 15029
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.05066040989636e-06,
+      "loss": 0.6353,
+      "step": 15030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.049492646788768e-06,
+      "loss": 0.5622,
+      "step": 15031
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.048324896760993e-06,
+      "loss": 0.6793,
+      "step": 15032
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.047157159829103e-06,
+      "loss": 0.5772,
+      "step": 15033
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.04598943600917e-06,
+      "loss": 0.6494,
+      "step": 15034
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.044821725317258e-06,
+      "loss": 0.6328,
+      "step": 15035
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.04365402776944e-06,
+      "loss": 0.5804,
+      "step": 15036
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.042486343381782e-06,
+      "loss": 0.5493,
+      "step": 15037
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.041318672170355e-06,
+      "loss": 0.5821,
+      "step": 15038
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.040151014151223e-06,
+      "loss": 0.6041,
+      "step": 15039
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.038983369340458e-06,
+      "loss": 0.6255,
+      "step": 15040
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.037815737754124e-06,
+      "loss": 0.6159,
+      "step": 15041
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.036648119408292e-06,
+      "loss": 0.5997,
+      "step": 15042
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.035480514319031e-06,
+      "loss": 0.6141,
+      "step": 15043
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.034312922502402e-06,
+      "loss": 0.5723,
+      "step": 15044
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.033145343974479e-06,
+      "loss": 0.6558,
+      "step": 15045
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.031977778751322e-06,
+      "loss": 0.5823,
+      "step": 15046
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.030810226849005e-06,
+      "loss": 0.6276,
+      "step": 15047
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.029642688283588e-06,
+      "loss": 0.681,
+      "step": 15048
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.028475163071142e-06,
+      "loss": 0.5673,
+      "step": 15049
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.02730765122773e-06,
+      "loss": 0.6236,
+      "step": 15050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.02614015276942e-06,
+      "loss": 0.5529,
+      "step": 15051
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.024972667712277e-06,
+      "loss": 0.6457,
+      "step": 15052
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.023805196072368e-06,
+      "loss": 0.6314,
+      "step": 15053
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.022637737865757e-06,
+      "loss": 0.568,
+      "step": 15054
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.021470293108507e-06,
+      "loss": 0.642,
+      "step": 15055
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.02030286181669e-06,
+      "loss": 0.5674,
+      "step": 15056
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.019135444006362e-06,
+      "loss": 0.632,
+      "step": 15057
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.017968039693595e-06,
+      "loss": 0.6168,
+      "step": 15058
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.01680064889445e-06,
+      "loss": 0.6798,
+      "step": 15059
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.015633271624993e-06,
+      "loss": 0.5887,
+      "step": 15060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.014465907901284e-06,
+      "loss": 0.5529,
+      "step": 15061
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.013298557739391e-06,
+      "loss": 0.6298,
+      "step": 15062
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.01213122115538e-06,
+      "loss": 0.5419,
+      "step": 15063
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.010963898165312e-06,
+      "loss": 0.6276,
+      "step": 15064
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.009796588785245e-06,
+      "loss": 0.6047,
+      "step": 15065
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.008629293031251e-06,
+      "loss": 0.5651,
+      "step": 15066
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.007462010919387e-06,
+      "loss": 0.576,
+      "step": 15067
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.00629474246572e-06,
+      "loss": 0.5089,
+      "step": 15068
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.005127487686306e-06,
+      "loss": 0.6283,
+      "step": 15069
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.003960246597215e-06,
+      "loss": 0.5603,
+      "step": 15070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.002793019214506e-06,
+      "loss": 0.5885,
+      "step": 15071
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.001625805554242e-06,
+      "loss": 0.5641,
+      "step": 15072
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.000458605632482e-06,
+      "loss": 0.5589,
+      "step": 15073
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.999291419465289e-06,
+      "loss": 0.5703,
+      "step": 15074
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.99812424706873e-06,
+      "loss": 0.5461,
+      "step": 15075
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.996957088458857e-06,
+      "loss": 0.5891,
+      "step": 15076
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.995789943651738e-06,
+      "loss": 0.6647,
+      "step": 15077
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.99462281266343e-06,
+      "loss": 0.5495,
+      "step": 15078
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.99345569551e-06,
+      "loss": 0.5982,
+      "step": 15079
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.992288592207499e-06,
+      "loss": 0.5873,
+      "step": 15080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.991121502771995e-06,
+      "loss": 0.6257,
+      "step": 15081
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.989954427219545e-06,
+      "loss": 0.5889,
+      "step": 15082
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.988787365566209e-06,
+      "loss": 0.553,
+      "step": 15083
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.98762031782805e-06,
+      "loss": 0.6471,
+      "step": 15084
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.98645328402112e-06,
+      "loss": 0.6554,
+      "step": 15085
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.985286264161487e-06,
+      "loss": 0.5872,
+      "step": 15086
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.984119258265205e-06,
+      "loss": 0.7037,
+      "step": 15087
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.982952266348339e-06,
+      "loss": 0.6112,
+      "step": 15088
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.981785288426938e-06,
+      "loss": 0.6655,
+      "step": 15089
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.98061832451707e-06,
+      "loss": 0.6065,
+      "step": 15090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.979451374634787e-06,
+      "loss": 0.6406,
+      "step": 15091
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.978284438796151e-06,
+      "loss": 0.5675,
+      "step": 15092
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.977117517017218e-06,
+      "loss": 0.5893,
+      "step": 15093
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.975950609314045e-06,
+      "loss": 0.5905,
+      "step": 15094
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.974783715702695e-06,
+      "loss": 0.593,
+      "step": 15095
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.97361683619922e-06,
+      "loss": 0.5846,
+      "step": 15096
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.972449970819682e-06,
+      "loss": 0.5894,
+      "step": 15097
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.971283119580132e-06,
+      "loss": 0.6125,
+      "step": 15098
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.970116282496633e-06,
+      "loss": 0.5689,
+      "step": 15099
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.968949459585236e-06,
+      "loss": 0.5424,
+      "step": 15100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.967782650862002e-06,
+      "loss": 0.6382,
+      "step": 15101
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.966615856342984e-06,
+      "loss": 0.621,
+      "step": 15102
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.965449076044243e-06,
+      "loss": 0.6605,
+      "step": 15103
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.964282309981828e-06,
+      "loss": 0.6187,
+      "step": 15104
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.9631155581718e-06,
+      "loss": 0.6235,
+      "step": 15105
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.961948820630215e-06,
+      "loss": 0.6631,
+      "step": 15106
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.960782097373124e-06,
+      "loss": 0.5658,
+      "step": 15107
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.959615388416584e-06,
+      "loss": 0.5487,
+      "step": 15108
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.95844869377665e-06,
+      "loss": 0.6579,
+      "step": 15109
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.957282013469381e-06,
+      "loss": 0.6442,
+      "step": 15110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.956115347510825e-06,
+      "loss": 0.5805,
+      "step": 15111
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.95494869591704e-06,
+      "loss": 0.5793,
+      "step": 15112
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.953782058704079e-06,
+      "loss": 0.6432,
+      "step": 15113
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.952615435887994e-06,
+      "loss": 0.5798,
+      "step": 15114
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.951448827484843e-06,
+      "loss": 0.5748,
+      "step": 15115
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.950282233510677e-06,
+      "loss": 0.6499,
+      "step": 15116
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.949115653981551e-06,
+      "loss": 0.593,
+      "step": 15117
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.947949088913516e-06,
+      "loss": 0.6472,
+      "step": 15118
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.946782538322628e-06,
+      "loss": 0.6303,
+      "step": 15119
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.945616002224934e-06,
+      "loss": 0.6412,
+      "step": 15120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.944449480636492e-06,
+      "loss": 0.5941,
+      "step": 15121
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.943282973573351e-06,
+      "loss": 0.6489,
+      "step": 15122
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.942116481051569e-06,
+      "loss": 0.5638,
+      "step": 15123
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.94095000308719e-06,
+      "loss": 0.5605,
+      "step": 15124
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.939783539696269e-06,
+      "loss": 0.5368,
+      "step": 15125
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.938617090894861e-06,
+      "loss": 0.5874,
+      "step": 15126
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.937450656699013e-06,
+      "loss": 0.6276,
+      "step": 15127
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.936284237124779e-06,
+      "loss": 0.641,
+      "step": 15128
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.935117832188205e-06,
+      "loss": 0.535,
+      "step": 15129
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.93395144190535e-06,
+      "loss": 0.6451,
+      "step": 15130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.932785066292257e-06,
+      "loss": 0.6481,
+      "step": 15131
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.931618705364981e-06,
+      "loss": 0.6489,
+      "step": 15132
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.930452359139568e-06,
+      "loss": 0.5672,
+      "step": 15133
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.92928602763207e-06,
+      "loss": 0.6709,
+      "step": 15134
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.928119710858539e-06,
+      "loss": 0.594,
+      "step": 15135
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.926953408835022e-06,
+      "loss": 0.5862,
+      "step": 15136
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.92578712157757e-06,
+      "loss": 0.6257,
+      "step": 15137
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.924620849102229e-06,
+      "loss": 0.5667,
+      "step": 15138
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.923454591425052e-06,
+      "loss": 0.5506,
+      "step": 15139
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.922288348562084e-06,
+      "loss": 0.7127,
+      "step": 15140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.921122120529377e-06,
+      "loss": 0.6527,
+      "step": 15141
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.919955907342977e-06,
+      "loss": 0.6123,
+      "step": 15142
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.918789709018934e-06,
+      "loss": 0.6537,
+      "step": 15143
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.917623525573292e-06,
+      "loss": 0.5536,
+      "step": 15144
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.916457357022101e-06,
+      "loss": 0.6245,
+      "step": 15145
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.915291203381413e-06,
+      "loss": 0.5596,
+      "step": 15146
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.914125064667267e-06,
+      "loss": 0.6325,
+      "step": 15147
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.912958940895717e-06,
+      "loss": 0.6236,
+      "step": 15148
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.911792832082805e-06,
+      "loss": 0.7031,
+      "step": 15149
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.910626738244583e-06,
+      "loss": 0.6391,
+      "step": 15150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90946065939709e-06,
+      "loss": 0.6495,
+      "step": 15151
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90829459555638e-06,
+      "loss": 0.6024,
+      "step": 15152
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.907128546738493e-06,
+      "loss": 0.5963,
+      "step": 15153
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90596251295948e-06,
+      "loss": 0.6091,
+      "step": 15154
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.904796494235382e-06,
+      "loss": 0.5483,
+      "step": 15155
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.903630490582246e-06,
+      "loss": 0.5632,
+      "step": 15156
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90246450201612e-06,
+      "loss": 0.6076,
+      "step": 15157
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.901298528553045e-06,
+      "loss": 0.6702,
+      "step": 15158
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90013257020907e-06,
+      "loss": 0.6238,
+      "step": 15159
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.898966627000235e-06,
+      "loss": 0.563,
+      "step": 15160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.897800698942589e-06,
+      "loss": 0.5461,
+      "step": 15161
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.896634786052172e-06,
+      "loss": 0.6355,
+      "step": 15162
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.895468888345032e-06,
+      "loss": 0.6313,
+      "step": 15163
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.894303005837208e-06,
+      "loss": 0.6561,
+      "step": 15164
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.893137138544747e-06,
+      "loss": 0.67,
+      "step": 15165
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.891971286483691e-06,
+      "loss": 0.65,
+      "step": 15166
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.890805449670083e-06,
+      "loss": 0.5798,
+      "step": 15167
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.889639628119971e-06,
+      "loss": 0.6053,
+      "step": 15168
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.888473821849388e-06,
+      "loss": 0.6154,
+      "step": 15169
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.887308030874387e-06,
+      "loss": 0.6538,
+      "step": 15170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.886142255211e-06,
+      "loss": 0.6198,
+      "step": 15171
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.884976494875277e-06,
+      "loss": 0.573,
+      "step": 15172
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.883810749883257e-06,
+      "loss": 0.5249,
+      "step": 15173
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.882645020250982e-06,
+      "loss": 0.5896,
+      "step": 15174
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.881479305994493e-06,
+      "loss": 0.5917,
+      "step": 15175
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.880313607129831e-06,
+      "loss": 0.5998,
+      "step": 15176
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.879147923673039e-06,
+      "loss": 0.5983,
+      "step": 15177
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.877982255640155e-06,
+      "loss": 0.6297,
+      "step": 15178
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.876816603047224e-06,
+      "loss": 0.6252,
+      "step": 15179
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.87565096591028e-06,
+      "loss": 0.5849,
+      "step": 15180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.874485344245369e-06,
+      "loss": 0.6825,
+      "step": 15181
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.873319738068527e-06,
+      "loss": 0.6133,
+      "step": 15182
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.872154147395797e-06,
+      "loss": 0.5344,
+      "step": 15183
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.870988572243214e-06,
+      "loss": 0.6474,
+      "step": 15184
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.869823012626824e-06,
+      "loss": 0.59,
+      "step": 15185
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.86865746856266e-06,
+      "loss": 0.6042,
+      "step": 15186
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.867491940066762e-06,
+      "loss": 0.6054,
+      "step": 15187
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.866326427155173e-06,
+      "loss": 0.6535,
+      "step": 15188
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.865160929843926e-06,
+      "loss": 0.6524,
+      "step": 15189
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.863995448149064e-06,
+      "loss": 0.6485,
+      "step": 15190
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.862829982086622e-06,
+      "loss": 0.6377,
+      "step": 15191
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.861664531672639e-06,
+      "loss": 0.5743,
+      "step": 15192
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.860499096923151e-06,
+      "loss": 0.6358,
+      "step": 15193
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.859333677854198e-06,
+      "loss": 0.5879,
+      "step": 15194
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.858168274481816e-06,
+      "loss": 0.5748,
+      "step": 15195
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.857002886822041e-06,
+      "loss": 0.5819,
+      "step": 15196
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85583751489091e-06,
+      "loss": 0.6074,
+      "step": 15197
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85467215870446e-06,
+      "loss": 0.6195,
+      "step": 15198
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85350681827873e-06,
+      "loss": 0.5571,
+      "step": 15199
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.852341493629752e-06,
+      "loss": 0.6109,
+      "step": 15200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.851176184773563e-06,
+      "loss": 0.6771,
+      "step": 15201
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8500108917262e-06,
+      "loss": 0.7084,
+      "step": 15202
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.848845614503698e-06,
+      "loss": 0.6113,
+      "step": 15203
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84768035312209e-06,
+      "loss": 0.5745,
+      "step": 15204
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.846515107597415e-06,
+      "loss": 0.6136,
+      "step": 15205
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.845349877945703e-06,
+      "loss": 0.6547,
+      "step": 15206
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.844184664182993e-06,
+      "loss": 0.5694,
+      "step": 15207
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.843019466325317e-06,
+      "loss": 0.6386,
+      "step": 15208
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84185428438871e-06,
+      "loss": 0.6308,
+      "step": 15209
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.840689118389207e-06,
+      "loss": 0.6252,
+      "step": 15210
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83952396834284e-06,
+      "loss": 0.6315,
+      "step": 15211
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.838358834265642e-06,
+      "loss": 0.6106,
+      "step": 15212
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.837193716173646e-06,
+      "loss": 0.53,
+      "step": 15213
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83602861408289e-06,
+      "loss": 0.6022,
+      "step": 15214
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.834863528009398e-06,
+      "loss": 0.5498,
+      "step": 15215
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.833698457969213e-06,
+      "loss": 0.5844,
+      "step": 15216
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.832533403978357e-06,
+      "loss": 0.5915,
+      "step": 15217
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83136836605287e-06,
+      "loss": 0.5786,
+      "step": 15218
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.830203344208781e-06,
+      "loss": 0.5973,
+      "step": 15219
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.82903833846212e-06,
+      "loss": 0.5955,
+      "step": 15220
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.827873348828922e-06,
+      "loss": 0.5866,
+      "step": 15221
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.826708375325215e-06,
+      "loss": 0.6069,
+      "step": 15222
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.825543417967035e-06,
+      "loss": 0.5575,
+      "step": 15223
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.824378476770407e-06,
+      "loss": 0.6509,
+      "step": 15224
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.823213551751366e-06,
+      "loss": 0.5497,
+      "step": 15225
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.822048642925938e-06,
+      "loss": 0.5654,
+      "step": 15226
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.820883750310159e-06,
+      "loss": 0.6152,
+      "step": 15227
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.819718873920052e-06,
+      "loss": 0.6258,
+      "step": 15228
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.818554013771653e-06,
+      "loss": 0.5847,
+      "step": 15229
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.817389169880988e-06,
+      "loss": 0.6383,
+      "step": 15230
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.816224342264085e-06,
+      "loss": 0.6235,
+      "step": 15231
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.81505953093698e-06,
+      "loss": 0.601,
+      "step": 15232
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.813894735915693e-06,
+      "loss": 0.5487,
+      "step": 15233
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.812729957216259e-06,
+      "loss": 0.6413,
+      "step": 15234
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.811565194854701e-06,
+      "loss": 0.5643,
+      "step": 15235
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.810400448847054e-06,
+      "loss": 0.616,
+      "step": 15236
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.809235719209339e-06,
+      "loss": 0.6186,
+      "step": 15237
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.808071005957586e-06,
+      "loss": 0.5947,
+      "step": 15238
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.806906309107828e-06,
+      "loss": 0.574,
+      "step": 15239
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.805741628676083e-06,
+      "loss": 0.6694,
+      "step": 15240
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.804576964678386e-06,
+      "loss": 0.5916,
+      "step": 15241
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.803412317130759e-06,
+      "loss": 0.5965,
+      "step": 15242
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80224768604923e-06,
+      "loss": 0.5713,
+      "step": 15243
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.801083071449824e-06,
+      "loss": 0.6227,
+      "step": 15244
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.79991847334857e-06,
+      "loss": 0.6136,
+      "step": 15245
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.798753891761492e-06,
+      "loss": 0.6869,
+      "step": 15246
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.797589326704616e-06,
+      "loss": 0.5416,
+      "step": 15247
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.796424778193966e-06,
+      "loss": 0.5185,
+      "step": 15248
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.795260246245569e-06,
+      "loss": 0.6204,
+      "step": 15249
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.794095730875452e-06,
+      "loss": 0.609,
+      "step": 15250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.792931232099633e-06,
+      "loss": 0.6129,
+      "step": 15251
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.791766749934145e-06,
+      "loss": 0.6706,
+      "step": 15252
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.790602284395006e-06,
+      "loss": 0.5572,
+      "step": 15253
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.789437835498243e-06,
+      "loss": 0.6155,
+      "step": 15254
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.788273403259879e-06,
+      "loss": 0.5908,
+      "step": 15255
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.787108987695937e-06,
+      "loss": 0.5486,
+      "step": 15256
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.785944588822443e-06,
+      "loss": 0.6722,
+      "step": 15257
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.784780206655419e-06,
+      "loss": 0.6229,
+      "step": 15258
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.783615841210883e-06,
+      "loss": 0.6254,
+      "step": 15259
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.782451492504861e-06,
+      "loss": 0.5987,
+      "step": 15260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.781287160553382e-06,
+      "loss": 0.5807,
+      "step": 15261
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.780122845372459e-06,
+      "loss": 0.618,
+      "step": 15262
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77895854697812e-06,
+      "loss": 0.6155,
+      "step": 15263
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77779426538638e-06,
+      "loss": 0.6106,
+      "step": 15264
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77663000061327e-06,
+      "loss": 0.6096,
+      "step": 15265
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.775465752674802e-06,
+      "loss": 0.5406,
+      "step": 15266
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.774301521587003e-06,
+      "loss": 0.6598,
+      "step": 15267
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.773137307365892e-06,
+      "loss": 0.633,
+      "step": 15268
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.771973110027487e-06,
+      "loss": 0.6106,
+      "step": 15269
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.770808929587815e-06,
+      "loss": 0.7553,
+      "step": 15270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.76964476606289e-06,
+      "loss": 0.5839,
+      "step": 15271
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.768480619468736e-06,
+      "loss": 0.5667,
+      "step": 15272
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.767316489821368e-06,
+      "loss": 0.5929,
+      "step": 15273
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.766152377136813e-06,
+      "loss": 0.5999,
+      "step": 15274
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.764988281431083e-06,
+      "loss": 0.5812,
+      "step": 15275
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.763824202720198e-06,
+      "loss": 0.6266,
+      "step": 15276
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.76266014102018e-06,
+      "loss": 0.5148,
+      "step": 15277
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.761496096347045e-06,
+      "loss": 0.6427,
+      "step": 15278
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.76033206871681e-06,
+      "loss": 0.6332,
+      "step": 15279
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.759168058145496e-06,
+      "loss": 0.6361,
+      "step": 15280
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75800406464912e-06,
+      "loss": 0.6046,
+      "step": 15281
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.756840088243701e-06,
+      "loss": 0.5744,
+      "step": 15282
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.755676128945251e-06,
+      "loss": 0.6044,
+      "step": 15283
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.754512186769796e-06,
+      "loss": 0.6167,
+      "step": 15284
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.753348261733343e-06,
+      "loss": 0.5996,
+      "step": 15285
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.752184353851917e-06,
+      "loss": 0.5567,
+      "step": 15286
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.751020463141528e-06,
+      "loss": 0.5839,
+      "step": 15287
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.749856589618197e-06,
+      "loss": 0.5878,
+      "step": 15288
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.748692733297935e-06,
+      "loss": 0.6487,
+      "step": 15289
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.747528894196764e-06,
+      "loss": 0.6499,
+      "step": 15290
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.746365072330692e-06,
+      "loss": 0.5712,
+      "step": 15291
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.745201267715739e-06,
+      "loss": 0.6251,
+      "step": 15292
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.744037480367922e-06,
+      "loss": 0.5839,
+      "step": 15293
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.74287371030325e-06,
+      "loss": 0.6453,
+      "step": 15294
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.741709957537742e-06,
+      "loss": 0.5711,
+      "step": 15295
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.74054622208741e-06,
+      "loss": 0.656,
+      "step": 15296
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.73938250396827e-06,
+      "loss": 0.6129,
+      "step": 15297
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.738218803196332e-06,
+      "loss": 0.5681,
+      "step": 15298
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.737055119787615e-06,
+      "loss": 0.6043,
+      "step": 15299
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.735891453758126e-06,
+      "loss": 0.5597,
+      "step": 15300
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.734727805123883e-06,
+      "loss": 0.6037,
+      "step": 15301
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.733564173900898e-06,
+      "loss": 0.616,
+      "step": 15302
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.73240056010518e-06,
+      "loss": 0.5707,
+      "step": 15303
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.731236963752746e-06,
+      "loss": 0.63,
+      "step": 15304
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.730073384859605e-06,
+      "loss": 0.5972,
+      "step": 15305
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.728909823441772e-06,
+      "loss": 0.6238,
+      "step": 15306
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.727746279515255e-06,
+      "loss": 0.6666,
+      "step": 15307
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72658275309607e-06,
+      "loss": 0.5921,
+      "step": 15308
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.725419244200222e-06,
+      "loss": 0.6808,
+      "step": 15309
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.724255752843726e-06,
+      "loss": 0.6387,
+      "step": 15310
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.723092279042591e-06,
+      "loss": 0.6216,
+      "step": 15311
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72192882281283e-06,
+      "loss": 0.592,
+      "step": 15312
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72076538417045e-06,
+      "loss": 0.6426,
+      "step": 15313
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.719601963131464e-06,
+      "loss": 0.5714,
+      "step": 15314
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.718438559711879e-06,
+      "loss": 0.6144,
+      "step": 15315
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.717275173927704e-06,
+      "loss": 0.6054,
+      "step": 15316
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.716111805794953e-06,
+      "loss": 0.6182,
+      "step": 15317
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.71494845532963e-06,
+      "loss": 0.543,
+      "step": 15318
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.713785122547746e-06,
+      "loss": 0.6069,
+      "step": 15319
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.712621807465307e-06,
+      "loss": 0.6758,
+      "step": 15320
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.711458510098326e-06,
+      "loss": 0.6338,
+      "step": 15321
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.710295230462805e-06,
+      "loss": 0.611,
+      "step": 15322
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.709131968574756e-06,
+      "loss": 0.6536,
+      "step": 15323
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.707968724450186e-06,
+      "loss": 0.5955,
+      "step": 15324
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.7068054981051e-06,
+      "loss": 0.6492,
+      "step": 15325
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.705642289555511e-06,
+      "loss": 0.6021,
+      "step": 15326
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.704479098817417e-06,
+      "loss": 0.6385,
+      "step": 15327
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.70331592590683e-06,
+      "loss": 0.6659,
+      "step": 15328
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.702152770839755e-06,
+      "loss": 0.5907,
+      "step": 15329
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.7009896336322e-06,
+      "loss": 0.5846,
+      "step": 15330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.699826514300168e-06,
+      "loss": 0.5645,
+      "step": 15331
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.698663412859667e-06,
+      "loss": 0.6243,
+      "step": 15332
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.697500329326697e-06,
+      "loss": 0.558,
+      "step": 15333
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.696337263717269e-06,
+      "loss": 0.6869,
+      "step": 15334
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.695174216047388e-06,
+      "loss": 0.6362,
+      "step": 15335
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.694011186333052e-06,
+      "loss": 0.6212,
+      "step": 15336
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.692848174590274e-06,
+      "loss": 0.6034,
+      "step": 15337
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.69168518083505e-06,
+      "loss": 0.6807,
+      "step": 15338
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.69052220508339e-06,
+      "loss": 0.6488,
+      "step": 15339
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.689359247351294e-06,
+      "loss": 0.5745,
+      "step": 15340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.688196307654769e-06,
+      "loss": 0.6434,
+      "step": 15341
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.687033386009811e-06,
+      "loss": 0.5877,
+      "step": 15342
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.685870482432429e-06,
+      "loss": 0.5445,
+      "step": 15343
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.684707596938624e-06,
+      "loss": 0.6013,
+      "step": 15344
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.683544729544398e-06,
+      "loss": 0.6565,
+      "step": 15345
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.682381880265756e-06,
+      "loss": 0.6191,
+      "step": 15346
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.681219049118691e-06,
+      "loss": 0.6488,
+      "step": 15347
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.680056236119218e-06,
+      "loss": 0.6326,
+      "step": 15348
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.678893441283325e-06,
+      "loss": 0.6704,
+      "step": 15349
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.677730664627023e-06,
+      "loss": 0.5776,
+      "step": 15350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.676567906166307e-06,
+      "loss": 0.6555,
+      "step": 15351
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.675405165917182e-06,
+      "loss": 0.6175,
+      "step": 15352
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.674242443895641e-06,
+      "loss": 0.6225,
+      "step": 15353
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.673079740117692e-06,
+      "loss": 0.6514,
+      "step": 15354
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.671917054599331e-06,
+      "loss": 0.6356,
+      "step": 15355
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.670754387356558e-06,
+      "loss": 0.6678,
+      "step": 15356
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.669591738405374e-06,
+      "loss": 0.594,
+      "step": 15357
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.668429107761775e-06,
+      "loss": 0.5998,
+      "step": 15358
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.667266495441764e-06,
+      "loss": 0.5593,
+      "step": 15359
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.666103901461333e-06,
+      "loss": 0.5418,
+      "step": 15360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.66494132583649e-06,
+      "loss": 0.5571,
+      "step": 15361
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.663778768583222e-06,
+      "loss": 0.6067,
+      "step": 15362
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.662616229717537e-06,
+      "loss": 0.6588,
+      "step": 15363
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.661453709255423e-06,
+      "loss": 0.6453,
+      "step": 15364
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.660291207212883e-06,
+      "loss": 0.5669,
+      "step": 15365
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.659128723605916e-06,
+      "loss": 0.6385,
+      "step": 15366
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.657966258450514e-06,
+      "loss": 0.6026,
+      "step": 15367
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.656803811762678e-06,
+      "loss": 0.6201,
+      "step": 15368
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.655641383558397e-06,
+      "loss": 0.5922,
+      "step": 15369
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.654478973853678e-06,
+      "loss": 0.6349,
+      "step": 15370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.653316582664504e-06,
+      "loss": 0.613,
+      "step": 15371
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.652154210006882e-06,
+      "loss": 0.5459,
+      "step": 15372
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.650991855896799e-06,
+      "loss": 0.6294,
+      "step": 15373
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.649829520350255e-06,
+      "loss": 0.6333,
+      "step": 15374
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.648667203383245e-06,
+      "loss": 0.5974,
+      "step": 15375
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.647504905011759e-06,
+      "loss": 0.5612,
+      "step": 15376
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.646342625251795e-06,
+      "loss": 0.6156,
+      "step": 15377
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.645180364119345e-06,
+      "loss": 0.6246,
+      "step": 15378
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.644018121630406e-06,
+      "loss": 0.5989,
+      "step": 15379
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.642855897800967e-06,
+      "loss": 0.6377,
+      "step": 15380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.641693692647025e-06,
+      "loss": 0.5712,
+      "step": 15381
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.640531506184568e-06,
+      "loss": 0.5879,
+      "step": 15382
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.639369338429596e-06,
+      "loss": 0.6145,
+      "step": 15383
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.638207189398096e-06,
+      "loss": 0.5623,
+      "step": 15384
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.63704505910606e-06,
+      "loss": 0.6833,
+      "step": 15385
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.635882947569484e-06,
+      "loss": 0.6555,
+      "step": 15386
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.634720854804354e-06,
+      "loss": 0.6081,
+      "step": 15387
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.633558780826668e-06,
+      "loss": 0.5606,
+      "step": 15388
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.632396725652412e-06,
+      "loss": 0.5359,
+      "step": 15389
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.63123468929758e-06,
+      "loss": 0.6419,
+      "step": 15390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.63007267177816e-06,
+      "loss": 0.5792,
+      "step": 15391
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.628910673110146e-06,
+      "loss": 0.6804,
+      "step": 15392
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.627748693309524e-06,
+      "loss": 0.5996,
+      "step": 15393
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.626586732392287e-06,
+      "loss": 0.5437,
+      "step": 15394
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.625424790374422e-06,
+      "loss": 0.6264,
+      "step": 15395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62426286727192e-06,
+      "loss": 0.6628,
+      "step": 15396
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62310096310077e-06,
+      "loss": 0.5905,
+      "step": 15397
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62193907787696e-06,
+      "loss": 0.5624,
+      "step": 15398
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62077721161648e-06,
+      "loss": 0.5643,
+      "step": 15399
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.619615364335316e-06,
+      "loss": 0.557,
+      "step": 15400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61845353604946e-06,
+      "loss": 0.6691,
+      "step": 15401
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.617291726774896e-06,
+      "loss": 0.6,
+      "step": 15402
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.616129936527612e-06,
+      "loss": 0.5935,
+      "step": 15403
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.614968165323598e-06,
+      "loss": 0.6556,
+      "step": 15404
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.613806413178836e-06,
+      "loss": 0.6466,
+      "step": 15405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61264468010932e-06,
+      "loss": 0.6683,
+      "step": 15406
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.611482966131027e-06,
+      "loss": 0.6633,
+      "step": 15407
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.610321271259954e-06,
+      "loss": 0.5906,
+      "step": 15408
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.609159595512077e-06,
+      "loss": 0.5294,
+      "step": 15409
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.607997938903389e-06,
+      "loss": 0.6929,
+      "step": 15410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.60683630144987e-06,
+      "loss": 0.5808,
+      "step": 15411
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.60567468316751e-06,
+      "loss": 0.6793,
+      "step": 15412
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.604513084072288e-06,
+      "loss": 0.5623,
+      "step": 15413
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.603351504180198e-06,
+      "loss": 0.5769,
+      "step": 15414
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.602189943507214e-06,
+      "loss": 0.5891,
+      "step": 15415
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.601028402069324e-06,
+      "loss": 0.5925,
+      "step": 15416
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.599866879882515e-06,
+      "loss": 0.5557,
+      "step": 15417
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.598705376962765e-06,
+      "loss": 0.6249,
+      "step": 15418
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.597543893326065e-06,
+      "loss": 0.607,
+      "step": 15419
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.59638242898839e-06,
+      "loss": 0.5363,
+      "step": 15420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.595220983965727e-06,
+      "loss": 0.63,
+      "step": 15421
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.594059558274056e-06,
+      "loss": 0.6416,
+      "step": 15422
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.592898151929364e-06,
+      "loss": 0.5913,
+      "step": 15423
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.591736764947628e-06,
+      "loss": 0.6241,
+      "step": 15424
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.590575397344831e-06,
+      "loss": 0.5408,
+      "step": 15425
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.589414049136955e-06,
+      "loss": 0.5747,
+      "step": 15426
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.58825272033998e-06,
+      "loss": 0.5768,
+      "step": 15427
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.587091410969893e-06,
+      "loss": 0.5514,
+      "step": 15428
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.585930121042666e-06,
+      "loss": 0.6799,
+      "step": 15429
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.584768850574284e-06,
+      "loss": 0.5584,
+      "step": 15430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.583607599580723e-06,
+      "loss": 0.5686,
+      "step": 15431
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.582446368077971e-06,
+      "loss": 0.6257,
+      "step": 15432
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.581285156081998e-06,
+      "loss": 0.6082,
+      "step": 15433
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.580123963608791e-06,
+      "loss": 0.5952,
+      "step": 15434
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.578962790674324e-06,
+      "loss": 0.6356,
+      "step": 15435
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.577801637294578e-06,
+      "loss": 0.5643,
+      "step": 15436
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.576640503485533e-06,
+      "loss": 0.6387,
+      "step": 15437
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.575479389263163e-06,
+      "loss": 0.5844,
+      "step": 15438
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.57431829464345e-06,
+      "loss": 0.5623,
+      "step": 15439
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.573157219642369e-06,
+      "loss": 0.6033,
+      "step": 15440
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.5719961642759e-06,
+      "loss": 0.5907,
+      "step": 15441
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.570835128560016e-06,
+      "loss": 0.6275,
+      "step": 15442
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.569674112510698e-06,
+      "loss": 0.5997,
+      "step": 15443
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.56851311614392e-06,
+      "loss": 0.6503,
+      "step": 15444
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.567352139475661e-06,
+      "loss": 0.6355,
+      "step": 15445
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.566191182521893e-06,
+      "loss": 0.551,
+      "step": 15446
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.565030245298594e-06,
+      "loss": 0.6143,
+      "step": 15447
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.563869327821744e-06,
+      "loss": 0.5392,
+      "step": 15448
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.562708430107309e-06,
+      "loss": 0.6547,
+      "step": 15449
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.561547552171272e-06,
+      "loss": 0.6307,
+      "step": 15450
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.560386694029603e-06,
+      "loss": 0.6159,
+      "step": 15451
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.55922585569828e-06,
+      "loss": 0.6219,
+      "step": 15452
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.558065037193272e-06,
+      "loss": 0.5593,
+      "step": 15453
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.55690423853056e-06,
+      "loss": 0.5889,
+      "step": 15454
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.55574345972611e-06,
+      "loss": 0.5865,
+      "step": 15455
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.554582700795902e-06,
+      "loss": 0.5576,
+      "step": 15456
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.553421961755904e-06,
+      "loss": 0.6717,
+      "step": 15457
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.55226124262209e-06,
+      "loss": 0.5908,
+      "step": 15458
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.551100543410437e-06,
+      "loss": 0.5994,
+      "step": 15459
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.549939864136912e-06,
+      "loss": 0.6925,
+      "step": 15460
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.548779204817487e-06,
+      "loss": 0.5849,
+      "step": 15461
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.547618565468137e-06,
+      "loss": 0.6156,
+      "step": 15462
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.546457946104832e-06,
+      "loss": 0.6675,
+      "step": 15463
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.545297346743541e-06,
+      "loss": 0.5252,
+      "step": 15464
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.54413676740024e-06,
+      "loss": 0.6056,
+      "step": 15465
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.542976208090893e-06,
+      "loss": 0.6487,
+      "step": 15466
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.541815668831474e-06,
+      "loss": 0.6009,
+      "step": 15467
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.540655149637955e-06,
+      "loss": 0.634,
+      "step": 15468
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.5394946505263e-06,
+      "loss": 0.6322,
+      "step": 15469
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.538334171512485e-06,
+      "loss": 0.6039,
+      "step": 15470
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.537173712612474e-06,
+      "loss": 0.5951,
+      "step": 15471
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.536013273842242e-06,
+      "loss": 0.6698,
+      "step": 15472
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.534852855217748e-06,
+      "loss": 0.6453,
+      "step": 15473
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.53369245675497e-06,
+      "loss": 0.6,
+      "step": 15474
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.53253207846987e-06,
+      "loss": 0.5941,
+      "step": 15475
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.531371720378421e-06,
+      "loss": 0.5928,
+      "step": 15476
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.530211382496584e-06,
+      "loss": 0.6365,
+      "step": 15477
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.529051064840332e-06,
+      "loss": 0.6051,
+      "step": 15478
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.52789076742563e-06,
+      "loss": 0.576,
+      "step": 15479
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.526730490268443e-06,
+      "loss": 0.5723,
+      "step": 15480
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.525570233384741e-06,
+      "loss": 0.6845,
+      "step": 15481
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.524409996790485e-06,
+      "loss": 0.675,
+      "step": 15482
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.523249780501647e-06,
+      "loss": 0.6027,
+      "step": 15483
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.522089584534189e-06,
+      "loss": 0.5728,
+      "step": 15484
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.520929408904079e-06,
+      "loss": 0.6045,
+      "step": 15485
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.51976925362728e-06,
+      "loss": 0.5209,
+      "step": 15486
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.518609118719753e-06,
+      "loss": 0.5757,
+      "step": 15487
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.517449004197468e-06,
+      "loss": 0.5766,
+      "step": 15488
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.516288910076385e-06,
+      "loss": 0.569,
+      "step": 15489
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.515128836372473e-06,
+      "loss": 0.5607,
+      "step": 15490
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.513968783101693e-06,
+      "loss": 0.6145,
+      "step": 15491
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.512808750280007e-06,
+      "loss": 0.5909,
+      "step": 15492
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.511648737923379e-06,
+      "loss": 0.5846,
+      "step": 15493
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.510488746047773e-06,
+      "loss": 0.6294,
+      "step": 15494
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.50932877466915e-06,
+      "loss": 0.6559,
+      "step": 15495
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.508168823803474e-06,
+      "loss": 0.6028,
+      "step": 15496
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.507008893466706e-06,
+      "loss": 0.6085,
+      "step": 15497
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.505848983674805e-06,
+      "loss": 0.6414,
+      "step": 15498
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.504689094443737e-06,
+      "loss": 0.603,
+      "step": 15499
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.503529225789461e-06,
+      "loss": 0.5598,
+      "step": 15500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.502369377727936e-06,
+      "loss": 0.6809,
+      "step": 15501
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.501209550275127e-06,
+      "loss": 0.5933,
+      "step": 15502
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.50004974344699e-06,
+      "loss": 0.5946,
+      "step": 15503
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.498889957259487e-06,
+      "loss": 0.5426,
+      "step": 15504
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.497730191728576e-06,
+      "loss": 0.6417,
+      "step": 15505
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49657044687022e-06,
+      "loss": 0.642,
+      "step": 15506
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.495410722700373e-06,
+      "loss": 0.5547,
+      "step": 15507
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.494251019234999e-06,
+      "loss": 0.6369,
+      "step": 15508
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.493091336490051e-06,
+      "loss": 0.5863,
+      "step": 15509
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.491931674481491e-06,
+      "loss": 0.5865,
+      "step": 15510
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49077203322528e-06,
+      "loss": 0.6192,
+      "step": 15511
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.489612412737366e-06,
+      "loss": 0.6115,
+      "step": 15512
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.488452813033716e-06,
+      "loss": 0.7196,
+      "step": 15513
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.487293234130283e-06,
+      "loss": 0.6169,
+      "step": 15514
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.486133676043025e-06,
+      "loss": 0.5713,
+      "step": 15515
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.484974138787897e-06,
+      "loss": 0.5632,
+      "step": 15516
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.483814622380856e-06,
+      "loss": 0.519,
+      "step": 15517
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.482655126837857e-06,
+      "loss": 0.6777,
+      "step": 15518
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.481495652174859e-06,
+      "loss": 0.602,
+      "step": 15519
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.480336198407814e-06,
+      "loss": 0.5297,
+      "step": 15520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.479176765552677e-06,
+      "loss": 0.5785,
+      "step": 15521
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.478017353625406e-06,
+      "loss": 0.6019,
+      "step": 15522
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.476857962641951e-06,
+      "loss": 0.674,
+      "step": 15523
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.475698592618272e-06,
+      "loss": 0.6082,
+      "step": 15524
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.474539243570316e-06,
+      "loss": 0.5807,
+      "step": 15525
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.473379915514045e-06,
+      "loss": 0.661,
+      "step": 15526
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.472220608465404e-06,
+      "loss": 0.6202,
+      "step": 15527
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.471061322440353e-06,
+      "loss": 0.6379,
+      "step": 15528
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.469902057454838e-06,
+      "loss": 0.5643,
+      "step": 15529
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.468742813524817e-06,
+      "loss": 0.5595,
+      "step": 15530
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.467583590666242e-06,
+      "loss": 0.5726,
+      "step": 15531
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46642438889506e-06,
+      "loss": 0.6765,
+      "step": 15532
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.465265208227229e-06,
+      "loss": 0.6394,
+      "step": 15533
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.464106048678696e-06,
+      "loss": 0.6552,
+      "step": 15534
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.462946910265413e-06,
+      "loss": 0.6499,
+      "step": 15535
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46178779300333e-06,
+      "loss": 0.6735,
+      "step": 15536
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.460628696908402e-06,
+      "loss": 0.5854,
+      "step": 15537
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.459469621996572e-06,
+      "loss": 0.5907,
+      "step": 15538
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.458310568283797e-06,
+      "loss": 0.6174,
+      "step": 15539
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.457151535786019e-06,
+      "loss": 0.6428,
+      "step": 15540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.455992524519194e-06,
+      "loss": 0.5804,
+      "step": 15541
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45483353449927e-06,
+      "loss": 0.5687,
+      "step": 15542
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.453674565742191e-06,
+      "loss": 0.5986,
+      "step": 15543
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.452515618263912e-06,
+      "loss": 0.6766,
+      "step": 15544
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.451356692080375e-06,
+      "loss": 0.5916,
+      "step": 15545
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.450197787207533e-06,
+      "loss": 0.5848,
+      "step": 15546
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.44903890366133e-06,
+      "loss": 0.5852,
+      "step": 15547
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.447880041457715e-06,
+      "loss": 0.6319,
+      "step": 15548
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.446721200612632e-06,
+      "loss": 0.5657,
+      "step": 15549
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.445562381142035e-06,
+      "loss": 0.5929,
+      "step": 15550
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.444403583061861e-06,
+      "loss": 0.5997,
+      "step": 15551
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.443244806388062e-06,
+      "loss": 0.6245,
+      "step": 15552
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.442086051136584e-06,
+      "loss": 0.5253,
+      "step": 15553
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.440927317323368e-06,
+      "loss": 0.6484,
+      "step": 15554
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.439768604964364e-06,
+      "loss": 0.5992,
+      "step": 15555
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.438609914075514e-06,
+      "loss": 0.6649,
+      "step": 15556
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.437451244672766e-06,
+      "loss": 0.6472,
+      "step": 15557
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.436292596772058e-06,
+      "loss": 0.5804,
+      "step": 15558
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.43513397038934e-06,
+      "loss": 0.5586,
+      "step": 15559
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.433975365540554e-06,
+      "loss": 0.5509,
+      "step": 15560
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.43281678224164e-06,
+      "loss": 0.5696,
+      "step": 15561
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.431658220508549e-06,
+      "loss": 0.6332,
+      "step": 15562
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.430499680357215e-06,
+      "loss": 0.6082,
+      "step": 15563
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.429341161803587e-06,
+      "loss": 0.6047,
+      "step": 15564
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.428182664863602e-06,
+      "loss": 0.5683,
+      "step": 15565
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.427024189553209e-06,
+      "loss": 0.6195,
+      "step": 15566
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42586573588834e-06,
+      "loss": 0.6076,
+      "step": 15567
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.424707303884945e-06,
+      "loss": 0.5702,
+      "step": 15568
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42354889355896e-06,
+      "loss": 0.6347,
+      "step": 15569
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42239050492633e-06,
+      "loss": 0.5902,
+      "step": 15570
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42123213800299e-06,
+      "loss": 0.6097,
+      "step": 15571
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.420073792804882e-06,
+      "loss": 0.5749,
+      "step": 15572
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.418915469347952e-06,
+      "loss": 0.5491,
+      "step": 15573
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41775716764813e-06,
+      "loss": 0.555,
+      "step": 15574
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.416598887721362e-06,
+      "loss": 0.663,
+      "step": 15575
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.415440629583583e-06,
+      "loss": 0.5738,
+      "step": 15576
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.414282393250734e-06,
+      "loss": 0.5963,
+      "step": 15577
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.413124178738753e-06,
+      "loss": 0.7017,
+      "step": 15578
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.411965986063578e-06,
+      "loss": 0.5738,
+      "step": 15579
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.410807815241143e-06,
+      "loss": 0.6875,
+      "step": 15580
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.409649666287393e-06,
+      "loss": 0.6685,
+      "step": 15581
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.408491539218258e-06,
+      "loss": 0.5705,
+      "step": 15582
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.407333434049678e-06,
+      "loss": 0.593,
+      "step": 15583
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.406175350797591e-06,
+      "loss": 0.5882,
+      "step": 15584
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.405017289477931e-06,
+      "loss": 0.5845,
+      "step": 15585
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.403859250106636e-06,
+      "loss": 0.5661,
+      "step": 15586
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.402701232699638e-06,
+      "loss": 0.5746,
+      "step": 15587
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.401543237272877e-06,
+      "loss": 0.5988,
+      "step": 15588
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.400385263842283e-06,
+      "loss": 0.6044,
+      "step": 15589
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.399227312423796e-06,
+      "loss": 0.6187,
+      "step": 15590
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.398069383033346e-06,
+      "loss": 0.5788,
+      "step": 15591
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.396911475686867e-06,
+      "loss": 0.6443,
+      "step": 15592
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.395753590400297e-06,
+      "loss": 0.6535,
+      "step": 15593
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.394595727189568e-06,
+      "loss": 0.6776,
+      "step": 15594
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.393437886070612e-06,
+      "loss": 0.6256,
+      "step": 15595
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.392280067059359e-06,
+      "loss": 0.5902,
+      "step": 15596
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.39112227017175e-06,
+      "loss": 0.5477,
+      "step": 15597
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.389964495423709e-06,
+      "loss": 0.6464,
+      "step": 15598
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.388806742831172e-06,
+      "loss": 0.6165,
+      "step": 15599
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.387649012410071e-06,
+      "loss": 0.6317,
+      "step": 15600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.386491304176337e-06,
+      "loss": 0.6281,
+      "step": 15601
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.385333618145896e-06,
+      "loss": 0.532,
+      "step": 15602
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.384175954334686e-06,
+      "loss": 0.5337,
+      "step": 15603
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.383018312758635e-06,
+      "loss": 0.5878,
+      "step": 15604
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.381860693433672e-06,
+      "loss": 0.6398,
+      "step": 15605
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.38070309637573e-06,
+      "loss": 0.6501,
+      "step": 15606
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.379545521600732e-06,
+      "loss": 0.5672,
+      "step": 15607
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.378387969124615e-06,
+      "loss": 0.6067,
+      "step": 15608
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.377230438963302e-06,
+      "loss": 0.634,
+      "step": 15609
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.376072931132727e-06,
+      "loss": 0.6691,
+      "step": 15610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.374915445648814e-06,
+      "loss": 0.6468,
+      "step": 15611
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.373757982527494e-06,
+      "loss": 0.6345,
+      "step": 15612
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.372600541784689e-06,
+      "loss": 0.6082,
+      "step": 15613
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.371443123436331e-06,
+      "loss": 0.6627,
+      "step": 15614
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.370285727498349e-06,
+      "loss": 0.6097,
+      "step": 15615
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.369128353986667e-06,
+      "loss": 0.588,
+      "step": 15616
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.367971002917213e-06,
+      "loss": 0.5959,
+      "step": 15617
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.366813674305909e-06,
+      "loss": 0.6402,
+      "step": 15618
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.365656368168686e-06,
+      "loss": 0.5696,
+      "step": 15619
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.364499084521467e-06,
+      "loss": 0.5604,
+      "step": 15620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.363341823380178e-06,
+      "loss": 0.672,
+      "step": 15621
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.362184584760742e-06,
+      "loss": 0.5705,
+      "step": 15622
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.361027368679085e-06,
+      "loss": 0.6076,
+      "step": 15623
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.359870175151135e-06,
+      "loss": 0.5479,
+      "step": 15624
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.358713004192809e-06,
+      "loss": 0.653,
+      "step": 15625
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.357555855820036e-06,
+      "loss": 0.5988,
+      "step": 15626
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.356398730048737e-06,
+      "loss": 0.5684,
+      "step": 15627
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.355241626894837e-06,
+      "loss": 0.6674,
+      "step": 15628
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.354084546374255e-06,
+      "loss": 0.6377,
+      "step": 15629
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.35292748850292e-06,
+      "loss": 0.6353,
+      "step": 15630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.351770453296746e-06,
+      "loss": 0.5687,
+      "step": 15631
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.350613440771661e-06,
+      "loss": 0.5606,
+      "step": 15632
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.349456450943583e-06,
+      "loss": 0.6321,
+      "step": 15633
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.348299483828433e-06,
+      "loss": 0.5425,
+      "step": 15634
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.347142539442136e-06,
+      "loss": 0.6169,
+      "step": 15635
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.345985617800609e-06,
+      "loss": 0.6658,
+      "step": 15636
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.344828718919775e-06,
+      "loss": 0.6286,
+      "step": 15637
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.34367184281555e-06,
+      "loss": 0.5746,
+      "step": 15638
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.342514989503857e-06,
+      "loss": 0.5876,
+      "step": 15639
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.341358159000613e-06,
+      "loss": 0.5837,
+      "step": 15640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.340201351321741e-06,
+      "loss": 0.5776,
+      "step": 15641
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.339044566483153e-06,
+      "loss": 0.5424,
+      "step": 15642
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.337887804500773e-06,
+      "loss": 0.5738,
+      "step": 15643
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.336731065390517e-06,
+      "loss": 0.5796,
+      "step": 15644
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.335574349168303e-06,
+      "loss": 0.5928,
+      "step": 15645
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.33441765585005e-06,
+      "loss": 0.5967,
+      "step": 15646
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.33326098545167e-06,
+      "loss": 0.6253,
+      "step": 15647
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.332104337989086e-06,
+      "loss": 0.5915,
+      "step": 15648
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.330947713478212e-06,
+      "loss": 0.6188,
+      "step": 15649
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.329791111934963e-06,
+      "loss": 0.5911,
+      "step": 15650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.328634533375258e-06,
+      "loss": 0.6076,
+      "step": 15651
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.327477977815009e-06,
+      "loss": 0.6231,
+      "step": 15652
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32632144527013e-06,
+      "loss": 0.6304,
+      "step": 15653
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32516493575654e-06,
+      "loss": 0.6478,
+      "step": 15654
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.324008449290155e-06,
+      "loss": 0.6064,
+      "step": 15655
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.322851985886884e-06,
+      "loss": 0.576,
+      "step": 15656
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.321695545562644e-06,
+      "loss": 0.594,
+      "step": 15657
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.320539128333346e-06,
+      "loss": 0.6435,
+      "step": 15658
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.319382734214909e-06,
+      "loss": 0.6014,
+      "step": 15659
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.31822636322324e-06,
+      "loss": 0.5161,
+      "step": 15660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.317070015374256e-06,
+      "loss": 0.6397,
+      "step": 15661
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.315913690683864e-06,
+      "loss": 0.6219,
+      "step": 15662
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.314757389167983e-06,
+      "loss": 0.6199,
+      "step": 15663
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.313601110842517e-06,
+      "loss": 0.6649,
+      "step": 15664
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.312444855723382e-06,
+      "loss": 0.5806,
+      "step": 15665
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.311288623826493e-06,
+      "loss": 0.5622,
+      "step": 15666
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.310132415167752e-06,
+      "loss": 0.6332,
+      "step": 15667
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.308976229763076e-06,
+      "loss": 0.6486,
+      "step": 15668
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.307820067628371e-06,
+      "loss": 0.6163,
+      "step": 15669
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.306663928779553e-06,
+      "loss": 0.6021,
+      "step": 15670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.305507813232522e-06,
+      "loss": 0.5944,
+      "step": 15671
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.304351721003197e-06,
+      "loss": 0.5617,
+      "step": 15672
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.303195652107478e-06,
+      "loss": 0.6022,
+      "step": 15673
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.302039606561282e-06,
+      "loss": 0.6078,
+      "step": 15674
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.30088358438051e-06,
+      "loss": 0.5849,
+      "step": 15675
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.299727585581073e-06,
+      "loss": 0.6505,
+      "step": 15676
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.298571610178879e-06,
+      "loss": 0.5771,
+      "step": 15677
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.297415658189834e-06,
+      "loss": 0.6171,
+      "step": 15678
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.296259729629847e-06,
+      "loss": 0.6566,
+      "step": 15679
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.295103824514821e-06,
+      "loss": 0.5658,
+      "step": 15680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.293947942860667e-06,
+      "loss": 0.6115,
+      "step": 15681
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.292792084683287e-06,
+      "loss": 0.6094,
+      "step": 15682
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.291636249998588e-06,
+      "loss": 0.5834,
+      "step": 15683
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.290480438822476e-06,
+      "loss": 0.6561,
+      "step": 15684
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.289324651170853e-06,
+      "loss": 0.6121,
+      "step": 15685
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.288168887059629e-06,
+      "loss": 0.5838,
+      "step": 15686
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.287013146504703e-06,
+      "loss": 0.6246,
+      "step": 15687
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.285857429521983e-06,
+      "loss": 0.6287,
+      "step": 15688
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.28470173612737e-06,
+      "loss": 0.5472,
+      "step": 15689
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.28354606633677e-06,
+      "loss": 0.6545,
+      "step": 15690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.28239042016608e-06,
+      "loss": 0.5617,
+      "step": 15691
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.281234797631212e-06,
+      "loss": 0.6177,
+      "step": 15692
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.280079198748061e-06,
+      "loss": 0.5646,
+      "step": 15693
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.278923623532532e-06,
+      "loss": 0.5737,
+      "step": 15694
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.277768072000526e-06,
+      "loss": 0.5532,
+      "step": 15695
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.276612544167947e-06,
+      "loss": 0.6742,
+      "step": 15696
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.27545704005069e-06,
+      "loss": 0.562,
+      "step": 15697
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.274301559664662e-06,
+      "loss": 0.6499,
+      "step": 15698
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.27314610302576e-06,
+      "loss": 0.613,
+      "step": 15699
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.271990670149885e-06,
+      "loss": 0.5611,
+      "step": 15700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.270835261052935e-06,
+      "loss": 0.5736,
+      "step": 15701
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.269679875750814e-06,
+      "loss": 0.6911,
+      "step": 15702
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.268524514259416e-06,
+      "loss": 0.6235,
+      "step": 15703
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.267369176594643e-06,
+      "loss": 0.5725,
+      "step": 15704
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.266213862772393e-06,
+      "loss": 0.5692,
+      "step": 15705
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.265058572808565e-06,
+      "loss": 0.5733,
+      "step": 15706
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.263903306719052e-06,
+      "loss": 0.6132,
+      "step": 15707
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.262748064519754e-06,
+      "loss": 0.5991,
+      "step": 15708
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.261592846226574e-06,
+      "loss": 0.5955,
+      "step": 15709
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2604376518554e-06,
+      "loss": 0.6436,
+      "step": 15710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.259282481422134e-06,
+      "loss": 0.6112,
+      "step": 15711
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.258127334942672e-06,
+      "loss": 0.6591,
+      "step": 15712
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.256972212432908e-06,
+      "loss": 0.6354,
+      "step": 15713
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.255817113908735e-06,
+      "loss": 0.6481,
+      "step": 15714
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.254662039386055e-06,
+      "loss": 0.6439,
+      "step": 15715
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.253506988880757e-06,
+      "loss": 0.5938,
+      "step": 15716
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.252351962408738e-06,
+      "loss": 0.6723,
+      "step": 15717
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.251196959985889e-06,
+      "loss": 0.6464,
+      "step": 15718
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.250041981628107e-06,
+      "loss": 0.588,
+      "step": 15719
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.24888702735129e-06,
+      "loss": 0.6827,
+      "step": 15720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.24773209717132e-06,
+      "loss": 0.6378,
+      "step": 15721
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.246577191104101e-06,
+      "loss": 0.6447,
+      "step": 15722
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.245422309165516e-06,
+      "loss": 0.6133,
+      "step": 15723
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.244267451371465e-06,
+      "loss": 0.6174,
+      "step": 15724
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.243112617737835e-06,
+      "loss": 0.6,
+      "step": 15725
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.241957808280522e-06,
+      "loss": 0.6097,
+      "step": 15726
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.24080302301541e-06,
+      "loss": 0.6106,
+      "step": 15727
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.239648261958394e-06,
+      "loss": 0.5375,
+      "step": 15728
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.238493525125369e-06,
+      "loss": 0.558,
+      "step": 15729
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.237338812532217e-06,
+      "loss": 0.6226,
+      "step": 15730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.236184124194834e-06,
+      "loss": 0.613,
+      "step": 15731
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.235029460129108e-06,
+      "loss": 0.6234,
+      "step": 15732
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.233874820350926e-06,
+      "loss": 0.5939,
+      "step": 15733
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.232720204876178e-06,
+      "loss": 0.6074,
+      "step": 15734
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.231565613720755e-06,
+      "loss": 0.575,
+      "step": 15735
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23041104690054e-06,
+      "loss": 0.6313,
+      "step": 15736
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.229256504431426e-06,
+      "loss": 0.6344,
+      "step": 15737
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.228101986329297e-06,
+      "loss": 0.558,
+      "step": 15738
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.226947492610041e-06,
+      "loss": 0.5303,
+      "step": 15739
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.22579302328955e-06,
+      "loss": 0.6733,
+      "step": 15740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2246385783837e-06,
+      "loss": 0.5828,
+      "step": 15741
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.223484157908386e-06,
+      "loss": 0.6182,
+      "step": 15742
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.22232976187949e-06,
+      "loss": 0.6807,
+      "step": 15743
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2211753903129e-06,
+      "loss": 0.6273,
+      "step": 15744
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2200210432245e-06,
+      "loss": 0.6172,
+      "step": 15745
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.218866720630174e-06,
+      "loss": 0.6951,
+      "step": 15746
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.217712422545805e-06,
+      "loss": 0.655,
+      "step": 15747
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.216558148987282e-06,
+      "loss": 0.5482,
+      "step": 15748
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.215403899970483e-06,
+      "loss": 0.5319,
+      "step": 15749
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.214249675511295e-06,
+      "loss": 0.6404,
+      "step": 15750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.213095475625602e-06,
+      "loss": 0.6505,
+      "step": 15751
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.211941300329284e-06,
+      "loss": 0.5942,
+      "step": 15752
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.210787149638227e-06,
+      "loss": 0.6108,
+      "step": 15753
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.20963302356831e-06,
+      "loss": 0.5717,
+      "step": 15754
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.208478922135416e-06,
+      "loss": 0.5771,
+      "step": 15755
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.207324845355425e-06,
+      "loss": 0.5938,
+      "step": 15756
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.20617079324422e-06,
+      "loss": 0.6052,
+      "step": 15757
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.205016765817679e-06,
+      "loss": 0.6206,
+      "step": 15758
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.203862763091686e-06,
+      "loss": 0.5877,
+      "step": 15759
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.202708785082122e-06,
+      "loss": 0.585,
+      "step": 15760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.20155483180486e-06,
+      "loss": 0.7279,
+      "step": 15761
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.200400903275788e-06,
+      "loss": 0.5694,
+      "step": 15762
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.199246999510779e-06,
+      "loss": 0.5508,
+      "step": 15763
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.198093120525713e-06,
+      "loss": 0.6276,
+      "step": 15764
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.19693926633647e-06,
+      "loss": 0.6001,
+      "step": 15765
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.195785436958928e-06,
+      "loss": 0.6289,
+      "step": 15766
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.19463163240896e-06,
+      "loss": 0.5709,
+      "step": 15767
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.193477852702451e-06,
+      "loss": 0.6057,
+      "step": 15768
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.19232409785527e-06,
+      "loss": 0.5565,
+      "step": 15769
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1911703678833e-06,
+      "loss": 0.5485,
+      "step": 15770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.190016662802417e-06,
+      "loss": 0.5638,
+      "step": 15771
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.188862982628492e-06,
+      "loss": 0.5995,
+      "step": 15772
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.187709327377407e-06,
+      "loss": 0.6029,
+      "step": 15773
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.186555697065031e-06,
+      "loss": 0.537,
+      "step": 15774
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.185402091707244e-06,
+      "loss": 0.6309,
+      "step": 15775
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.184248511319917e-06,
+      "loss": 0.609,
+      "step": 15776
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.18309495591893e-06,
+      "loss": 0.5434,
+      "step": 15777
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.181941425520148e-06,
+      "loss": 0.5931,
+      "step": 15778
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.180787920139452e-06,
+      "loss": 0.5917,
+      "step": 15779
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.179634439792712e-06,
+      "loss": 0.5821,
+      "step": 15780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1784809844958e-06,
+      "loss": 0.6558,
+      "step": 15781
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.177327554264592e-06,
+      "loss": 0.6708,
+      "step": 15782
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.176174149114957e-06,
+      "loss": 0.5954,
+      "step": 15783
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.175020769062772e-06,
+      "loss": 0.5467,
+      "step": 15784
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1738674141239e-06,
+      "loss": 0.513,
+      "step": 15785
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.17271408431422e-06,
+      "loss": 0.5671,
+      "step": 15786
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.171560779649598e-06,
+      "loss": 0.5653,
+      "step": 15787
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.17040750014591e-06,
+      "loss": 0.6055,
+      "step": 15788
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.169254245819018e-06,
+      "loss": 0.5067,
+      "step": 15789
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.168101016684796e-06,
+      "loss": 0.644,
+      "step": 15790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.166947812759118e-06,
+      "loss": 0.5239,
+      "step": 15791
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.165794634057843e-06,
+      "loss": 0.5557,
+      "step": 15792
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.164641480596852e-06,
+      "loss": 0.5691,
+      "step": 15793
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.163488352392001e-06,
+      "loss": 0.5693,
+      "step": 15794
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.162335249459169e-06,
+      "loss": 0.5841,
+      "step": 15795
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.161182171814214e-06,
+      "loss": 0.6001,
+      "step": 15796
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.160029119473014e-06,
+      "loss": 0.5955,
+      "step": 15797
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.158876092451424e-06,
+      "loss": 0.5763,
+      "step": 15798
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.157723090765321e-06,
+      "loss": 0.6306,
+      "step": 15799
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.156570114430564e-06,
+      "loss": 0.6627,
+      "step": 15800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.155417163463022e-06,
+      "loss": 0.6064,
+      "step": 15801
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.154264237878564e-06,
+      "loss": 0.6095,
+      "step": 15802
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.153111337693047e-06,
+      "loss": 0.6149,
+      "step": 15803
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.151958462922345e-06,
+      "loss": 0.566,
+      "step": 15804
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.150805613582317e-06,
+      "loss": 0.5627,
+      "step": 15805
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.149652789688828e-06,
+      "loss": 0.7061,
+      "step": 15806
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.148499991257744e-06,
+      "loss": 0.6152,
+      "step": 15807
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.147347218304927e-06,
+      "loss": 0.6327,
+      "step": 15808
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.146194470846237e-06,
+      "loss": 0.576,
+      "step": 15809
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.145041748897544e-06,
+      "loss": 0.5659,
+      "step": 15810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.143889052474702e-06,
+      "loss": 0.5491,
+      "step": 15811
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14273638159358e-06,
+      "loss": 0.6745,
+      "step": 15812
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.141583736270039e-06,
+      "loss": 0.6246,
+      "step": 15813
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.140431116519935e-06,
+      "loss": 0.6617,
+      "step": 15814
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.139278522359137e-06,
+      "loss": 0.6716,
+      "step": 15815
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.138125953803498e-06,
+      "loss": 0.6391,
+      "step": 15816
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.136973410868884e-06,
+      "loss": 0.651,
+      "step": 15817
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.135820893571152e-06,
+      "loss": 0.611,
+      "step": 15818
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.134668401926164e-06,
+      "loss": 0.6162,
+      "step": 15819
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.133515935949776e-06,
+      "loss": 0.5503,
+      "step": 15820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.132363495657849e-06,
+      "loss": 0.6478,
+      "step": 15821
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.131211081066241e-06,
+      "loss": 0.574,
+      "step": 15822
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.13005869219081e-06,
+      "loss": 0.6991,
+      "step": 15823
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.128906329047418e-06,
+      "loss": 0.5866,
+      "step": 15824
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.127753991651916e-06,
+      "loss": 0.5742,
+      "step": 15825
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.126601680020165e-06,
+      "loss": 0.5681,
+      "step": 15826
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.125449394168021e-06,
+      "loss": 0.5862,
+      "step": 15827
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.124297134111341e-06,
+      "loss": 0.6742,
+      "step": 15828
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.123144899865979e-06,
+      "loss": 0.5339,
+      "step": 15829
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.121992691447793e-06,
+      "loss": 0.6834,
+      "step": 15830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.120840508872639e-06,
+      "loss": 0.6046,
+      "step": 15831
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.119688352156369e-06,
+      "loss": 0.6298,
+      "step": 15832
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11853622131484e-06,
+      "loss": 0.6389,
+      "step": 15833
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.117384116363905e-06,
+      "loss": 0.6563,
+      "step": 15834
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11623203731942e-06,
+      "loss": 0.6266,
+      "step": 15835
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.115079984197238e-06,
+      "loss": 0.5823,
+      "step": 15836
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.113927957013212e-06,
+      "loss": 0.6261,
+      "step": 15837
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.112775955783192e-06,
+      "loss": 0.5627,
+      "step": 15838
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.111623980523036e-06,
+      "loss": 0.5798,
+      "step": 15839
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11047203124859e-06,
+      "loss": 0.6742,
+      "step": 15840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.109320107975711e-06,
+      "loss": 0.5981,
+      "step": 15841
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10816821072025e-06,
+      "loss": 0.5802,
+      "step": 15842
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.107016339498053e-06,
+      "loss": 0.6621,
+      "step": 15843
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.105864494324979e-06,
+      "loss": 0.5638,
+      "step": 15844
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.104712675216871e-06,
+      "loss": 0.5737,
+      "step": 15845
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.103560882189582e-06,
+      "loss": 0.5986,
+      "step": 15846
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.102409115258962e-06,
+      "loss": 0.7309,
+      "step": 15847
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10125737444086e-06,
+      "loss": 0.6024,
+      "step": 15848
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.100105659751126e-06,
+      "loss": 0.6343,
+      "step": 15849
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.098953971205607e-06,
+      "loss": 0.6424,
+      "step": 15850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09780230882015e-06,
+      "loss": 0.5343,
+      "step": 15851
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.096650672610605e-06,
+      "loss": 0.6401,
+      "step": 15852
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.095499062592822e-06,
+      "loss": 0.5876,
+      "step": 15853
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.094347478782643e-06,
+      "loss": 0.579,
+      "step": 15854
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.093195921195919e-06,
+      "loss": 0.7077,
+      "step": 15855
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.092044389848491e-06,
+      "loss": 0.6591,
+      "step": 15856
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.090892884756212e-06,
+      "loss": 0.5789,
+      "step": 15857
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.089741405934923e-06,
+      "loss": 0.5451,
+      "step": 15858
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.088589953400472e-06,
+      "loss": 0.5946,
+      "step": 15859
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.0874385271687e-06,
+      "loss": 0.5996,
+      "step": 15860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.086287127255457e-06,
+      "loss": 0.6403,
+      "step": 15861
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.085135753676583e-06,
+      "loss": 0.6113,
+      "step": 15862
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.083984406447925e-06,
+      "loss": 0.6013,
+      "step": 15863
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.082833085585324e-06,
+      "loss": 0.5487,
+      "step": 15864
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.081681791104624e-06,
+      "loss": 0.649,
+      "step": 15865
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.08053052302167e-06,
+      "loss": 0.7154,
+      "step": 15866
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.0793792813523e-06,
+      "loss": 0.6232,
+      "step": 15867
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.078228066112363e-06,
+      "loss": 0.5875,
+      "step": 15868
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.077076877317692e-06,
+      "loss": 0.6508,
+      "step": 15869
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.075925714984136e-06,
+      "loss": 0.6778,
+      "step": 15870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07477457912753e-06,
+      "loss": 0.6065,
+      "step": 15871
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07362346976372e-06,
+      "loss": 0.6121,
+      "step": 15872
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.072472386908543e-06,
+      "loss": 0.6526,
+      "step": 15873
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.071321330577838e-06,
+      "loss": 0.5763,
+      "step": 15874
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.070170300787448e-06,
+      "loss": 0.5264,
+      "step": 15875
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.069019297553208e-06,
+      "loss": 0.6564,
+      "step": 15876
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.067868320890961e-06,
+      "loss": 0.5597,
+      "step": 15877
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.066717370816543e-06,
+      "loss": 0.6457,
+      "step": 15878
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.065566447345794e-06,
+      "loss": 0.54,
+      "step": 15879
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.064415550494549e-06,
+      "loss": 0.6159,
+      "step": 15880
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.063264680278648e-06,
+      "loss": 0.6348,
+      "step": 15881
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.062113836713925e-06,
+      "loss": 0.5798,
+      "step": 15882
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.060963019816219e-06,
+      "loss": 0.6577,
+      "step": 15883
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.059812229601367e-06,
+      "loss": 0.6243,
+      "step": 15884
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.058661466085203e-06,
+      "loss": 0.5751,
+      "step": 15885
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.057510729283563e-06,
+      "loss": 0.6155,
+      "step": 15886
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.056360019212281e-06,
+      "loss": 0.5164,
+      "step": 15887
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.055209335887195e-06,
+      "loss": 0.6184,
+      "step": 15888
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.054058679324135e-06,
+      "loss": 0.5951,
+      "step": 15889
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.05290804953894e-06,
+      "loss": 0.5491,
+      "step": 15890
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.051757446547439e-06,
+      "loss": 0.5779,
+      "step": 15891
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.05060687036547e-06,
+      "loss": 0.6232,
+      "step": 15892
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.049456321008859e-06,
+      "loss": 0.5796,
+      "step": 15893
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.048305798493446e-06,
+      "loss": 0.651,
+      "step": 15894
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.04715530283506e-06,
+      "loss": 0.5966,
+      "step": 15895
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.046004834049531e-06,
+      "loss": 0.5961,
+      "step": 15896
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.044854392152696e-06,
+      "loss": 0.6287,
+      "step": 15897
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.043703977160378e-06,
+      "loss": 0.6232,
+      "step": 15898
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.042553589088416e-06,
+      "loss": 0.6215,
+      "step": 15899
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.041403227952635e-06,
+      "loss": 0.6104,
+      "step": 15900
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.040252893768868e-06,
+      "loss": 0.6174,
+      "step": 15901
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.039102586552941e-06,
+      "loss": 0.7008,
+      "step": 15902
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.037952306320684e-06,
+      "loss": 0.6082,
+      "step": 15903
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.036802053087932e-06,
+      "loss": 0.6318,
+      "step": 15904
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.035651826870505e-06,
+      "loss": 0.6652,
+      "step": 15905
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.034501627684237e-06,
+      "loss": 0.6993,
+      "step": 15906
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.033351455544957e-06,
+      "loss": 0.6408,
+      "step": 15907
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.032201310468484e-06,
+      "loss": 0.5801,
+      "step": 15908
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.03105119247065e-06,
+      "loss": 0.5685,
+      "step": 15909
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.029901101567284e-06,
+      "loss": 0.6164,
+      "step": 15910
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02875103777421e-06,
+      "loss": 0.6212,
+      "step": 15911
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02760100110725e-06,
+      "loss": 0.675,
+      "step": 15912
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.026450991582237e-06,
+      "loss": 0.6045,
+      "step": 15913
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02530100921499e-06,
+      "loss": 0.5615,
+      "step": 15914
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.024151054021337e-06,
+      "loss": 0.6206,
+      "step": 15915
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.023001126017102e-06,
+      "loss": 0.6628,
+      "step": 15916
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02185122521811e-06,
+      "loss": 0.6871,
+      "step": 15917
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.020701351640182e-06,
+      "loss": 0.605,
+      "step": 15918
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.01955150529914e-06,
+      "loss": 0.6517,
+      "step": 15919
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.018401686210815e-06,
+      "loss": 0.5582,
+      "step": 15920
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.01725189439102e-06,
+      "loss": 0.6112,
+      "step": 15921
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.016102129855584e-06,
+      "loss": 0.5333,
+      "step": 15922
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.014952392620322e-06,
+      "loss": 0.5958,
+      "step": 15923
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.013802682701063e-06,
+      "loss": 0.5386,
+      "step": 15924
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.012653000113621e-06,
+      "loss": 0.6071,
+      "step": 15925
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.011503344873821e-06,
+      "loss": 0.6332,
+      "step": 15926
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.010353716997483e-06,
+      "loss": 0.5923,
+      "step": 15927
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.009204116500426e-06,
+      "loss": 0.6326,
+      "step": 15928
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.008054543398471e-06,
+      "loss": 0.6894,
+      "step": 15929
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.006904997707432e-06,
+      "loss": 0.5988,
+      "step": 15930
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.005755479443136e-06,
+      "loss": 0.5894,
+      "step": 15931
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.004605988621394e-06,
+      "loss": 0.5592,
+      "step": 15932
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.003456525258028e-06,
+      "loss": 0.5982,
+      "step": 15933
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.002307089368853e-06,
+      "loss": 0.5349,
+      "step": 15934
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.00115768096969e-06,
+      "loss": 0.5919,
+      "step": 15935
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.000008300076351e-06,
+      "loss": 0.5956,
+      "step": 15936
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.998858946704657e-06,
+      "loss": 0.5655,
+      "step": 15937
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.997709620870421e-06,
+      "loss": 0.6582,
+      "step": 15938
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.996560322589462e-06,
+      "loss": 0.6213,
+      "step": 15939
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.995411051877593e-06,
+      "loss": 0.6164,
+      "step": 15940
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.994261808750628e-06,
+      "loss": 0.6189,
+      "step": 15941
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.993112593224384e-06,
+      "loss": 0.5788,
+      "step": 15942
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.991963405314672e-06,
+      "loss": 0.582,
+      "step": 15943
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.990814245037312e-06,
+      "loss": 0.5473,
+      "step": 15944
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.98966511240811e-06,
+      "loss": 0.6136,
+      "step": 15945
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.988516007442882e-06,
+      "loss": 0.6444,
+      "step": 15946
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.987366930157443e-06,
+      "loss": 0.6098,
+      "step": 15947
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.986217880567603e-06,
+      "loss": 0.6399,
+      "step": 15948
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.985068858689176e-06,
+      "loss": 0.6851,
+      "step": 15949
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.98391986453797e-06,
+      "loss": 0.6509,
+      "step": 15950
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.982770898129799e-06,
+      "loss": 0.6026,
+      "step": 15951
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.981621959480471e-06,
+      "loss": 0.5998,
+      "step": 15952
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.980473048605802e-06,
+      "loss": 0.5978,
+      "step": 15953
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.979324165521596e-06,
+      "loss": 0.6067,
+      "step": 15954
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.978175310243665e-06,
+      "loss": 0.6061,
+      "step": 15955
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.977026482787817e-06,
+      "loss": 0.6015,
+      "step": 15956
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.975877683169863e-06,
+      "loss": 0.6286,
+      "step": 15957
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.974728911405613e-06,
+      "loss": 0.6683,
+      "step": 15958
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97358016751087e-06,
+      "loss": 0.5778,
+      "step": 15959
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.972431451501446e-06,
+      "loss": 0.6097,
+      "step": 15960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.971282763393147e-06,
+      "loss": 0.5961,
+      "step": 15961
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97013410320178e-06,
+      "loss": 0.613,
+      "step": 15962
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96898547094315e-06,
+      "loss": 0.56,
+      "step": 15963
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.967836866633065e-06,
+      "loss": 0.6326,
+      "step": 15964
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96668829028733e-06,
+      "loss": 0.5402,
+      "step": 15965
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.965539741921753e-06,
+      "loss": 0.5432,
+      "step": 15966
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.964391221552133e-06,
+      "loss": 0.5998,
+      "step": 15967
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96324272919428e-06,
+      "loss": 0.6035,
+      "step": 15968
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.962094264863997e-06,
+      "loss": 0.6181,
+      "step": 15969
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.960945828577087e-06,
+      "loss": 0.6723,
+      "step": 15970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.959797420349356e-06,
+      "loss": 0.5902,
+      "step": 15971
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.958649040196602e-06,
+      "loss": 0.5513,
+      "step": 15972
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.957500688134635e-06,
+      "loss": 0.6341,
+      "step": 15973
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.95635236417925e-06,
+      "loss": 0.5759,
+      "step": 15974
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.955204068346254e-06,
+      "loss": 0.5322,
+      "step": 15975
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.954055800651445e-06,
+      "loss": 0.5781,
+      "step": 15976
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.952907561110625e-06,
+      "loss": 0.6987,
+      "step": 15977
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.9517593497396e-06,
+      "loss": 0.6251,
+      "step": 15978
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.95061116655416e-06,
+      "loss": 0.5851,
+      "step": 15979
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.949463011570117e-06,
+      "loss": 0.6175,
+      "step": 15980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.94831488480326e-06,
+      "loss": 0.6056,
+      "step": 15981
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.947166786269398e-06,
+      "loss": 0.6155,
+      "step": 15982
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.94601871598432e-06,
+      "loss": 0.5007,
+      "step": 15983
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.944870673963833e-06,
+      "loss": 0.5379,
+      "step": 15984
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.943722660223729e-06,
+      "loss": 0.6189,
+      "step": 15985
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.94257467477981e-06,
+      "loss": 0.5657,
+      "step": 15986
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.94142671764787e-06,
+      "loss": 0.6533,
+      "step": 15987
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.940278788843706e-06,
+      "loss": 0.5538,
+      "step": 15988
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.939130888383117e-06,
+      "loss": 0.5535,
+      "step": 15989
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.937983016281895e-06,
+      "loss": 0.6622,
+      "step": 15990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.936835172555843e-06,
+      "loss": 0.6192,
+      "step": 15991
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93568735722075e-06,
+      "loss": 0.619,
+      "step": 15992
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.934539570292415e-06,
+      "loss": 0.6286,
+      "step": 15993
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.933391811786625e-06,
+      "loss": 0.5549,
+      "step": 15994
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.932244081719186e-06,
+      "loss": 0.6691,
+      "step": 15995
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.931096380105881e-06,
+      "loss": 0.581,
+      "step": 15996
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.929948706962508e-06,
+      "loss": 0.613,
+      "step": 15997
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.928801062304862e-06,
+      "loss": 0.6531,
+      "step": 15998
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.927653446148732e-06,
+      "loss": 0.6065,
+      "step": 15999
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.926505858509911e-06,
+      "loss": 0.5776,
+      "step": 16000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92535829940419e-06,
+      "loss": 0.596,
+      "step": 16001
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.924210768847365e-06,
+      "loss": 0.6348,
+      "step": 16002
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92306326685522e-06,
+      "loss": 0.5937,
+      "step": 16003
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.921915793443553e-06,
+      "loss": 0.6324,
+      "step": 16004
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.920768348628149e-06,
+      "loss": 0.5709,
+      "step": 16005
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.919620932424801e-06,
+      "loss": 0.5561,
+      "step": 16006
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.918473544849295e-06,
+      "loss": 0.6624,
+      "step": 16007
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.917326185917422e-06,
+      "loss": 0.5945,
+      "step": 16008
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.916178855644972e-06,
+      "loss": 0.6568,
+      "step": 16009
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.915031554047732e-06,
+      "loss": 0.6456,
+      "step": 16010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.913884281141489e-06,
+      "loss": 0.6139,
+      "step": 16011
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.912737036942032e-06,
+      "loss": 0.5863,
+      "step": 16012
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.911589821465146e-06,
+      "loss": 0.6518,
+      "step": 16013
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.910442634726619e-06,
+      "loss": 0.5948,
+      "step": 16014
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90929547674224e-06,
+      "loss": 0.5974,
+      "step": 16015
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90814834752779e-06,
+      "loss": 0.5473,
+      "step": 16016
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.907001247099059e-06,
+      "loss": 0.5405,
+      "step": 16017
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.905854175471827e-06,
+      "loss": 0.5706,
+      "step": 16018
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.904707132661882e-06,
+      "loss": 0.6109,
+      "step": 16019
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90356011868501e-06,
+      "loss": 0.5976,
+      "step": 16020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90241313355699e-06,
+      "loss": 0.6731,
+      "step": 16021
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.901266177293612e-06,
+      "loss": 0.6769,
+      "step": 16022
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.900119249910653e-06,
+      "loss": 0.5737,
+      "step": 16023
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.898972351423899e-06,
+      "loss": 0.5917,
+      "step": 16024
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.897825481849131e-06,
+      "loss": 0.6297,
+      "step": 16025
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.896678641202133e-06,
+      "loss": 0.552,
+      "step": 16026
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.895531829498683e-06,
+      "loss": 0.6391,
+      "step": 16027
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.894385046754566e-06,
+      "loss": 0.5571,
+      "step": 16028
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.89323829298556e-06,
+      "loss": 0.6239,
+      "step": 16029
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.892091568207445e-06,
+      "loss": 0.591,
+      "step": 16030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.890944872436005e-06,
+      "loss": 0.5134,
+      "step": 16031
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.889798205687014e-06,
+      "loss": 0.6385,
+      "step": 16032
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.888651567976256e-06,
+      "loss": 0.586,
+      "step": 16033
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.887504959319507e-06,
+      "loss": 0.605,
+      "step": 16034
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.886358379732546e-06,
+      "loss": 0.608,
+      "step": 16035
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88521182923115e-06,
+      "loss": 0.605,
+      "step": 16036
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8840653078311e-06,
+      "loss": 0.5666,
+      "step": 16037
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.882918815548168e-06,
+      "loss": 0.5856,
+      "step": 16038
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.881772352398132e-06,
+      "loss": 0.5754,
+      "step": 16039
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.880625918396773e-06,
+      "loss": 0.6423,
+      "step": 16040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.879479513559861e-06,
+      "loss": 0.5816,
+      "step": 16041
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.878333137903177e-06,
+      "loss": 0.5925,
+      "step": 16042
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.87718679144249e-06,
+      "loss": 0.5558,
+      "step": 16043
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.87604047419358e-06,
+      "loss": 0.5737,
+      "step": 16044
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.87489418617222e-06,
+      "loss": 0.6445,
+      "step": 16045
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.873747927394182e-06,
+      "loss": 0.6314,
+      "step": 16046
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.872601697875242e-06,
+      "loss": 0.514,
+      "step": 16047
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.871455497631171e-06,
+      "loss": 0.603,
+      "step": 16048
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.870309326677742e-06,
+      "loss": 0.6449,
+      "step": 16049
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.869163185030728e-06,
+      "loss": 0.6077,
+      "step": 16050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.868017072705903e-06,
+      "loss": 0.6101,
+      "step": 16051
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.866870989719034e-06,
+      "loss": 0.5509,
+      "step": 16052
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.865724936085897e-06,
+      "loss": 0.5954,
+      "step": 16053
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.864578911822258e-06,
+      "loss": 0.5919,
+      "step": 16054
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.863432916943892e-06,
+      "loss": 0.5783,
+      "step": 16055
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.862286951466563e-06,
+      "loss": 0.6242,
+      "step": 16056
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.861141015406046e-06,
+      "loss": 0.6476,
+      "step": 16057
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.859995108778106e-06,
+      "loss": 0.6433,
+      "step": 16058
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.858849231598517e-06,
+      "loss": 0.6171,
+      "step": 16059
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.857703383883042e-06,
+      "loss": 0.6229,
+      "step": 16060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85655756564745e-06,
+      "loss": 0.6159,
+      "step": 16061
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85541177690751e-06,
+      "loss": 0.5432,
+      "step": 16062
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85426601767899e-06,
+      "loss": 0.6187,
+      "step": 16063
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.853120287977654e-06,
+      "loss": 0.6868,
+      "step": 16064
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.851974587819267e-06,
+      "loss": 0.5725,
+      "step": 16065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.850828917219601e-06,
+      "loss": 0.5975,
+      "step": 16066
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.849683276194416e-06,
+      "loss": 0.6451,
+      "step": 16067
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.848537664759477e-06,
+      "loss": 0.6601,
+      "step": 16068
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.847392082930552e-06,
+      "loss": 0.6632,
+      "step": 16069
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.846246530723401e-06,
+      "loss": 0.5703,
+      "step": 16070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.845101008153793e-06,
+      "loss": 0.6401,
+      "step": 16071
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.843955515237487e-06,
+      "loss": 0.5276,
+      "step": 16072
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.842810051990248e-06,
+      "loss": 0.5887,
+      "step": 16073
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.841664618427838e-06,
+      "loss": 0.5773,
+      "step": 16074
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84051921456602e-06,
+      "loss": 0.5875,
+      "step": 16075
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.839373840420555e-06,
+      "loss": 0.5495,
+      "step": 16076
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.838228496007206e-06,
+      "loss": 0.6053,
+      "step": 16077
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.837083181341728e-06,
+      "loss": 0.5922,
+      "step": 16078
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83593789643989e-06,
+      "loss": 0.603,
+      "step": 16079
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.834792641317447e-06,
+      "loss": 0.6019,
+      "step": 16080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.833647415990157e-06,
+      "loss": 0.5849,
+      "step": 16081
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.832502220473784e-06,
+      "loss": 0.6312,
+      "step": 16082
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.831357054784085e-06,
+      "loss": 0.6012,
+      "step": 16083
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83021191893682e-06,
+      "loss": 0.6168,
+      "step": 16084
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.829066812947743e-06,
+      "loss": 0.6269,
+      "step": 16085
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.827921736832617e-06,
+      "loss": 0.5494,
+      "step": 16086
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.826776690607194e-06,
+      "loss": 0.5772,
+      "step": 16087
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.825631674287235e-06,
+      "loss": 0.5768,
+      "step": 16088
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.824486687888493e-06,
+      "loss": 0.618,
+      "step": 16089
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.823341731426728e-06,
+      "loss": 0.615,
+      "step": 16090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.822196804917692e-06,
+      "loss": 0.6295,
+      "step": 16091
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.82105190837714e-06,
+      "loss": 0.5746,
+      "step": 16092
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.81990704182083e-06,
+      "loss": 0.647,
+      "step": 16093
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.818762205264514e-06,
+      "loss": 0.5648,
+      "step": 16094
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.817617398723949e-06,
+      "loss": 0.6062,
+      "step": 16095
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.816472622214884e-06,
+      "loss": 0.6274,
+      "step": 16096
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.815327875753077e-06,
+      "loss": 0.6806,
+      "step": 16097
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.814183159354274e-06,
+      "loss": 0.6157,
+      "step": 16098
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.813038473034233e-06,
+      "loss": 0.5321,
+      "step": 16099
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.811893816808704e-06,
+      "loss": 0.6248,
+      "step": 16100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.810749190693439e-06,
+      "loss": 0.5618,
+      "step": 16101
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.80960459470419e-06,
+      "loss": 0.5959,
+      "step": 16102
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.808460028856705e-06,
+      "loss": 0.6193,
+      "step": 16103
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.807315493166736e-06,
+      "loss": 0.6282,
+      "step": 16104
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.806170987650033e-06,
+      "loss": 0.5819,
+      "step": 16105
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.805026512322344e-06,
+      "loss": 0.5334,
+      "step": 16106
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.803882067199419e-06,
+      "loss": 0.5767,
+      "step": 16107
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.802737652297008e-06,
+      "loss": 0.5821,
+      "step": 16108
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.801593267630855e-06,
+      "loss": 0.5543,
+      "step": 16109
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.800448913216712e-06,
+      "loss": 0.6457,
+      "step": 16110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.799304589070325e-06,
+      "loss": 0.6334,
+      "step": 16111
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.798160295207438e-06,
+      "loss": 0.5874,
+      "step": 16112
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.797016031643803e-06,
+      "loss": 0.5991,
+      "step": 16113
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.795871798395162e-06,
+      "loss": 0.528,
+      "step": 16114
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.794727595477263e-06,
+      "loss": 0.5928,
+      "step": 16115
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.793583422905847e-06,
+      "loss": 0.5801,
+      "step": 16116
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.792439280696665e-06,
+      "loss": 0.6098,
+      "step": 16117
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.79129516886546e-06,
+      "loss": 0.6104,
+      "step": 16118
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.79015108742797e-06,
+      "loss": 0.5868,
+      "step": 16119
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.789007036399945e-06,
+      "loss": 0.5674,
+      "step": 16120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.787863015797123e-06,
+      "loss": 0.6496,
+      "step": 16121
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.786719025635254e-06,
+      "loss": 0.5487,
+      "step": 16122
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.785575065930072e-06,
+      "loss": 0.6284,
+      "step": 16123
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.784431136697324e-06,
+      "loss": 0.617,
+      "step": 16124
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.783287237952751e-06,
+      "loss": 0.623,
+      "step": 16125
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.782143369712093e-06,
+      "loss": 0.6366,
+      "step": 16126
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.780999531991092e-06,
+      "loss": 0.5749,
+      "step": 16127
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.779855724805483e-06,
+      "loss": 0.6063,
+      "step": 16128
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.778711948171016e-06,
+      "loss": 0.5881,
+      "step": 16129
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77756820210342e-06,
+      "loss": 0.5993,
+      "step": 16130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77642448661844e-06,
+      "loss": 0.584,
+      "step": 16131
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.775280801731813e-06,
+      "loss": 0.629,
+      "step": 16132
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.774137147459278e-06,
+      "loss": 0.6427,
+      "step": 16133
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77299352381657e-06,
+      "loss": 0.5888,
+      "step": 16134
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.771849930819427e-06,
+      "loss": 0.6025,
+      "step": 16135
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.770706368483589e-06,
+      "loss": 0.5938,
+      "step": 16136
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76956283682479e-06,
+      "loss": 0.5987,
+      "step": 16137
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.768419335858766e-06,
+      "loss": 0.5619,
+      "step": 16138
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76727586560125e-06,
+      "loss": 0.6501,
+      "step": 16139
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.766132426067983e-06,
+      "loss": 0.5516,
+      "step": 16140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.764989017274696e-06,
+      "loss": 0.5938,
+      "step": 16141
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.763845639237125e-06,
+      "loss": 0.5925,
+      "step": 16142
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.762702291971e-06,
+      "loss": 0.5185,
+      "step": 16143
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.761558975492059e-06,
+      "loss": 0.6263,
+      "step": 16144
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.760415689816035e-06,
+      "loss": 0.623,
+      "step": 16145
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.759272434958656e-06,
+      "loss": 0.6752,
+      "step": 16146
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.75812921093566e-06,
+      "loss": 0.5439,
+      "step": 16147
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.756986017762773e-06,
+      "loss": 0.6293,
+      "step": 16148
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.755842855455734e-06,
+      "loss": 0.632,
+      "step": 16149
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.754699724030263e-06,
+      "loss": 0.5762,
+      "step": 16150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.7535566235021e-06,
+      "loss": 0.5933,
+      "step": 16151
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.752413553886971e-06,
+      "loss": 0.5808,
+      "step": 16152
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.751270515200608e-06,
+      "loss": 0.6214,
+      "step": 16153
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.750127507458737e-06,
+      "loss": 0.5603,
+      "step": 16154
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.748984530677088e-06,
+      "loss": 0.6475,
+      "step": 16155
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.747841584871391e-06,
+      "loss": 0.6251,
+      "step": 16156
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.746698670057372e-06,
+      "loss": 0.5821,
+      "step": 16157
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.745555786250762e-06,
+      "loss": 0.6236,
+      "step": 16158
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.744412933467282e-06,
+      "loss": 0.6367,
+      "step": 16159
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.743270111722662e-06,
+      "loss": 0.5646,
+      "step": 16160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.742127321032628e-06,
+      "loss": 0.5356,
+      "step": 16161
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.74098456141291e-06,
+      "loss": 0.6189,
+      "step": 16162
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.739841832879225e-06,
+      "loss": 0.5957,
+      "step": 16163
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.738699135447305e-06,
+      "loss": 0.6382,
+      "step": 16164
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.73755646913287e-06,
+      "loss": 0.6277,
+      "step": 16165
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.736413833951646e-06,
+      "loss": 0.6141,
+      "step": 16166
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.735271229919358e-06,
+      "loss": 0.5488,
+      "step": 16167
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.734128657051725e-06,
+      "loss": 0.6485,
+      "step": 16168
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.732986115364477e-06,
+      "loss": 0.5677,
+      "step": 16169
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.731843604873328e-06,
+      "loss": 0.6454,
+      "step": 16170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.730701125594007e-06,
+      "loss": 0.5913,
+      "step": 16171
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72955867754223e-06,
+      "loss": 0.5716,
+      "step": 16172
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.728416260733723e-06,
+      "loss": 0.598,
+      "step": 16173
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.727273875184203e-06,
+      "loss": 0.6118,
+      "step": 16174
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72613152090939e-06,
+      "loss": 0.5997,
+      "step": 16175
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.724989197925008e-06,
+      "loss": 0.5502,
+      "step": 16176
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72384690624677e-06,
+      "loss": 0.5526,
+      "step": 16177
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.722704645890402e-06,
+      "loss": 0.6099,
+      "step": 16178
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.721562416871617e-06,
+      "loss": 0.6529,
+      "step": 16179
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.720420219206138e-06,
+      "loss": 0.6462,
+      "step": 16180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.719278052909678e-06,
+      "loss": 0.6438,
+      "step": 16181
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.718135917997956e-06,
+      "loss": 0.6278,
+      "step": 16182
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.716993814486689e-06,
+      "loss": 0.5862,
+      "step": 16183
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.715851742391594e-06,
+      "loss": 0.6698,
+      "step": 16184
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.714709701728383e-06,
+      "loss": 0.6103,
+      "step": 16185
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.713567692512775e-06,
+      "loss": 0.6485,
+      "step": 16186
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.712425714760488e-06,
+      "loss": 0.6448,
+      "step": 16187
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.71128376848723e-06,
+      "loss": 0.5921,
+      "step": 16188
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.710141853708722e-06,
+      "loss": 0.6042,
+      "step": 16189
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.70899997044067e-06,
+      "loss": 0.6577,
+      "step": 16190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.707858118698794e-06,
+      "loss": 0.5648,
+      "step": 16191
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.706716298498802e-06,
+      "loss": 0.5453,
+      "step": 16192
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.705574509856412e-06,
+      "loss": 0.6587,
+      "step": 16193
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.704432752787331e-06,
+      "loss": 0.5696,
+      "step": 16194
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.703291027307273e-06,
+      "loss": 0.6369,
+      "step": 16195
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.702149333431947e-06,
+      "loss": 0.6254,
+      "step": 16196
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.701007671177066e-06,
+      "loss": 0.6053,
+      "step": 16197
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69986604055834e-06,
+      "loss": 0.6179,
+      "step": 16198
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69872444159148e-06,
+      "loss": 0.5823,
+      "step": 16199
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.697582874292192e-06,
+      "loss": 0.574,
+      "step": 16200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.696441338676185e-06,
+      "loss": 0.6314,
+      "step": 16201
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.695299834759172e-06,
+      "loss": 0.5228,
+      "step": 16202
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.694158362556858e-06,
+      "loss": 0.5412,
+      "step": 16203
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69301692208495e-06,
+      "loss": 0.5743,
+      "step": 16204
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.691875513359156e-06,
+      "loss": 0.6965,
+      "step": 16205
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.690734136395183e-06,
+      "loss": 0.6658,
+      "step": 16206
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68959279120874e-06,
+      "loss": 0.6778,
+      "step": 16207
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.688451477815526e-06,
+      "loss": 0.5917,
+      "step": 16208
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.687310196231255e-06,
+      "loss": 0.5303,
+      "step": 16209
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.686168946471624e-06,
+      "loss": 0.6159,
+      "step": 16210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.685027728552344e-06,
+      "loss": 0.6211,
+      "step": 16211
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.683886542489117e-06,
+      "loss": 0.609,
+      "step": 16212
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.682745388297645e-06,
+      "loss": 0.6013,
+      "step": 16213
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.681604265993631e-06,
+      "loss": 0.6341,
+      "step": 16214
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.680463175592782e-06,
+      "loss": 0.5998,
+      "step": 16215
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.679322117110797e-06,
+      "loss": 0.6786,
+      "step": 16216
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.678181090563379e-06,
+      "loss": 0.5568,
+      "step": 16217
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.67704009596623e-06,
+      "loss": 0.6528,
+      "step": 16218
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.675899133335047e-06,
+      "loss": 0.664,
+      "step": 16219
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.674758202685538e-06,
+      "loss": 0.5307,
+      "step": 16220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.673617304033397e-06,
+      "loss": 0.5784,
+      "step": 16221
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.672476437394328e-06,
+      "loss": 0.5992,
+      "step": 16222
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.671335602784027e-06,
+      "loss": 0.7027,
+      "step": 16223
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.670194800218195e-06,
+      "loss": 0.5622,
+      "step": 16224
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.669054029712528e-06,
+      "loss": 0.5817,
+      "step": 16225
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.667913291282725e-06,
+      "loss": 0.5562,
+      "step": 16226
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.666772584944488e-06,
+      "loss": 0.5478,
+      "step": 16227
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.665631910713507e-06,
+      "loss": 0.6637,
+      "step": 16228
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.664491268605486e-06,
+      "loss": 0.6366,
+      "step": 16229
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.663350658636115e-06,
+      "loss": 0.5639,
+      "step": 16230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.662210080821092e-06,
+      "loss": 0.6069,
+      "step": 16231
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.661069535176111e-06,
+      "loss": 0.5966,
+      "step": 16232
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.659929021716871e-06,
+      "loss": 0.5886,
+      "step": 16233
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.658788540459063e-06,
+      "loss": 0.6106,
+      "step": 16234
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.657648091418381e-06,
+      "loss": 0.6205,
+      "step": 16235
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65650767461052e-06,
+      "loss": 0.6129,
+      "step": 16236
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65536729005117e-06,
+      "loss": 0.5672,
+      "step": 16237
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65422693775603e-06,
+      "loss": 0.5462,
+      "step": 16238
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.653086617740788e-06,
+      "loss": 0.5435,
+      "step": 16239
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.651946330021136e-06,
+      "loss": 0.6099,
+      "step": 16240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.650806074612765e-06,
+      "loss": 0.5333,
+      "step": 16241
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.649665851531366e-06,
+      "loss": 0.6454,
+      "step": 16242
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.648525660792631e-06,
+      "loss": 0.552,
+      "step": 16243
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.647385502412251e-06,
+      "loss": 0.5621,
+      "step": 16244
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.64624537640591e-06,
+      "loss": 0.5581,
+      "step": 16245
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.645105282789303e-06,
+      "loss": 0.555,
+      "step": 16246
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.643965221578116e-06,
+      "loss": 0.5955,
+      "step": 16247
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.642825192788037e-06,
+      "loss": 0.5413,
+      "step": 16248
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.641685196434756e-06,
+      "loss": 0.5693,
+      "step": 16249
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.640545232533957e-06,
+      "loss": 0.5559,
+      "step": 16250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.639405301101331e-06,
+      "loss": 0.5597,
+      "step": 16251
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63826540215256e-06,
+      "loss": 0.5952,
+      "step": 16252
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.637125535703336e-06,
+      "loss": 0.516,
+      "step": 16253
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.635985701769338e-06,
+      "loss": 0.6065,
+      "step": 16254
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.634845900366255e-06,
+      "loss": 0.5673,
+      "step": 16255
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.633706131509769e-06,
+      "loss": 0.5961,
+      "step": 16256
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.632566395215565e-06,
+      "loss": 0.5799,
+      "step": 16257
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63142669149933e-06,
+      "loss": 0.637,
+      "step": 16258
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.630287020376743e-06,
+      "loss": 0.6086,
+      "step": 16259
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.629147381863492e-06,
+      "loss": 0.5771,
+      "step": 16260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.628007775975254e-06,
+      "loss": 0.5545,
+      "step": 16261
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.626868202727714e-06,
+      "loss": 0.6167,
+      "step": 16262
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6257286621365516e-06,
+      "loss": 0.5498,
+      "step": 16263
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6245891542174495e-06,
+      "loss": 0.6393,
+      "step": 16264
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6234496789860875e-06,
+      "loss": 0.6385,
+      "step": 16265
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.622310236458148e-06,
+      "loss": 0.6232,
+      "step": 16266
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6211708266493065e-06,
+      "loss": 0.6146,
+      "step": 16267
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6200314495752446e-06,
+      "loss": 0.6163,
+      "step": 16268
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.618892105251644e-06,
+      "loss": 0.6986,
+      "step": 16269
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.617752793694178e-06,
+      "loss": 0.5587,
+      "step": 16270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.616613514918528e-06,
+      "loss": 0.6453,
+      "step": 16271
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.61547426894037e-06,
+      "loss": 0.5489,
+      "step": 16272
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6143350557753815e-06,
+      "loss": 0.5611,
+      "step": 16273
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.613195875439236e-06,
+      "loss": 0.6526,
+      "step": 16274
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.612056727947616e-06,
+      "loss": 0.5591,
+      "step": 16275
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.610917613316192e-06,
+      "loss": 0.5793,
+      "step": 16276
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.609778531560643e-06,
+      "loss": 0.5772,
+      "step": 16277
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.608639482696638e-06,
+      "loss": 0.6388,
+      "step": 16278
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.607500466739856e-06,
+      "loss": 0.538,
+      "step": 16279
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.606361483705973e-06,
+      "loss": 0.6195,
+      "step": 16280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.605222533610656e-06,
+      "loss": 0.5847,
+      "step": 16281
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.604083616469582e-06,
+      "loss": 0.5672,
+      "step": 16282
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.602944732298423e-06,
+      "loss": 0.6184,
+      "step": 16283
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.601805881112851e-06,
+      "loss": 0.6384,
+      "step": 16284
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.600667062928535e-06,
+      "loss": 0.588,
+      "step": 16285
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.599528277761152e-06,
+      "loss": 0.5537,
+      "step": 16286
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5983895256263664e-06,
+      "loss": 0.5443,
+      "step": 16287
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.597250806539851e-06,
+      "loss": 0.5734,
+      "step": 16288
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.596112120517279e-06,
+      "loss": 0.6005,
+      "step": 16289
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.594973467574313e-06,
+      "loss": 0.5808,
+      "step": 16290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.593834847726628e-06,
+      "loss": 0.57,
+      "step": 16291
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5926962609898865e-06,
+      "loss": 0.6099,
+      "step": 16292
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.591557707379763e-06,
+      "loss": 0.5432,
+      "step": 16293
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.590419186911919e-06,
+      "loss": 0.5952,
+      "step": 16294
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.589280699602028e-06,
+      "loss": 0.6537,
+      "step": 16295
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.588142245465748e-06,
+      "loss": 0.608,
+      "step": 16296
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.587003824518754e-06,
+      "loss": 0.5545,
+      "step": 16297
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.585865436776705e-06,
+      "loss": 0.6355,
+      "step": 16298
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.584727082255269e-06,
+      "loss": 0.5903,
+      "step": 16299
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.583588760970112e-06,
+      "loss": 0.5866,
+      "step": 16300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.582450472936895e-06,
+      "loss": 0.6471,
+      "step": 16301
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.581312218171286e-06,
+      "loss": 0.614,
+      "step": 16302
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.580173996688944e-06,
+      "loss": 0.6153,
+      "step": 16303
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.579035808505537e-06,
+      "loss": 0.595,
+      "step": 16304
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.577897653636721e-06,
+      "loss": 0.5705,
+      "step": 16305
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5767595320981645e-06,
+      "loss": 0.605,
+      "step": 16306
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.575621443905524e-06,
+      "loss": 0.5529,
+      "step": 16307
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.574483389074465e-06,
+      "loss": 0.6682,
+      "step": 16308
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.573345367620644e-06,
+      "loss": 0.6127,
+      "step": 16309
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.572207379559722e-06,
+      "loss": 0.6239,
+      "step": 16310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5710694249073615e-06,
+      "loss": 0.6428,
+      "step": 16311
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5699315036792185e-06,
+      "loss": 0.5805,
+      "step": 16312
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.568793615890955e-06,
+      "loss": 0.573,
+      "step": 16313
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5676557615582255e-06,
+      "loss": 0.5515,
+      "step": 16314
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.566517940696693e-06,
+      "loss": 0.636,
+      "step": 16315
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.565380153322008e-06,
+      "loss": 0.5866,
+      "step": 16316
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.564242399449836e-06,
+      "loss": 0.6761,
+      "step": 16317
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.563104679095825e-06,
+      "loss": 0.6227,
+      "step": 16318
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.561966992275636e-06,
+      "loss": 0.6661,
+      "step": 16319
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.560829339004923e-06,
+      "loss": 0.6242,
+      "step": 16320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.559691719299342e-06,
+      "loss": 0.6626,
+      "step": 16321
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.558554133174548e-06,
+      "loss": 0.6378,
+      "step": 16322
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.557416580646193e-06,
+      "loss": 0.603,
+      "step": 16323
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.556279061729934e-06,
+      "loss": 0.6101,
+      "step": 16324
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.55514157644142e-06,
+      "loss": 0.6204,
+      "step": 16325
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.554004124796308e-06,
+      "loss": 0.5864,
+      "step": 16326
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.552866706810247e-06,
+      "loss": 0.6078,
+      "step": 16327
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.551729322498892e-06,
+      "loss": 0.5657,
+      "step": 16328
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.550591971877895e-06,
+      "loss": 0.5922,
+      "step": 16329
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5494546549629015e-06,
+      "loss": 0.5721,
+      "step": 16330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.548317371769563e-06,
+      "loss": 0.6183,
+      "step": 16331
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.547180122313535e-06,
+      "loss": 0.5966,
+      "step": 16332
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.546042906610462e-06,
+      "loss": 0.6531,
+      "step": 16333
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.544905724675997e-06,
+      "loss": 0.5642,
+      "step": 16334
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.543768576525784e-06,
+      "loss": 0.609,
+      "step": 16335
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.542631462175476e-06,
+      "loss": 0.6246,
+      "step": 16336
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.541494381640716e-06,
+      "loss": 0.6632,
+      "step": 16337
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.540357334937156e-06,
+      "loss": 0.5698,
+      "step": 16338
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.539220322080437e-06,
+      "loss": 0.6278,
+      "step": 16339
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.538083343086212e-06,
+      "loss": 0.5791,
+      "step": 16340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.536946397970121e-06,
+      "loss": 0.6052,
+      "step": 16341
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.535809486747811e-06,
+      "loss": 0.6331,
+      "step": 16342
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.534672609434931e-06,
+      "loss": 0.5847,
+      "step": 16343
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.53353576604712e-06,
+      "loss": 0.6808,
+      "step": 16344
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.532398956600025e-06,
+      "loss": 0.5612,
+      "step": 16345
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.531262181109287e-06,
+      "loss": 0.5991,
+      "step": 16346
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.530125439590552e-06,
+      "loss": 0.6201,
+      "step": 16347
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5289887320594605e-06,
+      "loss": 0.5767,
+      "step": 16348
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.527852058531657e-06,
+      "loss": 0.5487,
+      "step": 16349
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.526715419022779e-06,
+      "loss": 0.6773,
+      "step": 16350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.525578813548472e-06,
+      "loss": 0.6117,
+      "step": 16351
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5244422421243724e-06,
+      "loss": 0.5522,
+      "step": 16352
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.523305704766124e-06,
+      "loss": 0.5908,
+      "step": 16353
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.522169201489367e-06,
+      "loss": 0.5952,
+      "step": 16354
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.521032732309736e-06,
+      "loss": 0.555,
+      "step": 16355
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.519896297242874e-06,
+      "loss": 0.5739,
+      "step": 16356
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.518759896304419e-06,
+      "loss": 0.5791,
+      "step": 16357
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.517623529510009e-06,
+      "loss": 0.6181,
+      "step": 16358
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.516487196875277e-06,
+      "loss": 0.5649,
+      "step": 16359
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.515350898415867e-06,
+      "loss": 0.6033,
+      "step": 16360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.514214634147409e-06,
+      "loss": 0.6088,
+      "step": 16361
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.513078404085542e-06,
+      "loss": 0.5476,
+      "step": 16362
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.511942208245904e-06,
+      "loss": 0.5327,
+      "step": 16363
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.510806046644125e-06,
+      "loss": 0.5985,
+      "step": 16364
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.509669919295843e-06,
+      "loss": 0.6686,
+      "step": 16365
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.50853382621669e-06,
+      "loss": 0.6615,
+      "step": 16366
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.507397767422303e-06,
+      "loss": 0.5772,
+      "step": 16367
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.50626174292831e-06,
+      "loss": 0.679,
+      "step": 16368
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.50512575275035e-06,
+      "loss": 0.5814,
+      "step": 16369
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.503989796904049e-06,
+      "loss": 0.5842,
+      "step": 16370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.502853875405043e-06,
+      "loss": 0.6491,
+      "step": 16371
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.50171798826896e-06,
+      "loss": 0.6849,
+      "step": 16372
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.500582135511434e-06,
+      "loss": 0.6264,
+      "step": 16373
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.499446317148096e-06,
+      "loss": 0.6185,
+      "step": 16374
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.49831053319457e-06,
+      "loss": 0.633,
+      "step": 16375
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.497174783666492e-06,
+      "loss": 0.6113,
+      "step": 16376
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.496039068579486e-06,
+      "loss": 0.5794,
+      "step": 16377
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4949033879491865e-06,
+      "loss": 0.6236,
+      "step": 16378
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.493767741791213e-06,
+      "loss": 0.5779,
+      "step": 16379
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4926321301212e-06,
+      "loss": 0.5946,
+      "step": 16380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.491496552954771e-06,
+      "loss": 0.6047,
+      "step": 16381
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.490361010307556e-06,
+      "loss": 0.5912,
+      "step": 16382
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.489225502195176e-06,
+      "loss": 0.5611,
+      "step": 16383
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.488090028633259e-06,
+      "loss": 0.6408,
+      "step": 16384
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.486954589637432e-06,
+      "loss": 0.5649,
+      "step": 16385
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.485819185223316e-06,
+      "loss": 0.5851,
+      "step": 16386
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.48468381540654e-06,
+      "loss": 0.5887,
+      "step": 16387
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.483548480202722e-06,
+      "loss": 0.6091,
+      "step": 16388
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4824131796274915e-06,
+      "loss": 0.6156,
+      "step": 16389
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.481277913696465e-06,
+      "loss": 0.6409,
+      "step": 16390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.480142682425268e-06,
+      "loss": 0.6098,
+      "step": 16391
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.479007485829523e-06,
+      "loss": 0.6209,
+      "step": 16392
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.477872323924847e-06,
+      "loss": 0.6228,
+      "step": 16393
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.476737196726868e-06,
+      "loss": 0.6593,
+      "step": 16394
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4756021042511995e-06,
+      "loss": 0.6152,
+      "step": 16395
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.474467046513467e-06,
+      "loss": 0.5444,
+      "step": 16396
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.473332023529284e-06,
+      "loss": 0.6106,
+      "step": 16397
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.472197035314274e-06,
+      "loss": 0.5952,
+      "step": 16398
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.471062081884053e-06,
+      "loss": 0.6237,
+      "step": 16399
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.469927163254243e-06,
+      "loss": 0.6397,
+      "step": 16400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.468792279440455e-06,
+      "loss": 0.5826,
+      "step": 16401
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4676574304583105e-06,
+      "loss": 0.6758,
+      "step": 16402
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.466522616323424e-06,
+      "loss": 0.5468,
+      "step": 16403
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.465387837051412e-06,
+      "loss": 0.5633,
+      "step": 16404
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4642530926578925e-06,
+      "loss": 0.6288,
+      "step": 16405
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4631183831584765e-06,
+      "loss": 0.5572,
+      "step": 16406
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.461983708568783e-06,
+      "loss": 0.6314,
+      "step": 16407
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4608490689044215e-06,
+      "loss": 0.5982,
+      "step": 16408
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.459714464181009e-06,
+      "loss": 0.6191,
+      "step": 16409
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.458579894414158e-06,
+      "loss": 0.6429,
+      "step": 16410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.457445359619481e-06,
+      "loss": 0.5568,
+      "step": 16411
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.456310859812588e-06,
+      "loss": 0.6455,
+      "step": 16412
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.455176395009095e-06,
+      "loss": 0.5715,
+      "step": 16413
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.454041965224608e-06,
+      "loss": 0.6097,
+      "step": 16414
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.452907570474741e-06,
+      "loss": 0.6077,
+      "step": 16415
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4517732107751075e-06,
+      "loss": 0.5839,
+      "step": 16416
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.450638886141311e-06,
+      "loss": 0.6255,
+      "step": 16417
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.449504596588964e-06,
+      "loss": 0.5837,
+      "step": 16418
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4483703421336736e-06,
+      "loss": 0.6249,
+      "step": 16419
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.447236122791052e-06,
+      "loss": 0.6499,
+      "step": 16420
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.446101938576702e-06,
+      "loss": 0.6562,
+      "step": 16421
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.444967789506234e-06,
+      "loss": 0.5762,
+      "step": 16422
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.443833675595254e-06,
+      "loss": 0.599,
+      "step": 16423
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.442699596859369e-06,
+      "loss": 0.5985,
+      "step": 16424
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.441565553314185e-06,
+      "loss": 0.5737,
+      "step": 16425
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.440431544975306e-06,
+      "loss": 0.5579,
+      "step": 16426
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.439297571858341e-06,
+      "loss": 0.6557,
+      "step": 16427
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.438163633978888e-06,
+      "loss": 0.6284,
+      "step": 16428
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.437029731352556e-06,
+      "loss": 0.5954,
+      "step": 16429
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.435895863994946e-06,
+      "loss": 0.6678,
+      "step": 16430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.434762031921663e-06,
+      "loss": 0.5624,
+      "step": 16431
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4336282351483065e-06,
+      "loss": 0.6101,
+      "step": 16432
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.432494473690484e-06,
+      "loss": 0.6351,
+      "step": 16433
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.431360747563791e-06,
+      "loss": 0.6074,
+      "step": 16434
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.430227056783829e-06,
+      "loss": 0.6488,
+      "step": 16435
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.429093401366205e-06,
+      "loss": 0.5875,
+      "step": 16436
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.427959781326512e-06,
+      "loss": 0.5414,
+      "step": 16437
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.426826196680354e-06,
+      "loss": 0.5975,
+      "step": 16438
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.425692647443328e-06,
+      "loss": 0.6363,
+      "step": 16439
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.424559133631035e-06,
+      "loss": 0.6641,
+      "step": 16440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.423425655259068e-06,
+      "loss": 0.5758,
+      "step": 16441
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.42229221234303e-06,
+      "loss": 0.6298,
+      "step": 16442
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.421158804898515e-06,
+      "loss": 0.6047,
+      "step": 16443
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.420025432941122e-06,
+      "loss": 0.6668,
+      "step": 16444
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4188920964864455e-06,
+      "loss": 0.5973,
+      "step": 16445
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.417758795550081e-06,
+      "loss": 0.6485,
+      "step": 16446
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4166255301476255e-06,
+      "loss": 0.6237,
+      "step": 16447
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.415492300294671e-06,
+      "loss": 0.583,
+      "step": 16448
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4143591060068154e-06,
+      "loss": 0.5593,
+      "step": 16449
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4132259472996496e-06,
+      "loss": 0.6158,
+      "step": 16450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.412092824188768e-06,
+      "loss": 0.6215,
+      "step": 16451
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.410959736689762e-06,
+      "loss": 0.6048,
+      "step": 16452
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.409826684818227e-06,
+      "loss": 0.6148,
+      "step": 16453
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.40869366858975e-06,
+      "loss": 0.5917,
+      "step": 16454
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.407560688019926e-06,
+      "loss": 0.5696,
+      "step": 16455
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.406427743124346e-06,
+      "loss": 0.6498,
+      "step": 16456
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.405294833918597e-06,
+      "loss": 0.5794,
+      "step": 16457
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4041619604182745e-06,
+      "loss": 0.6388,
+      "step": 16458
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.403029122638961e-06,
+      "loss": 0.6603,
+      "step": 16459
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4018963205962515e-06,
+      "loss": 0.5575,
+      "step": 16460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.400763554305729e-06,
+      "loss": 0.5247,
+      "step": 16461
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.399630823782986e-06,
+      "loss": 0.6535,
+      "step": 16462
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.398498129043605e-06,
+      "loss": 0.5849,
+      "step": 16463
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.39736547010318e-06,
+      "loss": 0.6321,
+      "step": 16464
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.396232846977289e-06,
+      "loss": 0.6412,
+      "step": 16465
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.395100259681523e-06,
+      "loss": 0.5815,
+      "step": 16466
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.393967708231466e-06,
+      "loss": 0.6023,
+      "step": 16467
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.392835192642702e-06,
+      "loss": 0.5199,
+      "step": 16468
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.39170271293082e-06,
+      "loss": 0.5911,
+      "step": 16469
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.390570269111399e-06,
+      "loss": 0.5971,
+      "step": 16470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.389437861200024e-06,
+      "loss": 0.621,
+      "step": 16471
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3883054892122774e-06,
+      "loss": 0.64,
+      "step": 16472
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.387173153163742e-06,
+      "loss": 0.6134,
+      "step": 16473
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3860408530700004e-06,
+      "loss": 0.5775,
+      "step": 16474
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.384908588946634e-06,
+      "loss": 0.6728,
+      "step": 16475
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3837763608092215e-06,
+      "loss": 0.5526,
+      "step": 16476
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.382644168673345e-06,
+      "loss": 0.6263,
+      "step": 16477
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.381512012554587e-06,
+      "loss": 0.5845,
+      "step": 16478
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.380379892468522e-06,
+      "loss": 0.6599,
+      "step": 16479
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.379247808430734e-06,
+      "loss": 0.5576,
+      "step": 16480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3781157604567965e-06,
+      "loss": 0.6052,
+      "step": 16481
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3769837485622916e-06,
+      "loss": 0.5816,
+      "step": 16482
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.375851772762795e-06,
+      "loss": 0.6261,
+      "step": 16483
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3747198330738834e-06,
+      "loss": 0.5798,
+      "step": 16484
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3735879295111325e-06,
+      "loss": 0.6471,
+      "step": 16485
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3724560620901194e-06,
+      "loss": 0.6377,
+      "step": 16486
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.371324230826422e-06,
+      "loss": 0.6122,
+      "step": 16487
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.37019243573561e-06,
+      "loss": 0.6421,
+      "step": 16488
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.369060676833264e-06,
+      "loss": 0.5422,
+      "step": 16489
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.367928954134953e-06,
+      "loss": 0.6764,
+      "step": 16490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.366797267656254e-06,
+      "loss": 0.5421,
+      "step": 16491
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.365665617412736e-06,
+      "loss": 0.5915,
+      "step": 16492
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.364534003419976e-06,
+      "loss": 0.5911,
+      "step": 16493
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.363402425693542e-06,
+      "loss": 0.5915,
+      "step": 16494
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3622708842490095e-06,
+      "loss": 0.5777,
+      "step": 16495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.361139379101945e-06,
+      "loss": 0.6715,
+      "step": 16496
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3600079102679215e-06,
+      "loss": 0.5674,
+      "step": 16497
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.35887647776251e-06,
+      "loss": 0.5837,
+      "step": 16498
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.357745081601278e-06,
+      "loss": 0.5438,
+      "step": 16499
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.356613721799797e-06,
+      "loss": 0.581,
+      "step": 16500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.355482398373631e-06,
+      "loss": 0.6131,
+      "step": 16501
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.354351111338354e-06,
+      "loss": 0.6141,
+      "step": 16502
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.353219860709529e-06,
+      "loss": 0.5297,
+      "step": 16503
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.352088646502724e-06,
+      "loss": 0.675,
+      "step": 16504
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.350957468733506e-06,
+      "loss": 0.65,
+      "step": 16505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.349826327417442e-06,
+      "loss": 0.6197,
+      "step": 16506
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.348695222570095e-06,
+      "loss": 0.5721,
+      "step": 16507
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.34756415420703e-06,
+      "loss": 0.5944,
+      "step": 16508
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.346433122343817e-06,
+      "loss": 0.6277,
+      "step": 16509
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.345302126996013e-06,
+      "loss": 0.6298,
+      "step": 16510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.344171168179185e-06,
+      "loss": 0.5731,
+      "step": 16511
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.343040245908894e-06,
+      "loss": 0.5677,
+      "step": 16512
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.341909360200706e-06,
+      "loss": 0.5873,
+      "step": 16513
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.340778511070179e-06,
+      "loss": 0.6979,
+      "step": 16514
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3396476985328795e-06,
+      "loss": 0.6612,
+      "step": 16515
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.338516922604362e-06,
+      "loss": 0.574,
+      "step": 16516
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.337386183300189e-06,
+      "loss": 0.6313,
+      "step": 16517
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.336255480635925e-06,
+      "loss": 0.5482,
+      "step": 16518
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.335124814627124e-06,
+      "loss": 0.6007,
+      "step": 16519
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.333994185289351e-06,
+      "loss": 0.5752,
+      "step": 16520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.332863592638155e-06,
+      "loss": 0.6439,
+      "step": 16521
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.331733036689104e-06,
+      "loss": 0.6466,
+      "step": 16522
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.330602517457748e-06,
+      "loss": 0.5556,
+      "step": 16523
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.329472034959649e-06,
+      "loss": 0.6164,
+      "step": 16524
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3283415892103596e-06,
+      "loss": 0.592,
+      "step": 16525
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.32721118022544e-06,
+      "loss": 0.628,
+      "step": 16526
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.32608080802044e-06,
+      "loss": 0.6152,
+      "step": 16527
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3249504726109175e-06,
+      "loss": 0.589,
+      "step": 16528
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.32382017401243e-06,
+      "loss": 0.6406,
+      "step": 16529
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.322689912240526e-06,
+      "loss": 0.6021,
+      "step": 16530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.321559687310763e-06,
+      "loss": 0.57,
+      "step": 16531
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.32042949923869e-06,
+      "loss": 0.5704,
+      "step": 16532
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.319299348039864e-06,
+      "loss": 0.6263,
+      "step": 16533
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3181692337298326e-06,
+      "loss": 0.5921,
+      "step": 16534
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.317039156324152e-06,
+      "loss": 0.5851,
+      "step": 16535
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.315909115838367e-06,
+      "loss": 0.5543,
+      "step": 16536
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.314779112288033e-06,
+      "loss": 0.6501,
+      "step": 16537
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.313649145688695e-06,
+      "loss": 0.6133,
+      "step": 16538
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.312519216055907e-06,
+      "loss": 0.5651,
+      "step": 16539
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.311389323405218e-06,
+      "loss": 0.6405,
+      "step": 16540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3102594677521735e-06,
+      "loss": 0.5393,
+      "step": 16541
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3091296491123194e-06,
+      "loss": 0.5604,
+      "step": 16542
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.307999867501209e-06,
+      "loss": 0.635,
+      "step": 16543
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.306870122934383e-06,
+      "loss": 0.6025,
+      "step": 16544
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3057404154273925e-06,
+      "loss": 0.5465,
+      "step": 16545
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.304610744995779e-06,
+      "loss": 0.6349,
+      "step": 16546
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.303481111655093e-06,
+      "loss": 0.6401,
+      "step": 16547
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.302351515420874e-06,
+      "loss": 0.5744,
+      "step": 16548
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.301221956308667e-06,
+      "loss": 0.5966,
+      "step": 16549
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.300092434334021e-06,
+      "loss": 0.6102,
+      "step": 16550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.298962949512474e-06,
+      "loss": 0.6225,
+      "step": 16551
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.29783350185957e-06,
+      "loss": 0.5647,
+      "step": 16552
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.29670409139085e-06,
+      "loss": 0.6218,
+      "step": 16553
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.295574718121859e-06,
+      "loss": 0.5633,
+      "step": 16554
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.294445382068136e-06,
+      "loss": 0.6042,
+      "step": 16555
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.293316083245221e-06,
+      "loss": 0.5386,
+      "step": 16556
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.292186821668655e-06,
+      "loss": 0.6381,
+      "step": 16557
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.291057597353979e-06,
+      "loss": 0.5518,
+      "step": 16558
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.289928410316729e-06,
+      "loss": 0.6045,
+      "step": 16559
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.288799260572446e-06,
+      "loss": 0.5171,
+      "step": 16560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.287670148136667e-06,
+      "loss": 0.5453,
+      "step": 16561
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2865410730249306e-06,
+      "loss": 0.5454,
+      "step": 16562
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.285412035252774e-06,
+      "loss": 0.6118,
+      "step": 16563
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.284283034835731e-06,
+      "loss": 0.5792,
+      "step": 16564
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.283154071789342e-06,
+      "loss": 0.6289,
+      "step": 16565
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.282025146129139e-06,
+      "loss": 0.6432,
+      "step": 16566
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2808962578706596e-06,
+      "loss": 0.5956,
+      "step": 16567
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.279767407029435e-06,
+      "loss": 0.6468,
+      "step": 16568
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.278638593621003e-06,
+      "loss": 0.5812,
+      "step": 16569
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.277509817660893e-06,
+      "loss": 0.5676,
+      "step": 16570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.276381079164639e-06,
+      "loss": 0.6112,
+      "step": 16571
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.275252378147778e-06,
+      "loss": 0.6089,
+      "step": 16572
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.274123714625836e-06,
+      "loss": 0.6539,
+      "step": 16573
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.272995088614348e-06,
+      "loss": 0.5796,
+      "step": 16574
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.271866500128844e-06,
+      "loss": 0.5364,
+      "step": 16575
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.270737949184854e-06,
+      "loss": 0.6025,
+      "step": 16576
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.269609435797906e-06,
+      "loss": 0.5857,
+      "step": 16577
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.268480959983535e-06,
+      "loss": 0.5997,
+      "step": 16578
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.267352521757263e-06,
+      "loss": 0.5983,
+      "step": 16579
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.266224121134622e-06,
+      "loss": 0.5805,
+      "step": 16580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2650957581311395e-06,
+      "loss": 0.6151,
+      "step": 16581
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.263967432762342e-06,
+      "loss": 0.673,
+      "step": 16582
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.262839145043759e-06,
+      "loss": 0.6178,
+      "step": 16583
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.261710894990912e-06,
+      "loss": 0.6511,
+      "step": 16584
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.26058268261933e-06,
+      "loss": 0.6047,
+      "step": 16585
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2594545079445366e-06,
+      "loss": 0.5975,
+      "step": 16586
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2583263709820595e-06,
+      "loss": 0.6852,
+      "step": 16587
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.257198271747418e-06,
+      "loss": 0.5962,
+      "step": 16588
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.256070210256141e-06,
+      "loss": 0.6287,
+      "step": 16589
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.254942186523746e-06,
+      "loss": 0.6334,
+      "step": 16590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2538142005657585e-06,
+      "loss": 0.6094,
+      "step": 16591
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.252686252397703e-06,
+      "loss": 0.5965,
+      "step": 16592
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.251558342035097e-06,
+      "loss": 0.5387,
+      "step": 16593
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.250430469493465e-06,
+      "loss": 0.5206,
+      "step": 16594
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.249302634788324e-06,
+      "loss": 0.5538,
+      "step": 16595
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.248174837935197e-06,
+      "loss": 0.5795,
+      "step": 16596
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2470470789496e-06,
+      "loss": 0.5618,
+      "step": 16597
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.245919357847057e-06,
+      "loss": 0.5189,
+      "step": 16598
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2447916746430815e-06,
+      "loss": 0.6625,
+      "step": 16599
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.243664029353195e-06,
+      "loss": 0.6596,
+      "step": 16600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.242536421992912e-06,
+      "loss": 0.5972,
+      "step": 16601
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2414088525777495e-06,
+      "loss": 0.5928,
+      "step": 16602
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.240281321123228e-06,
+      "loss": 0.5475,
+      "step": 16603
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2391538276448584e-06,
+      "loss": 0.5739,
+      "step": 16604
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.23802637215816e-06,
+      "loss": 0.5638,
+      "step": 16605
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.236898954678644e-06,
+      "loss": 0.6209,
+      "step": 16606
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.235771575221828e-06,
+      "loss": 0.6075,
+      "step": 16607
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.234644233803221e-06,
+      "loss": 0.6658,
+      "step": 16608
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.233516930438343e-06,
+      "loss": 0.5705,
+      "step": 16609
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.232389665142699e-06,
+      "loss": 0.575,
+      "step": 16610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.231262437931806e-06,
+      "loss": 0.5392,
+      "step": 16611
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.230135248821178e-06,
+      "loss": 0.6631,
+      "step": 16612
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.229008097826318e-06,
+      "loss": 0.6025,
+      "step": 16613
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.227880984962745e-06,
+      "loss": 0.5964,
+      "step": 16614
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.226753910245963e-06,
+      "loss": 0.587,
+      "step": 16615
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.225626873691487e-06,
+      "loss": 0.5698,
+      "step": 16616
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.22449987531482e-06,
+      "loss": 0.5653,
+      "step": 16617
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.223372915131476e-06,
+      "loss": 0.5445,
+      "step": 16618
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.222245993156958e-06,
+      "loss": 0.6729,
+      "step": 16619
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.221119109406778e-06,
+      "loss": 0.6208,
+      "step": 16620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.219992263896438e-06,
+      "loss": 0.5477,
+      "step": 16621
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.218865456641448e-06,
+      "loss": 0.5481,
+      "step": 16622
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.217738687657316e-06,
+      "loss": 0.5806,
+      "step": 16623
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.21661195695954e-06,
+      "loss": 0.6341,
+      "step": 16624
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.215485264563633e-06,
+      "loss": 0.5799,
+      "step": 16625
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2143586104850936e-06,
+      "loss": 0.6254,
+      "step": 16626
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.213231994739429e-06,
+      "loss": 0.5608,
+      "step": 16627
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.212105417342139e-06,
+      "loss": 0.5467,
+      "step": 16628
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.21097887830873e-06,
+      "loss": 0.6306,
+      "step": 16629
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2098523776547e-06,
+      "loss": 0.6446,
+      "step": 16630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2087259153955544e-06,
+      "loss": 0.5626,
+      "step": 16631
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.207599491546791e-06,
+      "loss": 0.5885,
+      "step": 16632
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.206473106123913e-06,
+      "loss": 0.5598,
+      "step": 16633
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.205346759142422e-06,
+      "loss": 0.6108,
+      "step": 16634
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.204220450617811e-06,
+      "loss": 0.5879,
+      "step": 16635
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.203094180565589e-06,
+      "loss": 0.6172,
+      "step": 16636
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.201967949001243e-06,
+      "loss": 0.5873,
+      "step": 16637
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.20084175594028e-06,
+      "loss": 0.6212,
+      "step": 16638
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.199715601398192e-06,
+      "loss": 0.5617,
+      "step": 16639
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1985894853904795e-06,
+      "loss": 0.5995,
+      "step": 16640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.197463407932636e-06,
+      "loss": 0.5549,
+      "step": 16641
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.196337369040158e-06,
+      "loss": 0.513,
+      "step": 16642
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.195211368728543e-06,
+      "loss": 0.5392,
+      "step": 16643
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.194085407013283e-06,
+      "loss": 0.5798,
+      "step": 16644
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1929594839098736e-06,
+      "loss": 0.6379,
+      "step": 16645
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.191833599433809e-06,
+      "loss": 0.6494,
+      "step": 16646
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.190707753600581e-06,
+      "loss": 0.6274,
+      "step": 16647
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.189581946425681e-06,
+      "loss": 0.619,
+      "step": 16648
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1884561779246055e-06,
+      "loss": 0.688,
+      "step": 16649
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.187330448112841e-06,
+      "loss": 0.6708,
+      "step": 16650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.186204757005884e-06,
+      "loss": 0.5796,
+      "step": 16651
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.185079104619219e-06,
+      "loss": 0.5427,
+      "step": 16652
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.18395349096834e-06,
+      "loss": 0.587,
+      "step": 16653
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.182827916068736e-06,
+      "loss": 0.5649,
+      "step": 16654
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.181702379935894e-06,
+      "loss": 0.6065,
+      "step": 16655
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.180576882585306e-06,
+      "loss": 0.6676,
+      "step": 16656
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.179451424032455e-06,
+      "loss": 0.6051,
+      "step": 16657
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.178326004292832e-06,
+      "loss": 0.6382,
+      "step": 16658
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.177200623381921e-06,
+      "loss": 0.6447,
+      "step": 16659
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1760752813152125e-06,
+      "loss": 0.5182,
+      "step": 16660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.174949978108188e-06,
+      "loss": 0.6338,
+      "step": 16661
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.173824713776334e-06,
+      "loss": 0.5607,
+      "step": 16662
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.172699488335134e-06,
+      "loss": 0.6174,
+      "step": 16663
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.171574301800074e-06,
+      "loss": 0.5625,
+      "step": 16664
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.170449154186638e-06,
+      "loss": 0.5928,
+      "step": 16665
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.169324045510306e-06,
+      "loss": 0.5653,
+      "step": 16666
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.168198975786565e-06,
+      "loss": 0.5704,
+      "step": 16667
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.167073945030892e-06,
+      "loss": 0.5906,
+      "step": 16668
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.165948953258773e-06,
+      "loss": 0.5829,
+      "step": 16669
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.164824000485685e-06,
+      "loss": 0.6045,
+      "step": 16670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.163699086727112e-06,
+      "loss": 0.6663,
+      "step": 16671
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.162574211998529e-06,
+      "loss": 0.5842,
+      "step": 16672
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1614493763154166e-06,
+      "loss": 0.6362,
+      "step": 16673
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.160324579693259e-06,
+      "loss": 0.5577,
+      "step": 16674
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.159199822147529e-06,
+      "loss": 0.598,
+      "step": 16675
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.158075103693706e-06,
+      "loss": 0.5535,
+      "step": 16676
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.156950424347264e-06,
+      "loss": 0.5763,
+      "step": 16677
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.155825784123687e-06,
+      "loss": 0.632,
+      "step": 16678
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.154701183038442e-06,
+      "loss": 0.6444,
+      "step": 16679
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.153576621107013e-06,
+      "loss": 0.6944,
+      "step": 16680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.152452098344866e-06,
+      "loss": 0.6206,
+      "step": 16681
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.151327614767484e-06,
+      "loss": 0.6051,
+      "step": 16682
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.150203170390335e-06,
+      "loss": 0.5865,
+      "step": 16683
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.149078765228894e-06,
+      "loss": 0.5532,
+      "step": 16684
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.147954399298635e-06,
+      "loss": 0.5657,
+      "step": 16685
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1468300726150295e-06,
+      "loss": 0.5822,
+      "step": 16686
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.145705785193551e-06,
+      "loss": 0.6575,
+      "step": 16687
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.144581537049666e-06,
+      "loss": 0.5872,
+      "step": 16688
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.143457328198851e-06,
+      "loss": 0.6563,
+      "step": 16689
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.14233315865657e-06,
+      "loss": 0.5927,
+      "step": 16690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1412090284382985e-06,
+      "loss": 0.6226,
+      "step": 16691
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1400849375595e-06,
+      "loss": 0.563,
+      "step": 16692
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.138960886035649e-06,
+      "loss": 0.5995,
+      "step": 16693
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.137836873882207e-06,
+      "loss": 0.6236,
+      "step": 16694
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.136712901114645e-06,
+      "loss": 0.6253,
+      "step": 16695
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.135588967748432e-06,
+      "loss": 0.6461,
+      "step": 16696
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.134465073799028e-06,
+      "loss": 0.5692,
+      "step": 16697
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.133341219281907e-06,
+      "loss": 0.676,
+      "step": 16698
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.132217404212525e-06,
+      "loss": 0.5755,
+      "step": 16699
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.131093628606355e-06,
+      "loss": 0.6285,
+      "step": 16700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.129969892478854e-06,
+      "loss": 0.6062,
+      "step": 16701
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.128846195845493e-06,
+      "loss": 0.6238,
+      "step": 16702
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.127722538721729e-06,
+      "loss": 0.6273,
+      "step": 16703
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.126598921123025e-06,
+      "loss": 0.5752,
+      "step": 16704
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.125475343064847e-06,
+      "loss": 0.5666,
+      "step": 16705
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1243518045626535e-06,
+      "loss": 0.5767,
+      "step": 16706
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.123228305631906e-06,
+      "loss": 0.6253,
+      "step": 16707
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.122104846288065e-06,
+      "loss": 0.6148,
+      "step": 16708
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1209814265465915e-06,
+      "loss": 0.5762,
+      "step": 16709
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.11985804642294e-06,
+      "loss": 0.5772,
+      "step": 16710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.118734705932575e-06,
+      "loss": 0.591,
+      "step": 16711
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.117611405090951e-06,
+      "loss": 0.5591,
+      "step": 16712
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.116488143913528e-06,
+      "loss": 0.5658,
+      "step": 16713
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1153649224157595e-06,
+      "loss": 0.6204,
+      "step": 16714
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.114241740613105e-06,
+      "loss": 0.5977,
+      "step": 16715
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1131185985210214e-06,
+      "loss": 0.6184,
+      "step": 16716
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.111995496154961e-06,
+      "loss": 0.5568,
+      "step": 16717
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.110872433530383e-06,
+      "loss": 0.5909,
+      "step": 16718
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.109749410662734e-06,
+      "loss": 0.5857,
+      "step": 16719
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.108626427567478e-06,
+      "loss": 0.5856,
+      "step": 16720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.107503484260059e-06,
+      "loss": 0.5881,
+      "step": 16721
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.106380580755936e-06,
+      "loss": 0.6017,
+      "step": 16722
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.105257717070556e-06,
+      "loss": 0.5908,
+      "step": 16723
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.104134893219376e-06,
+      "loss": 0.6088,
+      "step": 16724
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1030121092178435e-06,
+      "loss": 0.6378,
+      "step": 16725
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.101889365081409e-06,
+      "loss": 0.5483,
+      "step": 16726
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.100766660825525e-06,
+      "loss": 0.6165,
+      "step": 16727
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.099643996465637e-06,
+      "loss": 0.6249,
+      "step": 16728
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0985213720172e-06,
+      "loss": 0.5764,
+      "step": 16729
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.097398787495654e-06,
+      "loss": 0.6092,
+      "step": 16730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.096276242916454e-06,
+      "loss": 0.5934,
+      "step": 16731
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.095153738295043e-06,
+      "loss": 0.6066,
+      "step": 16732
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.094031273646871e-06,
+      "loss": 0.564,
+      "step": 16733
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.092908848987379e-06,
+      "loss": 0.6171,
+      "step": 16734
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.091786464332015e-06,
+      "loss": 0.6382,
+      "step": 16735
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.090664119696229e-06,
+      "loss": 0.6441,
+      "step": 16736
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0895418150954575e-06,
+      "loss": 0.6577,
+      "step": 16737
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.088419550545151e-06,
+      "loss": 0.5674,
+      "step": 16738
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.087297326060747e-06,
+      "loss": 0.605,
+      "step": 16739
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.086175141657694e-06,
+      "loss": 0.5923,
+      "step": 16740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.085052997351428e-06,
+      "loss": 0.6535,
+      "step": 16741
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.083930893157397e-06,
+      "loss": 0.6206,
+      "step": 16742
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0828088290910365e-06,
+      "loss": 0.5342,
+      "step": 16743
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.081686805167792e-06,
+      "loss": 0.6614,
+      "step": 16744
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.080564821403099e-06,
+      "loss": 0.54,
+      "step": 16745
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.079442877812401e-06,
+      "loss": 0.6675,
+      "step": 16746
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.078320974411136e-06,
+      "loss": 0.6289,
+      "step": 16747
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0771991112147386e-06,
+      "loss": 0.543,
+      "step": 16748
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.076077288238654e-06,
+      "loss": 0.6042,
+      "step": 16749
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.074955505498315e-06,
+      "loss": 0.5824,
+      "step": 16750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.073833763009156e-06,
+      "loss": 0.6154,
+      "step": 16751
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.072712060786618e-06,
+      "loss": 0.6278,
+      "step": 16752
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.071590398846129e-06,
+      "loss": 0.5725,
+      "step": 16753
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.070468777203135e-06,
+      "loss": 0.5512,
+      "step": 16754
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.069347195873059e-06,
+      "loss": 0.5581,
+      "step": 16755
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.068225654871346e-06,
+      "loss": 0.5875,
+      "step": 16756
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.06710415421342e-06,
+      "loss": 0.5999,
+      "step": 16757
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.065982693914719e-06,
+      "loss": 0.5662,
+      "step": 16758
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.064861273990675e-06,
+      "loss": 0.729,
+      "step": 16759
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.063739894456718e-06,
+      "loss": 0.566,
+      "step": 16760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0626185553282825e-06,
+      "loss": 0.566,
+      "step": 16761
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.061497256620793e-06,
+      "loss": 0.5628,
+      "step": 16762
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.060375998349686e-06,
+      "loss": 0.6804,
+      "step": 16763
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.059254780530387e-06,
+      "loss": 0.587,
+      "step": 16764
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.058133603178328e-06,
+      "loss": 0.6084,
+      "step": 16765
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.057012466308932e-06,
+      "loss": 0.6089,
+      "step": 16766
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.055891369937634e-06,
+      "loss": 0.6429,
+      "step": 16767
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.054770314079855e-06,
+      "loss": 0.5785,
+      "step": 16768
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.053649298751026e-06,
+      "loss": 0.5394,
+      "step": 16769
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.052528323966572e-06,
+      "loss": 0.6219,
+      "step": 16770
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.051407389741916e-06,
+      "loss": 0.5592,
+      "step": 16771
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.050286496092487e-06,
+      "loss": 0.592,
+      "step": 16772
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.049165643033706e-06,
+      "loss": 0.6013,
+      "step": 16773
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.048044830581001e-06,
+      "loss": 0.6232,
+      "step": 16774
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.046924058749791e-06,
+      "loss": 0.5901,
+      "step": 16775
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.045803327555502e-06,
+      "loss": 0.5814,
+      "step": 16776
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0446826370135535e-06,
+      "loss": 0.5961,
+      "step": 16777
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.043561987139369e-06,
+      "loss": 0.5332,
+      "step": 16778
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0424413779483705e-06,
+      "loss": 0.5925,
+      "step": 16779
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.041320809455976e-06,
+      "loss": 0.6235,
+      "step": 16780
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.040200281677609e-06,
+      "loss": 0.6954,
+      "step": 16781
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.039079794628685e-06,
+      "loss": 0.5844,
+      "step": 16782
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.037959348324626e-06,
+      "loss": 0.6113,
+      "step": 16783
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.036838942780846e-06,
+      "loss": 0.5848,
+      "step": 16784
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.03571857801277e-06,
+      "loss": 0.6259,
+      "step": 16785
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.034598254035807e-06,
+      "loss": 0.6621,
+      "step": 16786
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.033477970865381e-06,
+      "loss": 0.6302,
+      "step": 16787
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.032357728516903e-06,
+      "loss": 0.6638,
+      "step": 16788
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0312375270057895e-06,
+      "loss": 0.6026,
+      "step": 16789
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.030117366347458e-06,
+      "loss": 0.5637,
+      "step": 16790
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.028997246557321e-06,
+      "loss": 0.5434,
+      "step": 16791
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0278771676507915e-06,
+      "loss": 0.629,
+      "step": 16792
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.026757129643284e-06,
+      "loss": 0.5729,
+      "step": 16793
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.025637132550213e-06,
+      "loss": 0.6215,
+      "step": 16794
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.024517176386985e-06,
+      "loss": 0.5763,
+      "step": 16795
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.023397261169018e-06,
+      "loss": 0.6268,
+      "step": 16796
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.022277386911718e-06,
+      "loss": 0.6056,
+      "step": 16797
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0211575536305e-06,
+      "loss": 0.6573,
+      "step": 16798
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.020037761340769e-06,
+      "loss": 0.5858,
+      "step": 16799
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.018918010057936e-06,
+      "loss": 0.6022,
+      "step": 16800
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.017798299797414e-06,
+      "loss": 0.5731,
+      "step": 16801
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.016678630574605e-06,
+      "loss": 0.5221,
+      "step": 16802
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.015559002404921e-06,
+      "loss": 0.6175,
+      "step": 16803
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.014439415303768e-06,
+      "loss": 0.5748,
+      "step": 16804
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.013319869286552e-06,
+      "loss": 0.6018,
+      "step": 16805
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.012200364368677e-06,
+      "loss": 0.5734,
+      "step": 16806
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.011080900565552e-06,
+      "loss": 0.6085,
+      "step": 16807
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0099614778925785e-06,
+      "loss": 0.6311,
+      "step": 16808
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.008842096365163e-06,
+      "loss": 0.6483,
+      "step": 16809
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.007722755998709e-06,
+      "loss": 0.6107,
+      "step": 16810
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.006603456808618e-06,
+      "loss": 0.633,
+      "step": 16811
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.005484198810296e-06,
+      "loss": 0.6402,
+      "step": 16812
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.004364982019141e-06,
+      "loss": 0.662,
+      "step": 16813
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.003245806450557e-06,
+      "loss": 0.5861,
+      "step": 16814
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.002126672119942e-06,
+      "loss": 0.6762,
+      "step": 16815
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0010075790427e-06,
+      "loss": 0.551,
+      "step": 16816
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9998885272342276e-06,
+      "loss": 0.5389,
+      "step": 16817
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.998769516709927e-06,
+      "loss": 0.5969,
+      "step": 16818
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.997650547485193e-06,
+      "loss": 0.5849,
+      "step": 16819
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.996531619575425e-06,
+      "loss": 0.5473,
+      "step": 16820
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9954127329960235e-06,
+      "loss": 0.5524,
+      "step": 16821
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.99429388776238e-06,
+      "loss": 0.6397,
+      "step": 16822
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.993175083889897e-06,
+      "loss": 0.5634,
+      "step": 16823
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.992056321393964e-06,
+      "loss": 0.5693,
+      "step": 16824
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.990937600289982e-06,
+      "loss": 0.6359,
+      "step": 16825
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9898189205933385e-06,
+      "loss": 0.5608,
+      "step": 16826
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.988700282319436e-06,
+      "loss": 0.6076,
+      "step": 16827
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.987581685483659e-06,
+      "loss": 0.6214,
+      "step": 16828
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.986463130101411e-06,
+      "loss": 0.6558,
+      "step": 16829
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9853446161880726e-06,
+      "loss": 0.5651,
+      "step": 16830
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.984226143759043e-06,
+      "loss": 0.6043,
+      "step": 16831
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.983107712829713e-06,
+      "loss": 0.6291,
+      "step": 16832
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.981989323415469e-06,
+      "loss": 0.6319,
+      "step": 16833
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.980870975531706e-06,
+      "loss": 0.5654,
+      "step": 16834
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.979752669193811e-06,
+      "loss": 0.6008,
+      "step": 16835
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9786344044171725e-06,
+      "loss": 0.5737,
+      "step": 16836
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.977516181217179e-06,
+      "loss": 0.6363,
+      "step": 16837
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9763979996092215e-06,
+      "loss": 0.6067,
+      "step": 16838
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.97527985960868e-06,
+      "loss": 0.6103,
+      "step": 16839
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.974161761230947e-06,
+      "loss": 0.6145,
+      "step": 16840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.973043704491408e-06,
+      "loss": 0.6027,
+      "step": 16841
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9719256894054455e-06,
+      "loss": 0.6374,
+      "step": 16842
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.970807715988448e-06,
+      "loss": 0.6142,
+      "step": 16843
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.969689784255798e-06,
+      "loss": 0.5934,
+      "step": 16844
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9685718942228785e-06,
+      "loss": 0.603,
+      "step": 16845
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.967454045905073e-06,
+      "loss": 0.6024,
+      "step": 16846
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.966336239317766e-06,
+      "loss": 0.6199,
+      "step": 16847
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.965218474476336e-06,
+      "loss": 0.6506,
+      "step": 16848
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.964100751396168e-06,
+      "loss": 0.5146,
+      "step": 16849
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.962983070092639e-06,
+      "loss": 0.5453,
+      "step": 16850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.961865430581133e-06,
+      "loss": 0.5944,
+      "step": 16851
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.960747832877028e-06,
+      "loss": 0.6481,
+      "step": 16852
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.959630276995704e-06,
+      "loss": 0.5786,
+      "step": 16853
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.958512762952539e-06,
+      "loss": 0.659,
+      "step": 16854
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.957395290762909e-06,
+      "loss": 0.5957,
+      "step": 16855
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9562778604421965e-06,
+      "loss": 0.5075,
+      "step": 16856
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.955160472005772e-06,
+      "loss": 0.6894,
+      "step": 16857
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.954043125469017e-06,
+      "loss": 0.6611,
+      "step": 16858
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.952925820847302e-06,
+      "loss": 0.6526,
+      "step": 16859
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.95180855815601e-06,
+      "loss": 0.6626,
+      "step": 16860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9506913374105065e-06,
+      "loss": 0.5954,
+      "step": 16861
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.949574158626169e-06,
+      "loss": 0.6407,
+      "step": 16862
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.948457021818376e-06,
+      "loss": 0.5785,
+      "step": 16863
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.947339927002491e-06,
+      "loss": 0.5929,
+      "step": 16864
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.946222874193893e-06,
+      "loss": 0.5385,
+      "step": 16865
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.94510586340795e-06,
+      "loss": 0.6224,
+      "step": 16866
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.943988894660039e-06,
+      "loss": 0.6245,
+      "step": 16867
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.942871967965521e-06,
+      "loss": 0.5748,
+      "step": 16868
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.941755083339774e-06,
+      "loss": 0.5924,
+      "step": 16869
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.940638240798163e-06,
+      "loss": 0.5767,
+      "step": 16870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.939521440356056e-06,
+      "loss": 0.584,
+      "step": 16871
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.938404682028827e-06,
+      "loss": 0.588,
+      "step": 16872
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.937287965831838e-06,
+      "loss": 0.6039,
+      "step": 16873
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.936171291780458e-06,
+      "loss": 0.606,
+      "step": 16874
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.935054659890053e-06,
+      "loss": 0.6196,
+      "step": 16875
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.93393807017599e-06,
+      "loss": 0.6212,
+      "step": 16876
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.93282152265363e-06,
+      "loss": 0.6166,
+      "step": 16877
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.931705017338345e-06,
+      "loss": 0.5588,
+      "step": 16878
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.930588554245492e-06,
+      "loss": 0.5777,
+      "step": 16879
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.929472133390441e-06,
+      "loss": 0.5972,
+      "step": 16880
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.928355754788547e-06,
+      "loss": 0.6754,
+      "step": 16881
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9272394184551785e-06,
+      "loss": 0.6295,
+      "step": 16882
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.926123124405696e-06,
+      "loss": 0.5304,
+      "step": 16883
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.92500687265546e-06,
+      "loss": 0.5502,
+      "step": 16884
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.923890663219832e-06,
+      "loss": 0.5881,
+      "step": 16885
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.92277449611417e-06,
+      "loss": 0.6039,
+      "step": 16886
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.921658371353836e-06,
+      "loss": 0.6408,
+      "step": 16887
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.920542288954186e-06,
+      "loss": 0.5725,
+      "step": 16888
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.919426248930583e-06,
+      "loss": 0.6068,
+      "step": 16889
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.918310251298379e-06,
+      "loss": 0.6677,
+      "step": 16890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.917194296072936e-06,
+      "loss": 0.5844,
+      "step": 16891
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.916078383269607e-06,
+      "loss": 0.5707,
+      "step": 16892
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.914962512903748e-06,
+      "loss": 0.6014,
+      "step": 16893
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.913846684990717e-06,
+      "loss": 0.6128,
+      "step": 16894
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.912730899545867e-06,
+      "loss": 0.6105,
+      "step": 16895
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.911615156584554e-06,
+      "loss": 0.6604,
+      "step": 16896
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.910499456122128e-06,
+      "loss": 0.5152,
+      "step": 16897
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.909383798173945e-06,
+      "loss": 0.6433,
+      "step": 16898
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.908268182755356e-06,
+      "loss": 0.6472,
+      "step": 16899
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.907152609881716e-06,
+      "loss": 0.5625,
+      "step": 16900
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.906037079568371e-06,
+      "loss": 0.5715,
+      "step": 16901
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.904921591830673e-06,
+      "loss": 0.5259,
+      "step": 16902
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9038061466839765e-06,
+      "loss": 0.5247,
+      "step": 16903
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.902690744143627e-06,
+      "loss": 0.5765,
+      "step": 16904
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.901575384224974e-06,
+      "loss": 0.5739,
+      "step": 16905
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.900460066943366e-06,
+      "loss": 0.573,
+      "step": 16906
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8993447923141535e-06,
+      "loss": 0.5712,
+      "step": 16907
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.898229560352678e-06,
+      "loss": 0.5816,
+      "step": 16908
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.897114371074292e-06,
+      "loss": 0.6153,
+      "step": 16909
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.895999224494336e-06,
+      "loss": 0.5831,
+      "step": 16910
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.89488412062816e-06,
+      "loss": 0.6243,
+      "step": 16911
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.893769059491106e-06,
+      "loss": 0.6136,
+      "step": 16912
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.892654041098517e-06,
+      "loss": 0.6235,
+      "step": 16913
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8915390654657425e-06,
+      "loss": 0.5846,
+      "step": 16914
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.89042413260812e-06,
+      "loss": 0.5948,
+      "step": 16915
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8893092425409946e-06,
+      "loss": 0.6718,
+      "step": 16916
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.888194395279706e-06,
+      "loss": 0.5464,
+      "step": 16917
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.887079590839598e-06,
+      "loss": 0.5262,
+      "step": 16918
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.88596482923601e-06,
+      "loss": 0.5799,
+      "step": 16919
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.884850110484282e-06,
+      "loss": 0.5752,
+      "step": 16920
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.883735434599752e-06,
+      "loss": 0.5825,
+      "step": 16921
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.882620801597763e-06,
+      "loss": 0.6178,
+      "step": 16922
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.881506211493649e-06,
+      "loss": 0.6081,
+      "step": 16923
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.88039166430275e-06,
+      "loss": 0.5394,
+      "step": 16924
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.879277160040405e-06,
+      "loss": 0.5306,
+      "step": 16925
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.878162698721946e-06,
+      "loss": 0.5814,
+      "step": 16926
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.877048280362713e-06,
+      "loss": 0.5854,
+      "step": 16927
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.875933904978038e-06,
+      "loss": 0.5469,
+      "step": 16928
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.87481957258326e-06,
+      "loss": 0.6436,
+      "step": 16929
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.873705283193708e-06,
+      "loss": 0.6625,
+      "step": 16930
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8725910368247206e-06,
+      "loss": 0.5409,
+      "step": 16931
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.871476833491627e-06,
+      "loss": 0.5659,
+      "step": 16932
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.870362673209761e-06,
+      "loss": 0.5732,
+      "step": 16933
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.869248555994457e-06,
+      "loss": 0.5969,
+      "step": 16934
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.86813448186104e-06,
+      "loss": 0.6011,
+      "step": 16935
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.867020450824847e-06,
+      "loss": 0.5912,
+      "step": 16936
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.865906462901204e-06,
+      "loss": 0.6154,
+      "step": 16937
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.864792518105444e-06,
+      "loss": 0.5169,
+      "step": 16938
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.863678616452891e-06,
+      "loss": 0.5761,
+      "step": 16939
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8625647579588795e-06,
+      "loss": 0.5952,
+      "step": 16940
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.861450942638731e-06,
+      "loss": 0.5731,
+      "step": 16941
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.860337170507777e-06,
+      "loss": 0.6502,
+      "step": 16942
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.859223441581341e-06,
+      "loss": 0.6413,
+      "step": 16943
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.858109755874749e-06,
+      "loss": 0.6088,
+      "step": 16944
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.85699611340333e-06,
+      "loss": 0.6618,
+      "step": 16945
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.855882514182405e-06,
+      "loss": 0.6094,
+      "step": 16946
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8547689582273e-06,
+      "loss": 0.6189,
+      "step": 16947
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.853655445553336e-06,
+      "loss": 0.6059,
+      "step": 16948
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8525419761758405e-06,
+      "loss": 0.6054,
+      "step": 16949
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.851428550110131e-06,
+      "loss": 0.7,
+      "step": 16950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.850315167371533e-06,
+      "loss": 0.5806,
+      "step": 16951
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.849201827975362e-06,
+      "loss": 0.5931,
+      "step": 16952
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.848088531936947e-06,
+      "loss": 0.6502,
+      "step": 16953
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.846975279271599e-06,
+      "loss": 0.6462,
+      "step": 16954
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.845862069994644e-06,
+      "loss": 0.5733,
+      "step": 16955
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.844748904121399e-06,
+      "loss": 0.6286,
+      "step": 16956
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.843635781667179e-06,
+      "loss": 0.6808,
+      "step": 16957
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.842522702647307e-06,
+      "loss": 0.5886,
+      "step": 16958
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.841409667077096e-06,
+      "loss": 0.6308,
+      "step": 16959
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.840296674971864e-06,
+      "loss": 0.593,
+      "step": 16960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.83918372634693e-06,
+      "loss": 0.5779,
+      "step": 16961
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.838070821217598e-06,
+      "loss": 0.5987,
+      "step": 16962
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.836957959599194e-06,
+      "loss": 0.5311,
+      "step": 16963
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.835845141507025e-06,
+      "loss": 0.6012,
+      "step": 16964
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.834732366956407e-06,
+      "loss": 0.5343,
+      "step": 16965
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.833619635962656e-06,
+      "loss": 0.6274,
+      "step": 16966
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8325069485410775e-06,
+      "loss": 0.5302,
+      "step": 16967
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8313943047069905e-06,
+      "loss": 0.5685,
+      "step": 16968
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.830281704475699e-06,
+      "loss": 0.5578,
+      "step": 16969
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.829169147862517e-06,
+      "loss": 0.6403,
+      "step": 16970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.828056634882754e-06,
+      "loss": 0.5571,
+      "step": 16971
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.826944165551719e-06,
+      "loss": 0.5883,
+      "step": 16972
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.82583173988472e-06,
+      "loss": 0.6045,
+      "step": 16973
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.824719357897066e-06,
+      "loss": 0.6131,
+      "step": 16974
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.823607019604063e-06,
+      "loss": 0.6474,
+      "step": 16975
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.822494725021017e-06,
+      "loss": 0.5698,
+      "step": 16976
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.821382474163239e-06,
+      "loss": 0.6645,
+      "step": 16977
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.820270267046027e-06,
+      "loss": 0.6547,
+      "step": 16978
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.819158103684693e-06,
+      "loss": 0.5972,
+      "step": 16979
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.818045984094538e-06,
+      "loss": 0.5434,
+      "step": 16980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.816933908290867e-06,
+      "loss": 0.6317,
+      "step": 16981
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.815821876288979e-06,
+      "loss": 0.6306,
+      "step": 16982
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8147098881041855e-06,
+      "loss": 0.5762,
+      "step": 16983
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8135979437517775e-06,
+      "loss": 0.6634,
+      "step": 16984
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.812486043247064e-06,
+      "loss": 0.639,
+      "step": 16985
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.811374186605342e-06,
+      "loss": 0.6073,
+      "step": 16986
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.810262373841914e-06,
+      "loss": 0.5428,
+      "step": 16987
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.809150604972079e-06,
+      "loss": 0.5866,
+      "step": 16988
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.808038880011133e-06,
+      "loss": 0.5973,
+      "step": 16989
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.80692719897438e-06,
+      "loss": 0.6465,
+      "step": 16990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.805815561877111e-06,
+      "loss": 0.6191,
+      "step": 16991
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.804703968734629e-06,
+      "loss": 0.6534,
+      "step": 16992
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.803592419562228e-06,
+      "loss": 0.5587,
+      "step": 16993
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.802480914375205e-06,
+      "loss": 0.5648,
+      "step": 16994
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.801369453188851e-06,
+      "loss": 0.6418,
+      "step": 16995
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.800258036018463e-06,
+      "loss": 0.5912,
+      "step": 16996
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.79914666287934e-06,
+      "loss": 0.6204,
+      "step": 16997
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.798035333786769e-06,
+      "loss": 0.6057,
+      "step": 16998
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.796924048756047e-06,
+      "loss": 0.6095,
+      "step": 16999
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.795812807802462e-06,
+      "loss": 0.6266,
+      "step": 17000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.794701610941312e-06,
+      "loss": 0.6027,
+      "step": 17001
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.793590458187881e-06,
+      "loss": 0.557,
+      "step": 17002
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.792479349557465e-06,
+      "loss": 0.5584,
+      "step": 17003
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.791368285065349e-06,
+      "loss": 0.659,
+      "step": 17004
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.790257264726827e-06,
+      "loss": 0.5133,
+      "step": 17005
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.789146288557184e-06,
+      "loss": 0.592,
+      "step": 17006
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.788035356571709e-06,
+      "loss": 0.6116,
+      "step": 17007
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.786924468785691e-06,
+      "loss": 0.601,
+      "step": 17008
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.785813625214415e-06,
+      "loss": 0.6041,
+      "step": 17009
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.784702825873168e-06,
+      "loss": 0.6085,
+      "step": 17010
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.783592070777234e-06,
+      "loss": 0.6405,
+      "step": 17011
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.782481359941903e-06,
+      "loss": 0.6141,
+      "step": 17012
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.781370693382452e-06,
+      "loss": 0.6308,
+      "step": 17013
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7802600711141705e-06,
+      "loss": 0.659,
+      "step": 17014
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.779149493152337e-06,
+      "loss": 0.5504,
+      "step": 17015
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7780389595122395e-06,
+      "loss": 0.6015,
+      "step": 17016
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.776928470209154e-06,
+      "loss": 0.6356,
+      "step": 17017
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.775818025258366e-06,
+      "loss": 0.559,
+      "step": 17018
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.774707624675157e-06,
+      "loss": 0.6133,
+      "step": 17019
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.773597268474802e-06,
+      "loss": 0.5818,
+      "step": 17020
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.772486956672586e-06,
+      "loss": 0.5442,
+      "step": 17021
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.771376689283784e-06,
+      "loss": 0.6544,
+      "step": 17022
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.770266466323677e-06,
+      "loss": 0.5736,
+      "step": 17023
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.769156287807539e-06,
+      "loss": 0.5781,
+      "step": 17024
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.768046153750653e-06,
+      "loss": 0.5818,
+      "step": 17025
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7669360641682876e-06,
+      "loss": 0.6224,
+      "step": 17026
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.765826019075724e-06,
+      "loss": 0.5983,
+      "step": 17027
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7647160184882375e-06,
+      "loss": 0.5361,
+      "step": 17028
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7636060624211e-06,
+      "loss": 0.661,
+      "step": 17029
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7624961508895894e-06,
+      "loss": 0.63,
+      "step": 17030
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.761386283908973e-06,
+      "loss": 0.5675,
+      "step": 17031
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.76027646149453e-06,
+      "loss": 0.615,
+      "step": 17032
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.759166683661527e-06,
+      "loss": 0.6251,
+      "step": 17033
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.758056950425241e-06,
+      "loss": 0.6285,
+      "step": 17034
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7569472618009364e-06,
+      "loss": 0.5575,
+      "step": 17035
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.755837617803891e-06,
+      "loss": 0.6076,
+      "step": 17036
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.754728018449368e-06,
+      "loss": 0.5635,
+      "step": 17037
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.753618463752638e-06,
+      "loss": 0.6467,
+      "step": 17038
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.752508953728973e-06,
+      "loss": 0.5482,
+      "step": 17039
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.751399488393637e-06,
+      "loss": 0.6125,
+      "step": 17040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7502900677619e-06,
+      "loss": 0.5643,
+      "step": 17041
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.749180691849024e-06,
+      "loss": 0.5356,
+      "step": 17042
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.748071360670281e-06,
+      "loss": 0.6271,
+      "step": 17043
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.746962074240933e-06,
+      "loss": 0.5723,
+      "step": 17044
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.745852832576245e-06,
+      "loss": 0.6315,
+      "step": 17045
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7447436356914796e-06,
+      "loss": 0.5818,
+      "step": 17046
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.743634483601904e-06,
+      "loss": 0.6312,
+      "step": 17047
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.74252537632278e-06,
+      "loss": 0.6382,
+      "step": 17048
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7414163138693655e-06,
+      "loss": 0.5972,
+      "step": 17049
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7403072962569295e-06,
+      "loss": 0.6104,
+      "step": 17050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.739198323500727e-06,
+      "loss": 0.5651,
+      "step": 17051
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.738089395616023e-06,
+      "loss": 0.6658,
+      "step": 17052
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.736980512618072e-06,
+      "loss": 0.5817,
+      "step": 17053
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.73587167452214e-06,
+      "loss": 0.6476,
+      "step": 17054
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7347628813434784e-06,
+      "loss": 0.5485,
+      "step": 17055
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.733654133097351e-06,
+      "loss": 0.5531,
+      "step": 17056
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.732545429799011e-06,
+      "loss": 0.6179,
+      "step": 17057
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.731436771463718e-06,
+      "loss": 0.6045,
+      "step": 17058
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.730328158106727e-06,
+      "loss": 0.6152,
+      "step": 17059
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.729219589743292e-06,
+      "loss": 0.607,
+      "step": 17060
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.728111066388673e-06,
+      "loss": 0.6141,
+      "step": 17061
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.727002588058118e-06,
+      "loss": 0.6316,
+      "step": 17062
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.725894154766884e-06,
+      "loss": 0.6016,
+      "step": 17063
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.724785766530222e-06,
+      "loss": 0.6171,
+      "step": 17064
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.723677423363389e-06,
+      "loss": 0.5706,
+      "step": 17065
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.72256912528163e-06,
+      "loss": 0.594,
+      "step": 17066
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7214608723002015e-06,
+      "loss": 0.6095,
+      "step": 17067
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.720352664434351e-06,
+      "loss": 0.6127,
+      "step": 17068
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.719244501699329e-06,
+      "loss": 0.6117,
+      "step": 17069
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.718136384110388e-06,
+      "loss": 0.5632,
+      "step": 17070
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.717028311682771e-06,
+      "loss": 0.5875,
+      "step": 17071
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7159202844317325e-06,
+      "loss": 0.617,
+      "step": 17072
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7148123023725145e-06,
+      "loss": 0.5779,
+      "step": 17073
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.713704365520366e-06,
+      "loss": 0.5394,
+      "step": 17074
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.712596473890533e-06,
+      "loss": 0.5617,
+      "step": 17075
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.711488627498262e-06,
+      "loss": 0.6063,
+      "step": 17076
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.710380826358795e-06,
+      "loss": 0.6349,
+      "step": 17077
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.709273070487381e-06,
+      "loss": 0.5325,
+      "step": 17078
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.708165359899259e-06,
+      "loss": 0.5713,
+      "step": 17079
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.707057694609673e-06,
+      "loss": 0.6421,
+      "step": 17080
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70595007463387e-06,
+      "loss": 0.6387,
+      "step": 17081
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7048424999870855e-06,
+      "loss": 0.5894,
+      "step": 17082
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7037349706845656e-06,
+      "loss": 0.5739,
+      "step": 17083
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.702627486741548e-06,
+      "loss": 0.5848,
+      "step": 17084
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.701520048173274e-06,
+      "loss": 0.5999,
+      "step": 17085
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7004126549949814e-06,
+      "loss": 0.5562,
+      "step": 17086
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.699305307221912e-06,
+      "loss": 0.59,
+      "step": 17087
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6981980048692985e-06,
+      "loss": 0.5834,
+      "step": 17088
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.697090747952383e-06,
+      "loss": 0.4959,
+      "step": 17089
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.695983536486403e-06,
+      "loss": 0.5552,
+      "step": 17090
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.69487637048659e-06,
+      "loss": 0.6129,
+      "step": 17091
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.693769249968185e-06,
+      "loss": 0.5331,
+      "step": 17092
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.692662174946417e-06,
+      "loss": 0.596,
+      "step": 17093
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.691555145436526e-06,
+      "loss": 0.567,
+      "step": 17094
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.690448161453742e-06,
+      "loss": 0.6716,
+      "step": 17095
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.689341223013302e-06,
+      "loss": 0.5897,
+      "step": 17096
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.688234330130432e-06,
+      "loss": 0.5929,
+      "step": 17097
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.68712748282037e-06,
+      "loss": 0.6282,
+      "step": 17098
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.686020681098343e-06,
+      "loss": 0.6407,
+      "step": 17099
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.684913924979584e-06,
+      "loss": 0.6154,
+      "step": 17100
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.683807214479323e-06,
+      "loss": 0.6576,
+      "step": 17101
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6827005496127874e-06,
+      "loss": 0.5905,
+      "step": 17102
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.681593930395209e-06,
+      "loss": 0.6045,
+      "step": 17103
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.680487356841812e-06,
+      "loss": 0.5817,
+      "step": 17104
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.679380828967828e-06,
+      "loss": 0.5516,
+      "step": 17105
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.678274346788478e-06,
+      "loss": 0.6014,
+      "step": 17106
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.677167910318996e-06,
+      "loss": 0.5909,
+      "step": 17107
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.676061519574599e-06,
+      "loss": 0.6237,
+      "step": 17108
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.67495517457052e-06,
+      "loss": 0.5169,
+      "step": 17109
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.673848875321977e-06,
+      "loss": 0.5842,
+      "step": 17110
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6727426218441945e-06,
+      "loss": 0.6316,
+      "step": 17111
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.671636414152401e-06,
+      "loss": 0.657,
+      "step": 17112
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.670530252261813e-06,
+      "loss": 0.571,
+      "step": 17113
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6694241361876546e-06,
+      "loss": 0.5309,
+      "step": 17114
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.668318065945146e-06,
+      "loss": 0.6365,
+      "step": 17115
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.66721204154951e-06,
+      "loss": 0.625,
+      "step": 17116
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.666106063015962e-06,
+      "loss": 0.5554,
+      "step": 17117
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.665000130359726e-06,
+      "loss": 0.5818,
+      "step": 17118
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.663894243596018e-06,
+      "loss": 0.5553,
+      "step": 17119
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.662788402740057e-06,
+      "loss": 0.5512,
+      "step": 17120
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.661682607807061e-06,
+      "loss": 0.5514,
+      "step": 17121
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.660576858812242e-06,
+      "loss": 0.6375,
+      "step": 17122
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.659471155770824e-06,
+      "loss": 0.5929,
+      "step": 17123
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.658365498698015e-06,
+      "loss": 0.6135,
+      "step": 17124
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.657259887609036e-06,
+      "loss": 0.5634,
+      "step": 17125
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.656154322519095e-06,
+      "loss": 0.5858,
+      "step": 17126
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.65504880344341e-06,
+      "loss": 0.5937,
+      "step": 17127
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.653943330397193e-06,
+      "loss": 0.5215,
+      "step": 17128
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.652837903395655e-06,
+      "loss": 0.5873,
+      "step": 17129
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.651732522454008e-06,
+      "loss": 0.5907,
+      "step": 17130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.650627187587464e-06,
+      "loss": 0.6088,
+      "step": 17131
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.649521898811234e-06,
+      "loss": 0.6225,
+      "step": 17132
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.648416656140525e-06,
+      "loss": 0.6105,
+      "step": 17133
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6473114595905506e-06,
+      "loss": 0.5503,
+      "step": 17134
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.646206309176513e-06,
+      "loss": 0.6367,
+      "step": 17135
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6451012049136265e-06,
+      "loss": 0.5346,
+      "step": 17136
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.643996146817093e-06,
+      "loss": 0.6316,
+      "step": 17137
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.642891134902123e-06,
+      "loss": 0.5951,
+      "step": 17138
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6417861691839194e-06,
+      "loss": 0.6586,
+      "step": 17139
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.640681249677691e-06,
+      "loss": 0.6129,
+      "step": 17140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.639576376398637e-06,
+      "loss": 0.6403,
+      "step": 17141
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.638471549361967e-06,
+      "loss": 0.6369,
+      "step": 17142
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.637366768582882e-06,
+      "loss": 0.672,
+      "step": 17143
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.636262034076584e-06,
+      "loss": 0.6537,
+      "step": 17144
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.635157345858278e-06,
+      "loss": 0.5696,
+      "step": 17145
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.634052703943162e-06,
+      "loss": 0.5449,
+      "step": 17146
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.63294810834644e-06,
+      "loss": 0.5334,
+      "step": 17147
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.631843559083309e-06,
+      "loss": 0.6052,
+      "step": 17148
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.630739056168971e-06,
+      "loss": 0.5943,
+      "step": 17149
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.629634599618622e-06,
+      "loss": 0.6129,
+      "step": 17150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.628530189447463e-06,
+      "loss": 0.589,
+      "step": 17151
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.627425825670693e-06,
+      "loss": 0.5492,
+      "step": 17152
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6263215083035055e-06,
+      "loss": 0.6458,
+      "step": 17153
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.625217237361099e-06,
+      "loss": 0.6229,
+      "step": 17154
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.624113012858669e-06,
+      "loss": 0.5903,
+      "step": 17155
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6230088348114106e-06,
+      "loss": 0.6051,
+      "step": 17156
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.621904703234516e-06,
+      "loss": 0.6289,
+      "step": 17157
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.620800618143184e-06,
+      "loss": 0.5651,
+      "step": 17158
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.619696579552602e-06,
+      "loss": 0.5223,
+      "step": 17159
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.618592587477966e-06,
+      "loss": 0.5609,
+      "step": 17160
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.617488641934467e-06,
+      "loss": 0.5861,
+      "step": 17161
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.616384742937295e-06,
+      "loss": 0.5103,
+      "step": 17162
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.615280890501645e-06,
+      "loss": 0.6395,
+      "step": 17163
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.614177084642701e-06,
+      "loss": 0.5844,
+      "step": 17164
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.613073325375658e-06,
+      "loss": 0.5478,
+      "step": 17165
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6119696127157e-06,
+      "loss": 0.6059,
+      "step": 17166
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.61086594667802e-06,
+      "loss": 0.6469,
+      "step": 17167
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6097623272778e-06,
+      "loss": 0.5896,
+      "step": 17168
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.608658754530232e-06,
+      "loss": 0.6313,
+      "step": 17169
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.607555228450496e-06,
+      "loss": 0.6678,
+      "step": 17170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.606451749053784e-06,
+      "loss": 0.586,
+      "step": 17171
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.60534831635528e-06,
+      "loss": 0.5463,
+      "step": 17172
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6042449303701605e-06,
+      "loss": 0.6149,
+      "step": 17173
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.603141591113615e-06,
+      "loss": 0.5281,
+      "step": 17174
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.602038298600828e-06,
+      "loss": 0.6149,
+      "step": 17175
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.600935052846978e-06,
+      "loss": 0.6576,
+      "step": 17176
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.599831853867251e-06,
+      "loss": 0.5912,
+      "step": 17177
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.598728701676822e-06,
+      "loss": 0.5383,
+      "step": 17178
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.597625596290877e-06,
+      "loss": 0.6627,
+      "step": 17179
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.596522537724592e-06,
+      "loss": 0.6387,
+      "step": 17180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5954195259931495e-06,
+      "loss": 0.6287,
+      "step": 17181
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5943165611117244e-06,
+      "loss": 0.5679,
+      "step": 17182
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5932136430954965e-06,
+      "loss": 0.5412,
+      "step": 17183
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.592110771959641e-06,
+      "loss": 0.6339,
+      "step": 17184
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5910079477193345e-06,
+      "loss": 0.5629,
+      "step": 17185
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.589905170389757e-06,
+      "loss": 0.6248,
+      "step": 17186
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.58880243998608e-06,
+      "loss": 0.5501,
+      "step": 17187
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.587699756523479e-06,
+      "loss": 0.6229,
+      "step": 17188
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5865971200171265e-06,
+      "loss": 0.6264,
+      "step": 17189
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.585494530482199e-06,
+      "loss": 0.5602,
+      "step": 17190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.584391987933865e-06,
+      "loss": 0.6131,
+      "step": 17191
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5832894923873e-06,
+      "loss": 0.5154,
+      "step": 17192
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.582187043857671e-06,
+      "loss": 0.5542,
+      "step": 17193
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.581084642360153e-06,
+      "loss": 0.5538,
+      "step": 17194
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.579982287909916e-06,
+      "loss": 0.5795,
+      "step": 17195
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.578879980522125e-06,
+      "loss": 0.5316,
+      "step": 17196
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.577777720211954e-06,
+      "loss": 0.6307,
+      "step": 17197
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.576675506994567e-06,
+      "loss": 0.5731,
+      "step": 17198
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5755733408851355e-06,
+      "loss": 0.5864,
+      "step": 17199
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.57447122189882e-06,
+      "loss": 0.6606,
+      "step": 17200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.573369150050794e-06,
+      "loss": 0.5997,
+      "step": 17201
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.572267125356216e-06,
+      "loss": 0.6093,
+      "step": 17202
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.571165147830257e-06,
+      "loss": 0.5185,
+      "step": 17203
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.570063217488076e-06,
+      "loss": 0.606,
+      "step": 17204
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.568961334344841e-06,
+      "loss": 0.5622,
+      "step": 17205
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.567859498415713e-06,
+      "loss": 0.5646,
+      "step": 17206
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.566757709715852e-06,
+      "loss": 0.5542,
+      "step": 17207
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.565655968260424e-06,
+      "loss": 0.5815,
+      "step": 17208
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.564554274064585e-06,
+      "loss": 0.5448,
+      "step": 17209
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5634526271435e-06,
+      "loss": 0.6622,
+      "step": 17210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.562351027512326e-06,
+      "loss": 0.6022,
+      "step": 17211
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.561249475186222e-06,
+      "loss": 0.616,
+      "step": 17212
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.560147970180346e-06,
+      "loss": 0.6007,
+      "step": 17213
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.559046512509859e-06,
+      "loss": 0.5647,
+      "step": 17214
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.557945102189913e-06,
+      "loss": 0.5467,
+      "step": 17215
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.556843739235667e-06,
+      "loss": 0.5779,
+      "step": 17216
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.555742423662278e-06,
+      "loss": 0.5624,
+      "step": 17217
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.554641155484898e-06,
+      "loss": 0.6032,
+      "step": 17218
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.553539934718685e-06,
+      "loss": 0.6113,
+      "step": 17219
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5524387613787885e-06,
+      "loss": 0.572,
+      "step": 17220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.551337635480366e-06,
+      "loss": 0.5552,
+      "step": 17221
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.550236557038566e-06,
+      "loss": 0.5759,
+      "step": 17222
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5491355260685445e-06,
+      "loss": 0.5847,
+      "step": 17223
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.548034542585449e-06,
+      "loss": 0.5713,
+      "step": 17224
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.546933606604431e-06,
+      "loss": 0.5862,
+      "step": 17225
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.545832718140641e-06,
+      "loss": 0.6113,
+      "step": 17226
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.544731877209227e-06,
+      "loss": 0.6114,
+      "step": 17227
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.543631083825342e-06,
+      "loss": 0.5939,
+      "step": 17228
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.542530338004126e-06,
+      "loss": 0.6309,
+      "step": 17229
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.541429639760733e-06,
+      "loss": 0.5685,
+      "step": 17230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5403289891103065e-06,
+      "loss": 0.6267,
+      "step": 17231
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.539228386067994e-06,
+      "loss": 0.6348,
+      "step": 17232
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.538127830648937e-06,
+      "loss": 0.638,
+      "step": 17233
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.537027322868287e-06,
+      "loss": 0.5749,
+      "step": 17234
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.53592686274118e-06,
+      "loss": 0.6288,
+      "step": 17235
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.534826450282764e-06,
+      "loss": 0.6524,
+      "step": 17236
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.533726085508183e-06,
+      "loss": 0.5809,
+      "step": 17237
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.532625768432575e-06,
+      "loss": 0.5978,
+      "step": 17238
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.531525499071084e-06,
+      "loss": 0.6244,
+      "step": 17239
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.530425277438849e-06,
+      "loss": 0.6156,
+      "step": 17240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.529325103551012e-06,
+      "loss": 0.5699,
+      "step": 17241
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.528224977422709e-06,
+      "loss": 0.5821,
+      "step": 17242
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5271248990690835e-06,
+      "loss": 0.5925,
+      "step": 17243
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.526024868505269e-06,
+      "loss": 0.5599,
+      "step": 17244
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.524924885746407e-06,
+      "loss": 0.639,
+      "step": 17245
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.523824950807629e-06,
+      "loss": 0.5821,
+      "step": 17246
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.522725063704074e-06,
+      "loss": 0.5894,
+      "step": 17247
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.52162522445088e-06,
+      "loss": 0.6786,
+      "step": 17248
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.520525433063178e-06,
+      "loss": 0.6367,
+      "step": 17249
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.519425689556104e-06,
+      "loss": 0.5789,
+      "step": 17250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.518325993944792e-06,
+      "loss": 0.5918,
+      "step": 17251
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.517226346244372e-06,
+      "loss": 0.5981,
+      "step": 17252
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5161267464699775e-06,
+      "loss": 0.5829,
+      "step": 17253
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.515027194636741e-06,
+      "loss": 0.5975,
+      "step": 17254
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.513927690759792e-06,
+      "loss": 0.5508,
+      "step": 17255
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.51282823485426e-06,
+      "loss": 0.5887,
+      "step": 17256
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.511728826935278e-06,
+      "loss": 0.6197,
+      "step": 17257
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.51062946701797e-06,
+      "loss": 0.5662,
+      "step": 17258
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.509530155117469e-06,
+      "loss": 0.6097,
+      "step": 17259
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.508430891248898e-06,
+      "loss": 0.5622,
+      "step": 17260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.507331675427388e-06,
+      "loss": 0.5948,
+      "step": 17261
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.50623250766806e-06,
+      "loss": 0.5477,
+      "step": 17262
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.505133387986045e-06,
+      "loss": 0.6176,
+      "step": 17263
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.504034316396465e-06,
+      "loss": 0.6599,
+      "step": 17264
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5029352929144455e-06,
+      "loss": 0.5561,
+      "step": 17265
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.501836317555106e-06,
+      "loss": 0.5542,
+      "step": 17266
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.500737390333573e-06,
+      "loss": 0.6517,
+      "step": 17267
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4996385112649714e-06,
+      "loss": 0.6775,
+      "step": 17268
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.498539680364418e-06,
+      "loss": 0.6324,
+      "step": 17269
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.497440897647036e-06,
+      "loss": 0.6418,
+      "step": 17270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.496342163127943e-06,
+      "loss": 0.6463,
+      "step": 17271
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.495243476822264e-06,
+      "loss": 0.6226,
+      "step": 17272
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.49414483874511e-06,
+      "loss": 0.5682,
+      "step": 17273
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.493046248911606e-06,
+      "loss": 0.6502,
+      "step": 17274
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.491947707336865e-06,
+      "loss": 0.585,
+      "step": 17275
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4908492140360105e-06,
+      "loss": 0.5397,
+      "step": 17276
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.489750769024149e-06,
+      "loss": 0.5809,
+      "step": 17277
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.488652372316403e-06,
+      "loss": 0.5851,
+      "step": 17278
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.487554023927887e-06,
+      "loss": 0.5799,
+      "step": 17279
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.486455723873713e-06,
+      "loss": 0.588,
+      "step": 17280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.485357472168996e-06,
+      "loss": 0.5434,
+      "step": 17281
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.484259268828847e-06,
+      "loss": 0.6396,
+      "step": 17282
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4831611138683815e-06,
+      "loss": 0.586,
+      "step": 17283
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.482063007302708e-06,
+      "loss": 0.5887,
+      "step": 17284
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.480964949146939e-06,
+      "loss": 0.5925,
+      "step": 17285
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.479866939416184e-06,
+      "loss": 0.6502,
+      "step": 17286
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.478768978125554e-06,
+      "loss": 0.5845,
+      "step": 17287
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4776710652901564e-06,
+      "loss": 0.6262,
+      "step": 17288
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.476573200925099e-06,
+      "loss": 0.5901,
+      "step": 17289
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4754753850454925e-06,
+      "loss": 0.5815,
+      "step": 17290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4743776176664386e-06,
+      "loss": 0.6243,
+      "step": 17291
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.47327989880305e-06,
+      "loss": 0.5737,
+      "step": 17292
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.472182228470426e-06,
+      "loss": 0.6126,
+      "step": 17293
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.471084606683678e-06,
+      "loss": 0.5785,
+      "step": 17294
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4699870334579015e-06,
+      "loss": 0.6292,
+      "step": 17295
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.46888950880821e-06,
+      "loss": 0.6165,
+      "step": 17296
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.467792032749698e-06,
+      "loss": 0.6382,
+      "step": 17297
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.466694605297471e-06,
+      "loss": 0.6085,
+      "step": 17298
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.465597226466633e-06,
+      "loss": 0.5365,
+      "step": 17299
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4644998962722814e-06,
+      "loss": 0.5926,
+      "step": 17300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.46340261472952e-06,
+      "loss": 0.5888,
+      "step": 17301
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.462305381853444e-06,
+      "loss": 0.5788,
+      "step": 17302
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.461208197659157e-06,
+      "loss": 0.6018,
+      "step": 17303
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.460111062161751e-06,
+      "loss": 0.6707,
+      "step": 17304
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.459013975376331e-06,
+      "loss": 0.5433,
+      "step": 17305
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.457916937317986e-06,
+      "loss": 0.6305,
+      "step": 17306
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.45681994800182e-06,
+      "loss": 0.5826,
+      "step": 17307
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.455723007442922e-06,
+      "loss": 0.6407,
+      "step": 17308
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.45462611565639e-06,
+      "loss": 0.6228,
+      "step": 17309
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.453529272657319e-06,
+      "loss": 0.5765,
+      "step": 17310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.452432478460802e-06,
+      "loss": 0.567,
+      "step": 17311
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.451335733081931e-06,
+      "loss": 0.6302,
+      "step": 17312
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.450239036535797e-06,
+      "loss": 0.5903,
+      "step": 17313
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.449142388837495e-06,
+      "loss": 0.5853,
+      "step": 17314
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.448045790002111e-06,
+      "loss": 0.5859,
+      "step": 17315
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.446949240044742e-06,
+      "loss": 0.5708,
+      "step": 17316
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.445852738980469e-06,
+      "loss": 0.6216,
+      "step": 17317
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.444756286824387e-06,
+      "loss": 0.5707,
+      "step": 17318
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.443659883591584e-06,
+      "loss": 0.5974,
+      "step": 17319
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4425635292971435e-06,
+      "loss": 0.5622,
+      "step": 17320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.441467223956157e-06,
+      "loss": 0.5633,
+      "step": 17321
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.440370967583705e-06,
+      "loss": 0.6473,
+      "step": 17322
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4392747601948805e-06,
+      "loss": 0.5153,
+      "step": 17323
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.438178601804761e-06,
+      "loss": 0.6898,
+      "step": 17324
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4370824924284345e-06,
+      "loss": 0.6324,
+      "step": 17325
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.435986432080983e-06,
+      "loss": 0.6673,
+      "step": 17326
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.434890420777491e-06,
+      "loss": 0.5621,
+      "step": 17327
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.433794458533038e-06,
+      "loss": 0.5719,
+      "step": 17328
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.432698545362707e-06,
+      "loss": 0.584,
+      "step": 17329
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.43160268128158e-06,
+      "loss": 0.5426,
+      "step": 17330
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4305068663047335e-06,
+      "loss": 0.5885,
+      "step": 17331
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.429411100447252e-06,
+      "loss": 0.5878,
+      "step": 17332
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.428315383724208e-06,
+      "loss": 0.5927,
+      "step": 17333
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.427219716150685e-06,
+      "loss": 0.6609,
+      "step": 17334
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.426124097741755e-06,
+      "loss": 0.5642,
+      "step": 17335
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.425028528512501e-06,
+      "loss": 0.634,
+      "step": 17336
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.423933008477994e-06,
+      "loss": 0.6262,
+      "step": 17337
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.42283753765331e-06,
+      "loss": 0.5834,
+      "step": 17338
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.421742116053526e-06,
+      "loss": 0.6143,
+      "step": 17339
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.420646743693715e-06,
+      "loss": 0.5363,
+      "step": 17340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4195514205889495e-06,
+      "loss": 0.655,
+      "step": 17341
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.418456146754302e-06,
+      "loss": 0.5325,
+      "step": 17342
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.417360922204847e-06,
+      "loss": 0.5423,
+      "step": 17343
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.416265746955651e-06,
+      "loss": 0.575,
+      "step": 17344
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.415170621021789e-06,
+      "loss": 0.6213,
+      "step": 17345
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.414075544418328e-06,
+      "loss": 0.5627,
+      "step": 17346
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.412980517160341e-06,
+      "loss": 0.5901,
+      "step": 17347
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.41188553926289e-06,
+      "loss": 0.6027,
+      "step": 17348
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.410790610741048e-06,
+      "loss": 0.5359,
+      "step": 17349
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.409695731609882e-06,
+      "loss": 0.5581,
+      "step": 17350
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.408600901884456e-06,
+      "loss": 0.6456,
+      "step": 17351
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.40750612157984e-06,
+      "loss": 0.6179,
+      "step": 17352
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.406411390711094e-06,
+      "loss": 0.6058,
+      "step": 17353
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.405316709293287e-06,
+      "loss": 0.6339,
+      "step": 17354
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.404222077341478e-06,
+      "loss": 0.6062,
+      "step": 17355
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.403127494870735e-06,
+      "loss": 0.6603,
+      "step": 17356
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.402032961896116e-06,
+      "loss": 0.6384,
+      "step": 17357
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.400938478432688e-06,
+      "loss": 0.6185,
+      "step": 17358
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.399844044495504e-06,
+      "loss": 0.5733,
+      "step": 17359
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.398749660099633e-06,
+      "loss": 0.6353,
+      "step": 17360
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.39765532526013e-06,
+      "loss": 0.6129,
+      "step": 17361
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.396561039992055e-06,
+      "loss": 0.5864,
+      "step": 17362
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.395466804310468e-06,
+      "loss": 0.5963,
+      "step": 17363
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.394372618230422e-06,
+      "loss": 0.61,
+      "step": 17364
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.39327848176698e-06,
+      "loss": 0.6904,
+      "step": 17365
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.392184394935193e-06,
+      "loss": 0.5992,
+      "step": 17366
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3910903577501215e-06,
+      "loss": 0.5957,
+      "step": 17367
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.389996370226814e-06,
+      "loss": 0.5563,
+      "step": 17368
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.388902432380331e-06,
+      "loss": 0.5651,
+      "step": 17369
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.387808544225723e-06,
+      "loss": 0.5387,
+      "step": 17370
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3867147057780435e-06,
+      "loss": 0.5785,
+      "step": 17371
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.385620917052344e-06,
+      "loss": 0.5958,
+      "step": 17372
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.384527178063676e-06,
+      "loss": 0.6801,
+      "step": 17373
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.383433488827093e-06,
+      "loss": 0.5129,
+      "step": 17374
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.382339849357642e-06,
+      "loss": 0.5903,
+      "step": 17375
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3812462596703726e-06,
+      "loss": 0.6015,
+      "step": 17376
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.380152719780335e-06,
+      "loss": 0.5456,
+      "step": 17377
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.379059229702578e-06,
+      "loss": 0.5175,
+      "step": 17378
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.377965789452145e-06,
+      "loss": 0.5406,
+      "step": 17379
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.376872399044085e-06,
+      "loss": 0.6416,
+      "step": 17380
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.375779058493448e-06,
+      "loss": 0.6148,
+      "step": 17381
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.374685767815272e-06,
+      "loss": 0.603,
+      "step": 17382
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3735925270246105e-06,
+      "loss": 0.6475,
+      "step": 17383
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3724993361365e-06,
+      "loss": 0.5794,
+      "step": 17384
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.371406195165983e-06,
+      "loss": 0.6517,
+      "step": 17385
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3703131041281086e-06,
+      "loss": 0.5415,
+      "step": 17386
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.369220063037913e-06,
+      "loss": 0.6259,
+      "step": 17387
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.368127071910441e-06,
+      "loss": 0.5972,
+      "step": 17388
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.36703413076073e-06,
+      "loss": 0.6015,
+      "step": 17389
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.365941239603823e-06,
+      "loss": 0.6423,
+      "step": 17390
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.364848398454756e-06,
+      "loss": 0.628,
+      "step": 17391
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.36375560732857e-06,
+      "loss": 0.565,
+      "step": 17392
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.362662866240303e-06,
+      "loss": 0.611,
+      "step": 17393
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.36157017520499e-06,
+      "loss": 0.631,
+      "step": 17394
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.360477534237668e-06,
+      "loss": 0.5644,
+      "step": 17395
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3593849433533726e-06,
+      "loss": 0.6084,
+      "step": 17396
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.358292402567142e-06,
+      "loss": 0.6616,
+      "step": 17397
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.357199911894005e-06,
+      "loss": 0.5937,
+      "step": 17398
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.356107471349e-06,
+      "loss": 0.5864,
+      "step": 17399
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.355015080947158e-06,
+      "loss": 0.6692,
+      "step": 17400
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.353922740703511e-06,
+      "loss": 0.6468,
+      "step": 17401
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.35283045063309e-06,
+      "loss": 0.5635,
+      "step": 17402
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.351738210750925e-06,
+      "loss": 0.6266,
+      "step": 17403
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.350646021072051e-06,
+      "loss": 0.5962,
+      "step": 17404
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.349553881611492e-06,
+      "loss": 0.6397,
+      "step": 17405
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.348461792384282e-06,
+      "loss": 0.6036,
+      "step": 17406
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.347369753405444e-06,
+      "loss": 0.5585,
+      "step": 17407
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.346277764690009e-06,
+      "loss": 0.5614,
+      "step": 17408
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.345185826253001e-06,
+      "loss": 0.6269,
+      "step": 17409
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.344093938109449e-06,
+      "loss": 0.6504,
+      "step": 17410
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.343002100274375e-06,
+      "loss": 0.6343,
+      "step": 17411
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.341910312762804e-06,
+      "loss": 0.6003,
+      "step": 17412
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.340818575589767e-06,
+      "loss": 0.6183,
+      "step": 17413
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.339726888770277e-06,
+      "loss": 0.5726,
+      "step": 17414
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.338635252319363e-06,
+      "loss": 0.5997,
+      "step": 17415
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.337543666252044e-06,
+      "loss": 0.5966,
+      "step": 17416
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.336452130583344e-06,
+      "loss": 0.6048,
+      "step": 17417
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.335360645328279e-06,
+      "loss": 0.5861,
+      "step": 17418
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.334269210501876e-06,
+      "loss": 0.5921,
+      "step": 17419
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.333177826119145e-06,
+      "loss": 0.665,
+      "step": 17420
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.332086492195112e-06,
+      "loss": 0.5876,
+      "step": 17421
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.33099520874479e-06,
+      "loss": 0.5952,
+      "step": 17422
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.329903975783197e-06,
+      "loss": 0.5989,
+      "step": 17423
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3288127933253526e-06,
+      "loss": 0.5956,
+      "step": 17424
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.327721661386268e-06,
+      "loss": 0.6182,
+      "step": 17425
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.326630579980962e-06,
+      "loss": 0.5549,
+      "step": 17426
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.325539549124445e-06,
+      "loss": 0.591,
+      "step": 17427
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3244485688317355e-06,
+      "loss": 0.6088,
+      "step": 17428
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.32335763911784e-06,
+      "loss": 0.6183,
+      "step": 17429
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.322266759997776e-06,
+      "loss": 0.6071,
+      "step": 17430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.321175931486552e-06,
+      "loss": 0.6018,
+      "step": 17431
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.320085153599181e-06,
+      "loss": 0.5832,
+      "step": 17432
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.318994426350669e-06,
+      "loss": 0.5515,
+      "step": 17433
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3179037497560295e-06,
+      "loss": 0.6321,
+      "step": 17434
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.31681312383027e-06,
+      "loss": 0.6136,
+      "step": 17435
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.315722548588398e-06,
+      "loss": 0.6601,
+      "step": 17436
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.314632024045422e-06,
+      "loss": 0.5784,
+      "step": 17437
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3135415502163466e-06,
+      "loss": 0.6166,
+      "step": 17438
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3124511271161794e-06,
+      "loss": 0.6033,
+      "step": 17439
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.311360754759923e-06,
+      "loss": 0.6624,
+      "step": 17440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3102704331625865e-06,
+      "loss": 0.6035,
+      "step": 17441
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3091801623391676e-06,
+      "loss": 0.5554,
+      "step": 17442
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.308089942304675e-06,
+      "loss": 0.5645,
+      "step": 17443
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.306999773074109e-06,
+      "loss": 0.6737,
+      "step": 17444
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.30590965466247e-06,
+      "loss": 0.6184,
+      "step": 17445
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.30481958708476e-06,
+      "loss": 0.5824,
+      "step": 17446
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.303729570355981e-06,
+      "loss": 0.6073,
+      "step": 17447
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.30263960449113e-06,
+      "loss": 0.5913,
+      "step": 17448
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.301549689505207e-06,
+      "loss": 0.5246,
+      "step": 17449
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.30045982541321e-06,
+      "loss": 0.5515,
+      "step": 17450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.299370012230135e-06,
+      "loss": 0.5719,
+      "step": 17451
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.298280249970983e-06,
+      "loss": 0.592,
+      "step": 17452
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.297190538650745e-06,
+      "loss": 0.5927,
+      "step": 17453
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.296100878284419e-06,
+      "loss": 0.5792,
+      "step": 17454
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.295011268887e-06,
+      "loss": 0.5819,
+      "step": 17455
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.293921710473481e-06,
+      "loss": 0.5678,
+      "step": 17456
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.292832203058857e-06,
+      "loss": 0.6218,
+      "step": 17457
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2917427466581184e-06,
+      "loss": 0.6217,
+      "step": 17458
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.290653341286259e-06,
+      "loss": 0.5622,
+      "step": 17459
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.289563986958267e-06,
+      "loss": 0.5419,
+      "step": 17460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2884746836891365e-06,
+      "loss": 0.6141,
+      "step": 17461
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2873854314938544e-06,
+      "loss": 0.6383,
+      "step": 17462
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.286296230387412e-06,
+      "loss": 0.6506,
+      "step": 17463
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.285207080384795e-06,
+      "loss": 0.6614,
+      "step": 17464
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.284117981500993e-06,
+      "loss": 0.5706,
+      "step": 17465
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.283028933750994e-06,
+      "loss": 0.6435,
+      "step": 17466
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.281939937149781e-06,
+      "loss": 0.6258,
+      "step": 17467
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.280850991712345e-06,
+      "loss": 0.5506,
+      "step": 17468
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.279762097453664e-06,
+      "loss": 0.5647,
+      "step": 17469
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2786732543887274e-06,
+      "loss": 0.5831,
+      "step": 17470
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.277584462532516e-06,
+      "loss": 0.6034,
+      "step": 17471
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.276495721900013e-06,
+      "loss": 0.6548,
+      "step": 17472
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2754070325062e-06,
+      "loss": 0.621,
+      "step": 17473
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.274318394366058e-06,
+      "loss": 0.5694,
+      "step": 17474
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.27322980749457e-06,
+      "loss": 0.6261,
+      "step": 17475
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.272141271906714e-06,
+      "loss": 0.5881,
+      "step": 17476
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.271052787617471e-06,
+      "loss": 0.6075,
+      "step": 17477
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.269964354641817e-06,
+      "loss": 0.5758,
+      "step": 17478
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2688759729947325e-06,
+      "loss": 0.5923,
+      "step": 17479
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.267787642691191e-06,
+      "loss": 0.5586,
+      "step": 17480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.266699363746173e-06,
+      "loss": 0.5598,
+      "step": 17481
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.26561113617465e-06,
+      "loss": 0.6226,
+      "step": 17482
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.264522959991601e-06,
+      "loss": 0.5751,
+      "step": 17483
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2634348352119966e-06,
+      "loss": 0.5929,
+      "step": 17484
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.262346761850813e-06,
+      "loss": 0.6517,
+      "step": 17485
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.261258739923022e-06,
+      "loss": 0.5859,
+      "step": 17486
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2601707694435945e-06,
+      "loss": 0.5495,
+      "step": 17487
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.259082850427507e-06,
+      "loss": 0.6311,
+      "step": 17488
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.257994982889723e-06,
+      "loss": 0.6155,
+      "step": 17489
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.256907166845219e-06,
+      "loss": 0.6133,
+      "step": 17490
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.255819402308958e-06,
+      "loss": 0.6411,
+      "step": 17491
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.254731689295914e-06,
+      "loss": 0.5735,
+      "step": 17492
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.253644027821052e-06,
+      "loss": 0.6384,
+      "step": 17493
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.252556417899342e-06,
+      "loss": 0.5709,
+      "step": 17494
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.251468859545746e-06,
+      "loss": 0.6185,
+      "step": 17495
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.250381352775232e-06,
+      "loss": 0.6035,
+      "step": 17496
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.249293897602769e-06,
+      "loss": 0.606,
+      "step": 17497
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.248206494043313e-06,
+      "loss": 0.7093,
+      "step": 17498
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.247119142111835e-06,
+      "loss": 0.5761,
+      "step": 17499
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.246031841823295e-06,
+      "loss": 0.5897,
+      "step": 17500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.244944593192655e-06,
+      "loss": 0.5806,
+      "step": 17501
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.243857396234876e-06,
+      "loss": 0.6906,
+      "step": 17502
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.242770250964921e-06,
+      "loss": 0.6004,
+      "step": 17503
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.241683157397748e-06,
+      "loss": 0.5769,
+      "step": 17504
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.240596115548316e-06,
+      "loss": 0.5852,
+      "step": 17505
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.239509125431586e-06,
+      "loss": 0.6234,
+      "step": 17506
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.238422187062515e-06,
+      "loss": 0.5585,
+      "step": 17507
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23733530045606e-06,
+      "loss": 0.6064,
+      "step": 17508
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.236248465627177e-06,
+      "loss": 0.5667,
+      "step": 17509
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.235161682590824e-06,
+      "loss": 0.6107,
+      "step": 17510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23407495136195e-06,
+      "loss": 0.5409,
+      "step": 17511
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.232988271955518e-06,
+      "loss": 0.5749,
+      "step": 17512
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.231901644386476e-06,
+      "loss": 0.5939,
+      "step": 17513
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2308150686697775e-06,
+      "loss": 0.5454,
+      "step": 17514
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.229728544820376e-06,
+      "loss": 0.6147,
+      "step": 17515
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2286420728532215e-06,
+      "loss": 0.6155,
+      "step": 17516
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.227555652783268e-06,
+      "loss": 0.6843,
+      "step": 17517
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.226469284625462e-06,
+      "loss": 0.6098,
+      "step": 17518
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.225382968394755e-06,
+      "loss": 0.5991,
+      "step": 17519
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.224296704106096e-06,
+      "loss": 0.5802,
+      "step": 17520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.223210491774432e-06,
+      "loss": 0.5775,
+      "step": 17521
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.222124331414707e-06,
+      "loss": 0.608,
+      "step": 17522
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2210382230418744e-06,
+      "loss": 0.6246,
+      "step": 17523
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.219952166670873e-06,
+      "loss": 0.5964,
+      "step": 17524
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2188661623166545e-06,
+      "loss": 0.5867,
+      "step": 17525
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.217780209994157e-06,
+      "loss": 0.5593,
+      "step": 17526
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2166943097183285e-06,
+      "loss": 0.6323,
+      "step": 17527
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.215608461504112e-06,
+      "loss": 0.6194,
+      "step": 17528
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.214522665366446e-06,
+      "loss": 0.6147,
+      "step": 17529
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.213436921320277e-06,
+      "loss": 0.605,
+      "step": 17530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.21235122938054e-06,
+      "loss": 0.6155,
+      "step": 17531
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2112655895621806e-06,
+      "loss": 0.6456,
+      "step": 17532
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.210180001880135e-06,
+      "loss": 0.5698,
+      "step": 17533
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.209094466349344e-06,
+      "loss": 0.6265,
+      "step": 17534
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.208008982984741e-06,
+      "loss": 0.5837,
+      "step": 17535
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.206923551801267e-06,
+      "loss": 0.6454,
+      "step": 17536
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.20583817281386e-06,
+      "loss": 0.5702,
+      "step": 17537
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.204752846037452e-06,
+      "loss": 0.6075,
+      "step": 17538
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.20366757148698e-06,
+      "loss": 0.6314,
+      "step": 17539
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.202582349177377e-06,
+      "loss": 0.5601,
+      "step": 17540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.201497179123579e-06,
+      "loss": 0.5846,
+      "step": 17541
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.200412061340517e-06,
+      "loss": 0.642,
+      "step": 17542
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.199326995843125e-06,
+      "loss": 0.6041,
+      "step": 17543
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1982419826463304e-06,
+      "loss": 0.5772,
+      "step": 17544
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.197157021765069e-06,
+      "loss": 0.5687,
+      "step": 17545
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.196072113214266e-06,
+      "loss": 0.7006,
+      "step": 17546
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.194987257008855e-06,
+      "loss": 0.5946,
+      "step": 17547
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.193902453163762e-06,
+      "loss": 0.5845,
+      "step": 17548
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1928177016939165e-06,
+      "loss": 0.5859,
+      "step": 17549
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.191733002614244e-06,
+      "loss": 0.5057,
+      "step": 17550
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.190648355939671e-06,
+      "loss": 0.6105,
+      "step": 17551
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.189563761685125e-06,
+      "loss": 0.6164,
+      "step": 17552
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.188479219865529e-06,
+      "loss": 0.5708,
+      "step": 17553
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.18739473049581e-06,
+      "loss": 0.6749,
+      "step": 17554
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.186310293590887e-06,
+      "loss": 0.5327,
+      "step": 17555
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.185225909165686e-06,
+      "loss": 0.6323,
+      "step": 17556
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.184141577235129e-06,
+      "loss": 0.6198,
+      "step": 17557
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.183057297814135e-06,
+      "loss": 0.6325,
+      "step": 17558
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.18197307091763e-06,
+      "loss": 0.5813,
+      "step": 17559
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1808888965605265e-06,
+      "loss": 0.5958,
+      "step": 17560
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.179804774757751e-06,
+      "loss": 0.5405,
+      "step": 17561
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.178720705524216e-06,
+      "loss": 0.6348,
+      "step": 17562
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.177636688874844e-06,
+      "loss": 0.5435,
+      "step": 17563
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.176552724824547e-06,
+      "loss": 0.6465,
+      "step": 17564
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.175468813388247e-06,
+      "loss": 0.5611,
+      "step": 17565
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1743849545808545e-06,
+      "loss": 0.6281,
+      "step": 17566
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.173301148417286e-06,
+      "loss": 0.614,
+      "step": 17567
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.172217394912459e-06,
+      "loss": 0.5273,
+      "step": 17568
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.17113369408128e-06,
+      "loss": 0.6177,
+      "step": 17569
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.17005004593867e-06,
+      "loss": 0.5581,
+      "step": 17570
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.168966450499534e-06,
+      "loss": 0.6475,
+      "step": 17571
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1678829077787864e-06,
+      "loss": 0.5686,
+      "step": 17572
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.166799417791337e-06,
+      "loss": 0.5645,
+      "step": 17573
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.165715980552097e-06,
+      "loss": 0.6013,
+      "step": 17574
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.164632596075972e-06,
+      "loss": 0.5833,
+      "step": 17575
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.163549264377875e-06,
+      "loss": 0.5668,
+      "step": 17576
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.16246598547271e-06,
+      "loss": 0.6175,
+      "step": 17577
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.161382759375383e-06,
+      "loss": 0.5672,
+      "step": 17578
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.160299586100806e-06,
+      "loss": 0.614,
+      "step": 17579
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1592164656638775e-06,
+      "loss": 0.5338,
+      "step": 17580
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1581333980795065e-06,
+      "loss": 0.6359,
+      "step": 17581
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.157050383362595e-06,
+      "loss": 0.5386,
+      "step": 17582
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.155967421528048e-06,
+      "loss": 0.5951,
+      "step": 17583
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.154884512590766e-06,
+      "loss": 0.5988,
+      "step": 17584
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.153801656565654e-06,
+      "loss": 0.6543,
+      "step": 17585
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.152718853467607e-06,
+      "loss": 0.5967,
+      "step": 17586
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1516361033115335e-06,
+      "loss": 0.6516,
+      "step": 17587
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.150553406112325e-06,
+      "loss": 0.5551,
+      "step": 17588
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.149470761884884e-06,
+      "loss": 0.5771,
+      "step": 17589
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1483881706441105e-06,
+      "loss": 0.6096,
+      "step": 17590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1473056324049e-06,
+      "loss": 0.5471,
+      "step": 17591
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.146223147182148e-06,
+      "loss": 0.5615,
+      "step": 17592
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.145140714990754e-06,
+      "loss": 0.5885,
+      "step": 17593
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.144058335845612e-06,
+      "loss": 0.5534,
+      "step": 17594
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.142976009761614e-06,
+      "loss": 0.6133,
+      "step": 17595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.141893736753652e-06,
+      "loss": 0.5892,
+      "step": 17596
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.140811516836627e-06,
+      "loss": 0.5939,
+      "step": 17597
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.139729350025423e-06,
+      "loss": 0.6029,
+      "step": 17598
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.138647236334936e-06,
+      "loss": 0.5779,
+      "step": 17599
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.137565175780056e-06,
+      "loss": 0.6098,
+      "step": 17600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1364831683756705e-06,
+      "loss": 0.642,
+      "step": 17601
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.135401214136676e-06,
+      "loss": 0.6469,
+      "step": 17602
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.134319313077952e-06,
+      "loss": 0.6321,
+      "step": 17603
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.133237465214394e-06,
+      "loss": 0.6077,
+      "step": 17604
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1321556705608856e-06,
+      "loss": 0.5608,
+      "step": 17605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.131073929132314e-06,
+      "loss": 0.5878,
+      "step": 17606
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.129992240943563e-06,
+      "loss": 0.5705,
+      "step": 17607
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.128910606009522e-06,
+      "loss": 0.572,
+      "step": 17608
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.12782902434507e-06,
+      "loss": 0.5688,
+      "step": 17609
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.126747495965093e-06,
+      "loss": 0.6115,
+      "step": 17610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.125666020884476e-06,
+      "loss": 0.6166,
+      "step": 17611
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.124584599118096e-06,
+      "loss": 0.6074,
+      "step": 17612
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.123503230680841e-06,
+      "loss": 0.5646,
+      "step": 17613
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.122421915587584e-06,
+      "loss": 0.4984,
+      "step": 17614
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1213406538532105e-06,
+      "loss": 0.5056,
+      "step": 17615
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.120259445492596e-06,
+      "loss": 0.5251,
+      "step": 17616
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.119178290520623e-06,
+      "loss": 0.6802,
+      "step": 17617
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1180971889521644e-06,
+      "loss": 0.5994,
+      "step": 17618
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.117016140802101e-06,
+      "loss": 0.5861,
+      "step": 17619
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.115935146085307e-06,
+      "loss": 0.603,
+      "step": 17620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.114854204816657e-06,
+      "loss": 0.5978,
+      "step": 17621
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1137733170110295e-06,
+      "loss": 0.5243,
+      "step": 17622
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.112692482683293e-06,
+      "loss": 0.6179,
+      "step": 17623
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1116117018483256e-06,
+      "loss": 0.509,
+      "step": 17624
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1105309745209965e-06,
+      "loss": 0.697,
+      "step": 17625
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.109450300716181e-06,
+      "loss": 0.5875,
+      "step": 17626
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1083696804487445e-06,
+      "loss": 0.5368,
+      "step": 17627
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.107289113733566e-06,
+      "loss": 0.5713,
+      "step": 17628
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.106208600585504e-06,
+      "loss": 0.634,
+      "step": 17629
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.105128141019434e-06,
+      "loss": 0.6169,
+      "step": 17630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.104047735050227e-06,
+      "loss": 0.7218,
+      "step": 17631
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.102967382692742e-06,
+      "loss": 0.611,
+      "step": 17632
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.101887083961852e-06,
+      "loss": 0.6113,
+      "step": 17633
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.100806838872422e-06,
+      "loss": 0.567,
+      "step": 17634
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0997266474393145e-06,
+      "loss": 0.6262,
+      "step": 17635
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.098646509677395e-06,
+      "loss": 0.6275,
+      "step": 17636
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.097566425601531e-06,
+      "loss": 0.5061,
+      "step": 17637
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.096486395226577e-06,
+      "loss": 0.5205,
+      "step": 17638
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.095406418567405e-06,
+      "loss": 0.6112,
+      "step": 17639
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.094326495638867e-06,
+      "loss": 0.6518,
+      "step": 17640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.09324662645583e-06,
+      "loss": 0.6119,
+      "step": 17641
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.092166811033154e-06,
+      "loss": 0.5829,
+      "step": 17642
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.091087049385693e-06,
+      "loss": 0.6813,
+      "step": 17643
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.090007341528313e-06,
+      "loss": 0.5695,
+      "step": 17644
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.088927687475863e-06,
+      "loss": 0.636,
+      "step": 17645
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.087848087243209e-06,
+      "loss": 0.6721,
+      "step": 17646
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0867685408452005e-06,
+      "loss": 0.6174,
+      "step": 17647
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.085689048296695e-06,
+      "loss": 0.5588,
+      "step": 17648
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0846096096125486e-06,
+      "loss": 0.6068,
+      "step": 17649
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.083530224807614e-06,
+      "loss": 0.5667,
+      "step": 17650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.082450893896746e-06,
+      "loss": 0.5592,
+      "step": 17651
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.081371616894792e-06,
+      "loss": 0.5655,
+      "step": 17652
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.080292393816612e-06,
+      "loss": 0.6182,
+      "step": 17653
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.079213224677049e-06,
+      "loss": 0.6073,
+      "step": 17654
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.078134109490962e-06,
+      "loss": 0.5445,
+      "step": 17655
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.077055048273193e-06,
+      "loss": 0.6282,
+      "step": 17656
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.075976041038594e-06,
+      "loss": 0.6522,
+      "step": 17657
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.074897087802012e-06,
+      "loss": 0.582,
+      "step": 17658
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.073818188578295e-06,
+      "loss": 0.5303,
+      "step": 17659
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.07273934338229e-06,
+      "loss": 0.5861,
+      "step": 17660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.071660552228842e-06,
+      "loss": 0.5834,
+      "step": 17661
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.070581815132797e-06,
+      "loss": 0.5578,
+      "step": 17662
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.069503132108998e-06,
+      "loss": 0.5617,
+      "step": 17663
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.068424503172292e-06,
+      "loss": 0.6385,
+      "step": 17664
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.067345928337516e-06,
+      "loss": 0.6347,
+      "step": 17665
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.06626740761952e-06,
+      "loss": 0.509,
+      "step": 17666
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0651889410331355e-06,
+      "loss": 0.5875,
+      "step": 17667
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.064110528593212e-06,
+      "loss": 0.6452,
+      "step": 17668
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.063032170314585e-06,
+      "loss": 0.5733,
+      "step": 17669
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.061953866212096e-06,
+      "loss": 0.5788,
+      "step": 17670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.060875616300581e-06,
+      "loss": 0.5268,
+      "step": 17671
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0597974205948775e-06,
+      "loss": 0.5939,
+      "step": 17672
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.058719279109826e-06,
+      "loss": 0.5893,
+      "step": 17673
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.057641191860259e-06,
+      "loss": 0.5774,
+      "step": 17674
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.056563158861014e-06,
+      "loss": 0.5639,
+      "step": 17675
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0554851801269236e-06,
+      "loss": 0.5388,
+      "step": 17676
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.054407255672825e-06,
+      "loss": 0.5506,
+      "step": 17677
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.053329385513548e-06,
+      "loss": 0.66,
+      "step": 17678
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.052251569663927e-06,
+      "loss": 0.5635,
+      "step": 17679
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.051173808138793e-06,
+      "loss": 0.6295,
+      "step": 17680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.050096100952978e-06,
+      "loss": 0.6059,
+      "step": 17681
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0490184481213095e-06,
+      "loss": 0.6477,
+      "step": 17682
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.04794084965862e-06,
+      "loss": 0.5843,
+      "step": 17683
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.046863305579738e-06,
+      "loss": 0.5648,
+      "step": 17684
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.045785815899488e-06,
+      "loss": 0.6501,
+      "step": 17685
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0447083806327025e-06,
+      "loss": 0.5634,
+      "step": 17686
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.043630999794202e-06,
+      "loss": 0.6081,
+      "step": 17687
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.042553673398819e-06,
+      "loss": 0.6041,
+      "step": 17688
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.04147640146137e-06,
+      "loss": 0.6023,
+      "step": 17689
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.040399183996688e-06,
+      "loss": 0.612,
+      "step": 17690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.039322021019591e-06,
+      "loss": 0.6158,
+      "step": 17691
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.038244912544902e-06,
+      "loss": 0.6646,
+      "step": 17692
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0371678585874475e-06,
+      "loss": 0.566,
+      "step": 17693
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.036090859162041e-06,
+      "loss": 0.5994,
+      "step": 17694
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0350139142835105e-06,
+      "loss": 0.642,
+      "step": 17695
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.033937023966671e-06,
+      "loss": 0.6261,
+      "step": 17696
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.032860188226344e-06,
+      "loss": 0.6169,
+      "step": 17697
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.031783407077345e-06,
+      "loss": 0.629,
+      "step": 17698
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.030706680534495e-06,
+      "loss": 0.583,
+      "step": 17699
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.029630008612608e-06,
+      "loss": 0.59,
+      "step": 17700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.028553391326502e-06,
+      "loss": 0.6687,
+      "step": 17701
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.027476828690988e-06,
+      "loss": 0.5501,
+      "step": 17702
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.026400320720884e-06,
+      "loss": 0.629,
+      "step": 17703
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.025323867431005e-06,
+      "loss": 0.6046,
+      "step": 17704
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.024247468836161e-06,
+      "loss": 0.5934,
+      "step": 17705
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0231711249511655e-06,
+      "loss": 0.544,
+      "step": 17706
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.022094835790829e-06,
+      "loss": 0.5691,
+      "step": 17707
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.021018601369966e-06,
+      "loss": 0.6015,
+      "step": 17708
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.019942421703379e-06,
+      "loss": 0.5967,
+      "step": 17709
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.018866296805883e-06,
+      "loss": 0.6074,
+      "step": 17710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.017790226692285e-06,
+      "loss": 0.6535,
+      "step": 17711
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.016714211377393e-06,
+      "loss": 0.5529,
+      "step": 17712
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.015638250876013e-06,
+      "loss": 0.5986,
+      "step": 17713
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.01456234520295e-06,
+      "loss": 0.6048,
+      "step": 17714
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.013486494373014e-06,
+      "loss": 0.6196,
+      "step": 17715
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.012410698401006e-06,
+      "loss": 0.6104,
+      "step": 17716
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.011334957301731e-06,
+      "loss": 0.6378,
+      "step": 17717
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.010259271089989e-06,
+      "loss": 0.5163,
+      "step": 17718
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.009183639780588e-06,
+      "loss": 0.57,
+      "step": 17719
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.008108063388325e-06,
+      "loss": 0.5695,
+      "step": 17720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.007032541928004e-06,
+      "loss": 0.61,
+      "step": 17721
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.005957075414423e-06,
+      "loss": 0.6117,
+      "step": 17722
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.004881663862381e-06,
+      "loss": 0.5667,
+      "step": 17723
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.003806307286679e-06,
+      "loss": 0.5269,
+      "step": 17724
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.002731005702114e-06,
+      "loss": 0.5359,
+      "step": 17725
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.001655759123481e-06,
+      "loss": 0.5722,
+      "step": 17726
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.000580567565579e-06,
+      "loss": 0.6026,
+      "step": 17727
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.999505431043204e-06,
+      "loss": 0.543,
+      "step": 17728
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9984303495711474e-06,
+      "loss": 0.6189,
+      "step": 17729
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.997355323164207e-06,
+      "loss": 0.6422,
+      "step": 17730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.996280351837173e-06,
+      "loss": 0.6103,
+      "step": 17731
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.995205435604841e-06,
+      "loss": 0.6883,
+      "step": 17732
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.994130574481997e-06,
+      "loss": 0.6354,
+      "step": 17733
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.99305576848344e-06,
+      "loss": 0.5498,
+      "step": 17734
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9919810176239554e-06,
+      "loss": 0.6092,
+      "step": 17735
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.990906321918333e-06,
+      "loss": 0.5926,
+      "step": 17736
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.989831681381364e-06,
+      "loss": 0.6289,
+      "step": 17737
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.988757096027833e-06,
+      "loss": 0.6047,
+      "step": 17738
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.987682565872532e-06,
+      "loss": 0.5635,
+      "step": 17739
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.986608090930241e-06,
+      "loss": 0.592,
+      "step": 17740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.985533671215751e-06,
+      "loss": 0.5794,
+      "step": 17741
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.984459306743844e-06,
+      "loss": 0.5784,
+      "step": 17742
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9833849975293065e-06,
+      "loss": 0.6381,
+      "step": 17743
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9823107435869184e-06,
+      "loss": 0.5618,
+      "step": 17744
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.981236544931466e-06,
+      "loss": 0.5764,
+      "step": 17745
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.980162401577732e-06,
+      "loss": 0.5532,
+      "step": 17746
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.979088313540492e-06,
+      "loss": 0.6114,
+      "step": 17747
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.978014280834532e-06,
+      "loss": 0.5892,
+      "step": 17748
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9769403034746295e-06,
+      "loss": 0.6006,
+      "step": 17749
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.975866381475564e-06,
+      "loss": 0.6364,
+      "step": 17750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.974792514852111e-06,
+      "loss": 0.5225,
+      "step": 17751
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.973718703619051e-06,
+      "loss": 0.503,
+      "step": 17752
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.97264494779116e-06,
+      "loss": 0.5649,
+      "step": 17753
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.971571247383211e-06,
+      "loss": 0.5792,
+      "step": 17754
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.970497602409984e-06,
+      "loss": 0.5957,
+      "step": 17755
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.969424012886249e-06,
+      "loss": 0.5744,
+      "step": 17756
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.968350478826782e-06,
+      "loss": 0.6194,
+      "step": 17757
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9672770002463545e-06,
+      "loss": 0.4944,
+      "step": 17758
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.96620357715974e-06,
+      "loss": 0.5763,
+      "step": 17759
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.965130209581706e-06,
+      "loss": 0.5558,
+      "step": 17760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.964056897527028e-06,
+      "loss": 0.6296,
+      "step": 17761
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.962983641010473e-06,
+      "loss": 0.5376,
+      "step": 17762
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.961910440046811e-06,
+      "loss": 0.6319,
+      "step": 17763
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.960837294650807e-06,
+      "loss": 0.6475,
+      "step": 17764
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.959764204837233e-06,
+      "loss": 0.6401,
+      "step": 17765
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.958691170620854e-06,
+      "loss": 0.6151,
+      "step": 17766
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9576181920164336e-06,
+      "loss": 0.6341,
+      "step": 17767
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.95654526903874e-06,
+      "loss": 0.5782,
+      "step": 17768
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.955472401702535e-06,
+      "loss": 0.5977,
+      "step": 17769
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.954399590022587e-06,
+      "loss": 0.6194,
+      "step": 17770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.953326834013653e-06,
+      "loss": 0.5052,
+      "step": 17771
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.952254133690498e-06,
+      "loss": 0.6307,
+      "step": 17772
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.951181489067882e-06,
+      "loss": 0.5804,
+      "step": 17773
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.950108900160568e-06,
+      "loss": 0.539,
+      "step": 17774
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.949036366983311e-06,
+      "loss": 0.656,
+      "step": 17775
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.947963889550875e-06,
+      "loss": 0.6667,
+      "step": 17776
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.946891467878018e-06,
+      "loss": 0.5334,
+      "step": 17777
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.945819101979493e-06,
+      "loss": 0.6062,
+      "step": 17778
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.944746791870062e-06,
+      "loss": 0.5767,
+      "step": 17779
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.943674537564475e-06,
+      "loss": 0.5754,
+      "step": 17780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.942602339077495e-06,
+      "loss": 0.5601,
+      "step": 17781
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9415301964238685e-06,
+      "loss": 0.586,
+      "step": 17782
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.940458109618354e-06,
+      "loss": 0.6053,
+      "step": 17783
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.939386078675702e-06,
+      "loss": 0.5427,
+      "step": 17784
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.938314103610665e-06,
+      "loss": 0.5895,
+      "step": 17785
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.937242184437997e-06,
+      "loss": 0.661,
+      "step": 17786
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.936170321172444e-06,
+      "loss": 0.6347,
+      "step": 17787
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.935098513828761e-06,
+      "loss": 0.5556,
+      "step": 17788
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.934026762421692e-06,
+      "loss": 0.5991,
+      "step": 17789
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.932955066965989e-06,
+      "loss": 0.6524,
+      "step": 17790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.931883427476396e-06,
+      "loss": 0.6187,
+      "step": 17791
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.930811843967663e-06,
+      "loss": 0.5427,
+      "step": 17792
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.929740316454533e-06,
+      "loss": 0.5712,
+      "step": 17793
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.928668844951755e-06,
+      "loss": 0.5166,
+      "step": 17794
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9275974294740705e-06,
+      "loss": 0.6603,
+      "step": 17795
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.926526070036221e-06,
+      "loss": 0.6208,
+      "step": 17796
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.925454766652955e-06,
+      "loss": 0.5269,
+      "step": 17797
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.92438351933901e-06,
+      "loss": 0.619,
+      "step": 17798
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.923312328109131e-06,
+      "loss": 0.6707,
+      "step": 17799
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.922241192978053e-06,
+      "loss": 0.6029,
+      "step": 17800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.921170113960524e-06,
+      "loss": 0.6042,
+      "step": 17801
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.920099091071274e-06,
+      "loss": 0.5793,
+      "step": 17802
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.919028124325048e-06,
+      "loss": 0.5997,
+      "step": 17803
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.917957213736578e-06,
+      "loss": 0.5785,
+      "step": 17804
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.91688635932061e-06,
+      "loss": 0.6127,
+      "step": 17805
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9158155610918685e-06,
+      "loss": 0.5636,
+      "step": 17806
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.914744819065093e-06,
+      "loss": 0.5295,
+      "step": 17807
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.913674133255019e-06,
+      "loss": 0.5923,
+      "step": 17808
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.912603503676381e-06,
+      "loss": 0.6452,
+      "step": 17809
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.911532930343909e-06,
+      "loss": 0.5436,
+      "step": 17810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.910462413272337e-06,
+      "loss": 0.6393,
+      "step": 17811
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9093919524763935e-06,
+      "loss": 0.6161,
+      "step": 17812
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9083215479708134e-06,
+      "loss": 0.5204,
+      "step": 17813
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.907251199770323e-06,
+      "loss": 0.5276,
+      "step": 17814
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.906180907889654e-06,
+      "loss": 0.6109,
+      "step": 17815
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.905110672343529e-06,
+      "loss": 0.5784,
+      "step": 17816
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.904040493146683e-06,
+      "loss": 0.5841,
+      "step": 17817
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.902970370313837e-06,
+      "loss": 0.6141,
+      "step": 17818
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.901900303859717e-06,
+      "loss": 0.6067,
+      "step": 17819
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.900830293799054e-06,
+      "loss": 0.6138,
+      "step": 17820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.899760340146564e-06,
+      "loss": 0.534,
+      "step": 17821
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.898690442916978e-06,
+      "loss": 0.5695,
+      "step": 17822
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.897620602125011e-06,
+      "loss": 0.6187,
+      "step": 17823
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.896550817785394e-06,
+      "loss": 0.6261,
+      "step": 17824
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.89548108991284e-06,
+      "loss": 0.6405,
+      "step": 17825
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.894411418522075e-06,
+      "loss": 0.5807,
+      "step": 17826
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.893341803627814e-06,
+      "loss": 0.6013,
+      "step": 17827
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.892272245244779e-06,
+      "loss": 0.6604,
+      "step": 17828
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.891202743387689e-06,
+      "loss": 0.5918,
+      "step": 17829
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.890133298071257e-06,
+      "loss": 0.5944,
+      "step": 17830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.889063909310204e-06,
+      "loss": 0.6577,
+      "step": 17831
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.887994577119242e-06,
+      "loss": 0.626,
+      "step": 17832
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.88692530151309e-06,
+      "loss": 0.5948,
+      "step": 17833
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.885856082506458e-06,
+      "loss": 0.6585,
+      "step": 17834
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.884786920114063e-06,
+      "loss": 0.5762,
+      "step": 17835
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.883717814350614e-06,
+      "loss": 0.5487,
+      "step": 17836
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8826487652308275e-06,
+      "loss": 0.6438,
+      "step": 17837
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.881579772769408e-06,
+      "loss": 0.5444,
+      "step": 17838
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.880510836981071e-06,
+      "loss": 0.5987,
+      "step": 17839
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.879441957880526e-06,
+      "loss": 0.5629,
+      "step": 17840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.878373135482479e-06,
+      "loss": 0.5976,
+      "step": 17841
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.877304369801641e-06,
+      "loss": 0.6471,
+      "step": 17842
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.876235660852715e-06,
+      "loss": 0.5328,
+      "step": 17843
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.875167008650412e-06,
+      "loss": 0.5664,
+      "step": 17844
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8740984132094326e-06,
+      "loss": 0.5926,
+      "step": 17845
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.873029874544487e-06,
+      "loss": 0.5793,
+      "step": 17846
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.871961392670275e-06,
+      "loss": 0.6065,
+      "step": 17847
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.870892967601504e-06,
+      "loss": 0.5829,
+      "step": 17848
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8698245993528715e-06,
+      "loss": 0.6052,
+      "step": 17849
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.86875628793908e-06,
+      "loss": 0.5107,
+      "step": 17850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.867688033374836e-06,
+      "loss": 0.6376,
+      "step": 17851
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.866619835674832e-06,
+      "loss": 0.6376,
+      "step": 17852
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.865551694853774e-06,
+      "loss": 0.6005,
+      "step": 17853
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.864483610926355e-06,
+      "loss": 0.6044,
+      "step": 17854
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.863415583907277e-06,
+      "loss": 0.5755,
+      "step": 17855
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.862347613811234e-06,
+      "loss": 0.5581,
+      "step": 17856
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.861279700652927e-06,
+      "loss": 0.5863,
+      "step": 17857
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.860211844447043e-06,
+      "loss": 0.5434,
+      "step": 17858
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.859144045208283e-06,
+      "loss": 0.6274,
+      "step": 17859
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.858076302951341e-06,
+      "loss": 0.6416,
+      "step": 17860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.857008617690907e-06,
+      "loss": 0.5807,
+      "step": 17861
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8559409894416765e-06,
+      "loss": 0.5768,
+      "step": 17862
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.854873418218337e-06,
+      "loss": 0.6455,
+      "step": 17863
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.853805904035583e-06,
+      "loss": 0.6342,
+      "step": 17864
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.852738446908103e-06,
+      "loss": 0.5316,
+      "step": 17865
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.851671046850587e-06,
+      "loss": 0.5354,
+      "step": 17866
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.85060370387772e-06,
+      "loss": 0.6367,
+      "step": 17867
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.849536418004194e-06,
+      "loss": 0.5864,
+      "step": 17868
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8484691892446925e-06,
+      "loss": 0.6381,
+      "step": 17869
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.847402017613902e-06,
+      "loss": 0.5951,
+      "step": 17870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.846334903126511e-06,
+      "loss": 0.5896,
+      "step": 17871
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.845267845797199e-06,
+      "loss": 0.6293,
+      "step": 17872
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.844200845640654e-06,
+      "loss": 0.6363,
+      "step": 17873
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.843133902671556e-06,
+      "loss": 0.5403,
+      "step": 17874
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.842067016904589e-06,
+      "loss": 0.5845,
+      "step": 17875
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.841000188354432e-06,
+      "loss": 0.5889,
+      "step": 17876
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.839933417035768e-06,
+      "loss": 0.5667,
+      "step": 17877
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.838866702963275e-06,
+      "loss": 0.6331,
+      "step": 17878
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.837800046151631e-06,
+      "loss": 0.5959,
+      "step": 17879
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.836733446615519e-06,
+      "loss": 0.5962,
+      "step": 17880
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.835666904369611e-06,
+      "loss": 0.5733,
+      "step": 17881
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.834600419428583e-06,
+      "loss": 0.5814,
+      "step": 17882
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.833533991807117e-06,
+      "loss": 0.6159,
+      "step": 17883
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.832467621519884e-06,
+      "loss": 0.6653,
+      "step": 17884
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.831401308581559e-06,
+      "loss": 0.5866,
+      "step": 17885
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.83033505300681e-06,
+      "loss": 0.5546,
+      "step": 17886
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.829268854810319e-06,
+      "loss": 0.5533,
+      "step": 17887
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.828202714006753e-06,
+      "loss": 0.613,
+      "step": 17888
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.827136630610783e-06,
+      "loss": 0.617,
+      "step": 17889
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.826070604637078e-06,
+      "loss": 0.5966,
+      "step": 17890
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8250046361003095e-06,
+      "loss": 0.6267,
+      "step": 17891
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.823938725015148e-06,
+      "loss": 0.6327,
+      "step": 17892
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.822872871396255e-06,
+      "loss": 0.6355,
+      "step": 17893
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.821807075258303e-06,
+      "loss": 0.5436,
+      "step": 17894
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.820741336615961e-06,
+      "loss": 0.6082,
+      "step": 17895
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.819675655483888e-06,
+      "loss": 0.5363,
+      "step": 17896
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.818610031876746e-06,
+      "loss": 0.5681,
+      "step": 17897
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.817544465809209e-06,
+      "loss": 0.6827,
+      "step": 17898
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.816478957295937e-06,
+      "loss": 0.5922,
+      "step": 17899
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.815413506351584e-06,
+      "loss": 0.6113,
+      "step": 17900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.814348112990824e-06,
+      "loss": 0.6235,
+      "step": 17901
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8132827772283105e-06,
+      "loss": 0.6155,
+      "step": 17902
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.812217499078705e-06,
+      "loss": 0.6142,
+      "step": 17903
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8111522785566635e-06,
+      "loss": 0.5931,
+      "step": 17904
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.81008711567685e-06,
+      "loss": 0.5639,
+      "step": 17905
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.80902201045392e-06,
+      "loss": 0.5578,
+      "step": 17906
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.80795696290253e-06,
+      "loss": 0.5979,
+      "step": 17907
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.806891973037331e-06,
+      "loss": 0.6875,
+      "step": 17908
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.805827040872986e-06,
+      "loss": 0.5676,
+      "step": 17909
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.804762166424147e-06,
+      "loss": 0.5722,
+      "step": 17910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.803697349705464e-06,
+      "loss": 0.6677,
+      "step": 17911
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.802632590731596e-06,
+      "loss": 0.6136,
+      "step": 17912
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.801567889517191e-06,
+      "loss": 0.6159,
+      "step": 17913
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.800503246076902e-06,
+      "loss": 0.6407,
+      "step": 17914
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.799438660425374e-06,
+      "loss": 0.5799,
+      "step": 17915
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.798374132577264e-06,
+      "loss": 0.5629,
+      "step": 17916
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.797309662547218e-06,
+      "loss": 0.5971,
+      "step": 17917
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.796245250349885e-06,
+      "loss": 0.6555,
+      "step": 17918
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.795180895999905e-06,
+      "loss": 0.5892,
+      "step": 17919
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7941165995119364e-06,
+      "loss": 0.6028,
+      "step": 17920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.793052360900619e-06,
+      "loss": 0.5837,
+      "step": 17921
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.791988180180593e-06,
+      "loss": 0.5196,
+      "step": 17922
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.790924057366511e-06,
+      "loss": 0.5565,
+      "step": 17923
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.789859992473013e-06,
+      "loss": 0.5843,
+      "step": 17924
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.788795985514742e-06,
+      "loss": 0.5782,
+      "step": 17925
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.787732036506332e-06,
+      "loss": 0.5464,
+      "step": 17926
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.786668145462435e-06,
+      "loss": 0.5605,
+      "step": 17927
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.785604312397688e-06,
+      "loss": 0.5629,
+      "step": 17928
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.784540537326727e-06,
+      "loss": 0.569,
+      "step": 17929
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.783476820264189e-06,
+      "loss": 0.6574,
+      "step": 17930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.782413161224718e-06,
+      "loss": 0.6226,
+      "step": 17931
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7813495602229495e-06,
+      "loss": 0.6219,
+      "step": 17932
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.780286017273513e-06,
+      "loss": 0.6076,
+      "step": 17933
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.779222532391054e-06,
+      "loss": 0.6044,
+      "step": 17934
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7781591055902e-06,
+      "loss": 0.6205,
+      "step": 17935
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.777095736885585e-06,
+      "loss": 0.6146,
+      "step": 17936
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.776032426291842e-06,
+      "loss": 0.598,
+      "step": 17937
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.774969173823607e-06,
+      "loss": 0.5384,
+      "step": 17938
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.773905979495509e-06,
+      "loss": 0.5664,
+      "step": 17939
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.772842843322177e-06,
+      "loss": 0.572,
+      "step": 17940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.771779765318239e-06,
+      "loss": 0.5704,
+      "step": 17941
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.77071674549833e-06,
+      "loss": 0.6019,
+      "step": 17942
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.769653783877076e-06,
+      "loss": 0.5446,
+      "step": 17943
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.768590880469098e-06,
+      "loss": 0.5791,
+      "step": 17944
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.767528035289032e-06,
+      "loss": 0.6105,
+      "step": 17945
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.766465248351499e-06,
+      "loss": 0.5983,
+      "step": 17946
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.765402519671124e-06,
+      "loss": 0.5393,
+      "step": 17947
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.764339849262528e-06,
+      "loss": 0.5528,
+      "step": 17948
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.76327723714034e-06,
+      "loss": 0.5715,
+      "step": 17949
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.76221468331918e-06,
+      "loss": 0.5669,
+      "step": 17950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.761152187813671e-06,
+      "loss": 0.6306,
+      "step": 17951
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.760089750638427e-06,
+      "loss": 0.5637,
+      "step": 17952
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.759027371808079e-06,
+      "loss": 0.5634,
+      "step": 17953
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7579650513372395e-06,
+      "loss": 0.5838,
+      "step": 17954
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.756902789240525e-06,
+      "loss": 0.5728,
+      "step": 17955
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7558405855325594e-06,
+      "loss": 0.5653,
+      "step": 17956
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.754778440227959e-06,
+      "loss": 0.547,
+      "step": 17957
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.753716353341336e-06,
+      "loss": 0.5429,
+      "step": 17958
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.752654324887302e-06,
+      "loss": 0.594,
+      "step": 17959
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7515923548804816e-06,
+      "loss": 0.6183,
+      "step": 17960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7505304433354825e-06,
+      "loss": 0.645,
+      "step": 17961
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.749468590266916e-06,
+      "loss": 0.6049,
+      "step": 17962
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.748406795689399e-06,
+      "loss": 0.5719,
+      "step": 17963
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.747345059617541e-06,
+      "loss": 0.6147,
+      "step": 17964
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.746283382065951e-06,
+      "loss": 0.609,
+      "step": 17965
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.745221763049237e-06,
+      "loss": 0.5539,
+      "step": 17966
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7441602025820135e-06,
+      "loss": 0.5424,
+      "step": 17967
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.743098700678884e-06,
+      "loss": 0.5752,
+      "step": 17968
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.742037257354458e-06,
+      "loss": 0.6256,
+      "step": 17969
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.740975872623335e-06,
+      "loss": 0.6352,
+      "step": 17970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.739914546500133e-06,
+      "loss": 0.5804,
+      "step": 17971
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7388532789994476e-06,
+      "loss": 0.5086,
+      "step": 17972
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.737792070135883e-06,
+      "loss": 0.5994,
+      "step": 17973
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.736730919924047e-06,
+      "loss": 0.5939,
+      "step": 17974
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.735669828378541e-06,
+      "loss": 0.6091,
+      "step": 17975
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.734608795513965e-06,
+      "loss": 0.578,
+      "step": 17976
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.733547821344916e-06,
+      "loss": 0.6453,
+      "step": 17977
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.732486905886001e-06,
+      "loss": 0.6148,
+      "step": 17978
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.731426049151816e-06,
+      "loss": 0.6208,
+      "step": 17979
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.73036525115696e-06,
+      "loss": 0.5883,
+      "step": 17980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.729304511916025e-06,
+      "loss": 0.6211,
+      "step": 17981
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7282438314436165e-06,
+      "loss": 0.6104,
+      "step": 17982
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7271832097543256e-06,
+      "loss": 0.5723,
+      "step": 17983
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7261226468627466e-06,
+      "loss": 0.5923,
+      "step": 17984
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.725062142783477e-06,
+      "loss": 0.6015,
+      "step": 17985
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7240016975311094e-06,
+      "loss": 0.5515,
+      "step": 17986
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.722941311120236e-06,
+      "loss": 0.5871,
+      "step": 17987
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.721880983565444e-06,
+      "loss": 0.5908,
+      "step": 17988
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7208207148813334e-06,
+      "loss": 0.5756,
+      "step": 17989
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.719760505082489e-06,
+      "loss": 0.6548,
+      "step": 17990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7187003541835014e-06,
+      "loss": 0.6019,
+      "step": 17991
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7176402621989535e-06,
+      "loss": 0.6432,
+      "step": 17992
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.716580229143445e-06,
+      "loss": 0.5886,
+      "step": 17993
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7155202550315546e-06,
+      "loss": 0.6355,
+      "step": 17994
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.714460339877865e-06,
+      "loss": 0.595,
+      "step": 17995
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.713400483696974e-06,
+      "loss": 0.6705,
+      "step": 17996
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.712340686503456e-06,
+      "loss": 0.5865,
+      "step": 17997
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.711280948311899e-06,
+      "loss": 0.5502,
+      "step": 17998
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.71022126913688e-06,
+      "loss": 0.5914,
+      "step": 17999
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.709161648992989e-06,
+      "loss": 0.5309,
+      "step": 18000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.708102087894805e-06,
+      "loss": 0.5744,
+      "step": 18001
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.707042585856908e-06,
+      "loss": 0.528,
+      "step": 18002
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7059831428938715e-06,
+      "loss": 0.6379,
+      "step": 18003
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.704923759020285e-06,
+      "loss": 0.5835,
+      "step": 18004
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.703864434250721e-06,
+      "loss": 0.6387,
+      "step": 18005
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.702805168599754e-06,
+      "loss": 0.5881,
+      "step": 18006
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.701745962081968e-06,
+      "loss": 0.5207,
+      "step": 18007
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.700686814711933e-06,
+      "loss": 0.5656,
+      "step": 18008
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6996277265042265e-06,
+      "loss": 0.5755,
+      "step": 18009
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.698568697473416e-06,
+      "loss": 0.6408,
+      "step": 18010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.697509727634085e-06,
+      "loss": 0.5994,
+      "step": 18011
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.696450817000801e-06,
+      "loss": 0.5177,
+      "step": 18012
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.695391965588133e-06,
+      "loss": 0.5652,
+      "step": 18013
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.694333173410651e-06,
+      "loss": 0.5527,
+      "step": 18014
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.693274440482932e-06,
+      "loss": 0.6272,
+      "step": 18015
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.692215766819542e-06,
+      "loss": 0.4814,
+      "step": 18016
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.691157152435047e-06,
+      "loss": 0.6379,
+      "step": 18017
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.690098597344016e-06,
+      "loss": 0.5445,
+      "step": 18018
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.689040101561014e-06,
+      "loss": 0.6159,
+      "step": 18019
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.687981665100606e-06,
+      "loss": 0.5448,
+      "step": 18020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.686923287977362e-06,
+      "loss": 0.5348,
+      "step": 18021
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.685864970205843e-06,
+      "loss": 0.6007,
+      "step": 18022
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.684806711800614e-06,
+      "loss": 0.6306,
+      "step": 18023
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.683748512776231e-06,
+      "loss": 0.6405,
+      "step": 18024
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.682690373147266e-06,
+      "loss": 0.5792,
+      "step": 18025
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.681632292928275e-06,
+      "loss": 0.5423,
+      "step": 18026
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.680574272133814e-06,
+      "loss": 0.5777,
+      "step": 18027
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.67951631077845e-06,
+      "loss": 0.6141,
+      "step": 18028
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.678458408876739e-06,
+      "loss": 0.5832,
+      "step": 18029
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.677400566443236e-06,
+      "loss": 0.6281,
+      "step": 18030
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.676342783492496e-06,
+      "loss": 0.6128,
+      "step": 18031
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.675285060039083e-06,
+      "loss": 0.5973,
+      "step": 18032
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6742273960975465e-06,
+      "loss": 0.6344,
+      "step": 18033
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.673169791682442e-06,
+      "loss": 0.6208,
+      "step": 18034
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.672112246808321e-06,
+      "loss": 0.6239,
+      "step": 18035
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.67105476148974e-06,
+      "loss": 0.5862,
+      "step": 18036
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6699973357412505e-06,
+      "loss": 0.601,
+      "step": 18037
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.668939969577398e-06,
+      "loss": 0.6257,
+      "step": 18038
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.66788266301274e-06,
+      "loss": 0.5829,
+      "step": 18039
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.666825416061825e-06,
+      "loss": 0.5711,
+      "step": 18040
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.665768228739199e-06,
+      "loss": 0.6184,
+      "step": 18041
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6647111010594055e-06,
+      "loss": 0.5849,
+      "step": 18042
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.663654033037003e-06,
+      "loss": 0.6011,
+      "step": 18043
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6625970246865295e-06,
+      "loss": 0.6299,
+      "step": 18044
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.661540076022533e-06,
+      "loss": 0.5863,
+      "step": 18045
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.660483187059551e-06,
+      "loss": 0.5524,
+      "step": 18046
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.65942635781214e-06,
+      "loss": 0.5443,
+      "step": 18047
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.658369588294835e-06,
+      "loss": 0.5148,
+      "step": 18048
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.657312878522175e-06,
+      "loss": 0.5227,
+      "step": 18049
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.656256228508711e-06,
+      "loss": 0.629,
+      "step": 18050
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.655199638268978e-06,
+      "loss": 0.5404,
+      "step": 18051
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.654143107817513e-06,
+      "loss": 0.5789,
+      "step": 18052
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.653086637168857e-06,
+      "loss": 0.587,
+      "step": 18053
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.652030226337551e-06,
+      "loss": 0.6065,
+      "step": 18054
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.650973875338128e-06,
+      "loss": 0.609,
+      "step": 18055
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.649917584185128e-06,
+      "loss": 0.5885,
+      "step": 18056
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.648861352893079e-06,
+      "loss": 0.5672,
+      "step": 18057
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.647805181476526e-06,
+      "loss": 0.6409,
+      "step": 18058
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.646749069949997e-06,
+      "loss": 0.5989,
+      "step": 18059
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.645693018328023e-06,
+      "loss": 0.6286,
+      "step": 18060
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.644637026625141e-06,
+      "loss": 0.6354,
+      "step": 18061
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.643581094855883e-06,
+      "loss": 0.6217,
+      "step": 18062
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.642525223034776e-06,
+      "loss": 0.5566,
+      "step": 18063
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.641469411176346e-06,
+      "loss": 0.574,
+      "step": 18064
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.640413659295132e-06,
+      "loss": 0.6533,
+      "step": 18065
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.639357967405655e-06,
+      "loss": 0.5686,
+      "step": 18066
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.638302335522441e-06,
+      "loss": 0.6009,
+      "step": 18067
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.637246763660024e-06,
+      "loss": 0.5313,
+      "step": 18068
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.636191251832923e-06,
+      "loss": 0.5678,
+      "step": 18069
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.635135800055667e-06,
+      "loss": 0.6439,
+      "step": 18070
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.634080408342772e-06,
+      "loss": 0.5979,
+      "step": 18071
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.633025076708773e-06,
+      "loss": 0.5097,
+      "step": 18072
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.631969805168184e-06,
+      "loss": 0.5796,
+      "step": 18073
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.630914593735529e-06,
+      "loss": 0.6357,
+      "step": 18074
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.629859442425325e-06,
+      "loss": 0.5407,
+      "step": 18075
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.628804351252099e-06,
+      "loss": 0.6457,
+      "step": 18076
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.627749320230365e-06,
+      "loss": 0.6211,
+      "step": 18077
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.626694349374638e-06,
+      "loss": 0.5774,
+      "step": 18078
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.625639438699444e-06,
+      "loss": 0.557,
+      "step": 18079
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.624584588219295e-06,
+      "loss": 0.6924,
+      "step": 18080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.623529797948708e-06,
+      "loss": 0.6197,
+      "step": 18081
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.622475067902191e-06,
+      "loss": 0.5728,
+      "step": 18082
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.621420398094267e-06,
+      "loss": 0.6097,
+      "step": 18083
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.620365788539446e-06,
+      "loss": 0.6089,
+      "step": 18084
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.619311239252241e-06,
+      "loss": 0.633,
+      "step": 18085
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6182567502471575e-06,
+      "loss": 0.5867,
+      "step": 18086
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.617202321538716e-06,
+      "loss": 0.5712,
+      "step": 18087
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.616147953141422e-06,
+      "loss": 0.5408,
+      "step": 18088
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.615093645069779e-06,
+      "loss": 0.5207,
+      "step": 18089
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6140393973383046e-06,
+      "loss": 0.617,
+      "step": 18090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.612985209961502e-06,
+      "loss": 0.5121,
+      "step": 18091
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.611931082953879e-06,
+      "loss": 0.6205,
+      "step": 18092
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.610877016329934e-06,
+      "loss": 0.6007,
+      "step": 18093
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.609823010104183e-06,
+      "loss": 0.5686,
+      "step": 18094
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.608769064291124e-06,
+      "loss": 0.6284,
+      "step": 18095
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.607715178905262e-06,
+      "loss": 0.5836,
+      "step": 18096
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.606661353961095e-06,
+      "loss": 0.5862,
+      "step": 18097
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.605607589473131e-06,
+      "loss": 0.6155,
+      "step": 18098
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.604553885455868e-06,
+      "loss": 0.5511,
+      "step": 18099
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.603500241923803e-06,
+      "loss": 0.5911,
+      "step": 18100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.602446658891441e-06,
+      "loss": 0.6078,
+      "step": 18101
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.601393136373276e-06,
+      "loss": 0.5765,
+      "step": 18102
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.600339674383808e-06,
+      "loss": 0.6823,
+      "step": 18103
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.599286272937527e-06,
+      "loss": 0.592,
+      "step": 18104
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.598232932048937e-06,
+      "loss": 0.577,
+      "step": 18105
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5971796517325306e-06,
+      "loss": 0.5948,
+      "step": 18106
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5961264320028e-06,
+      "loss": 0.6324,
+      "step": 18107
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.595073272874236e-06,
+      "loss": 0.615,
+      "step": 18108
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.594020174361339e-06,
+      "loss": 0.6179,
+      "step": 18109
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5929671364785955e-06,
+      "loss": 0.6186,
+      "step": 18110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.591914159240491e-06,
+      "loss": 0.5961,
+      "step": 18111
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.590861242661526e-06,
+      "loss": 0.5808,
+      "step": 18112
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.589808386756184e-06,
+      "loss": 0.6356,
+      "step": 18113
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5887555915389545e-06,
+      "loss": 0.5607,
+      "step": 18114
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.58770285702432e-06,
+      "loss": 0.5781,
+      "step": 18115
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.586650183226776e-06,
+      "loss": 0.5618,
+      "step": 18116
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.585597570160802e-06,
+      "loss": 0.5852,
+      "step": 18117
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.584545017840886e-06,
+      "loss": 0.6202,
+      "step": 18118
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.583492526281504e-06,
+      "loss": 0.6492,
+      "step": 18119
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.58244009549715e-06,
+      "loss": 0.6436,
+      "step": 18120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.581387725502303e-06,
+      "loss": 0.6363,
+      "step": 18121
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.580335416311439e-06,
+      "loss": 0.544,
+      "step": 18122
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.579283167939047e-06,
+      "loss": 0.6696,
+      "step": 18123
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5782309803996026e-06,
+      "loss": 0.6627,
+      "step": 18124
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.577178853707586e-06,
+      "loss": 0.6158,
+      "step": 18125
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.576126787877469e-06,
+      "loss": 0.6418,
+      "step": 18126
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.575074782923741e-06,
+      "loss": 0.5835,
+      "step": 18127
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5740228388608705e-06,
+      "loss": 0.5972,
+      "step": 18128
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.572970955703332e-06,
+      "loss": 0.5768,
+      "step": 18129
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.571919133465605e-06,
+      "loss": 0.6219,
+      "step": 18130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.570867372162163e-06,
+      "loss": 0.6464,
+      "step": 18131
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.569815671807477e-06,
+      "loss": 0.664,
+      "step": 18132
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.568764032416018e-06,
+      "loss": 0.5925,
+      "step": 18133
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.567712454002261e-06,
+      "loss": 0.5064,
+      "step": 18134
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.566660936580677e-06,
+      "loss": 0.5815,
+      "step": 18135
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.565609480165732e-06,
+      "loss": 0.6147,
+      "step": 18136
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.564558084771895e-06,
+      "loss": 0.5567,
+      "step": 18137
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.563506750413639e-06,
+      "loss": 0.5914,
+      "step": 18138
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.562455477105428e-06,
+      "loss": 0.5755,
+      "step": 18139
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5614042648617265e-06,
+      "loss": 0.5079,
+      "step": 18140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5603531136970054e-06,
+      "loss": 0.6021,
+      "step": 18141
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.559302023625726e-06,
+      "loss": 0.566,
+      "step": 18142
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.558250994662353e-06,
+      "loss": 0.5781,
+      "step": 18143
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.557200026821346e-06,
+      "loss": 0.561,
+      "step": 18144
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.556149120117174e-06,
+      "loss": 0.6374,
+      "step": 18145
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.555098274564295e-06,
+      "loss": 0.6202,
+      "step": 18146
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.554047490177168e-06,
+      "loss": 0.6093,
+      "step": 18147
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.552996766970251e-06,
+      "loss": 0.5341,
+      "step": 18148
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.55194610495801e-06,
+      "loss": 0.599,
+      "step": 18149
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.550895504154898e-06,
+      "loss": 0.6538,
+      "step": 18150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.549844964575371e-06,
+      "loss": 0.5723,
+      "step": 18151
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.54879448623389e-06,
+      "loss": 0.6208,
+      "step": 18152
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5477440691449095e-06,
+      "loss": 0.5538,
+      "step": 18153
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5466937133228815e-06,
+      "loss": 0.5932,
+      "step": 18154
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5456434187822574e-06,
+      "loss": 0.6242,
+      "step": 18155
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.544593185537498e-06,
+      "loss": 0.6452,
+      "step": 18156
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.543543013603051e-06,
+      "loss": 0.5561,
+      "step": 18157
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.542492902993369e-06,
+      "loss": 0.5752,
+      "step": 18158
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5414428537228955e-06,
+      "loss": 0.5812,
+      "step": 18159
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.54039286580609e-06,
+      "loss": 0.5518,
+      "step": 18160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.539342939257399e-06,
+      "loss": 0.5698,
+      "step": 18161
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.538293074091264e-06,
+      "loss": 0.5224,
+      "step": 18162
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.537243270322139e-06,
+      "loss": 0.5924,
+      "step": 18163
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.536193527964471e-06,
+      "loss": 0.6045,
+      "step": 18164
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5351438470327e-06,
+      "loss": 0.6132,
+      "step": 18165
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.534094227541269e-06,
+      "loss": 0.6049,
+      "step": 18166
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.53304466950463e-06,
+      "loss": 0.588,
+      "step": 18167
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.531995172937221e-06,
+      "loss": 0.5757,
+      "step": 18168
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.530945737853485e-06,
+      "loss": 0.6122,
+      "step": 18169
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.529896364267857e-06,
+      "loss": 0.6332,
+      "step": 18170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.528847052194789e-06,
+      "loss": 0.5642,
+      "step": 18171
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5277978016487135e-06,
+      "loss": 0.5549,
+      "step": 18172
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.526748612644065e-06,
+      "loss": 0.5813,
+      "step": 18173
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.52569948519529e-06,
+      "loss": 0.5659,
+      "step": 18174
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.524650419316824e-06,
+      "loss": 0.6492,
+      "step": 18175
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.523601415023099e-06,
+      "loss": 0.5975,
+      "step": 18176
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.522552472328546e-06,
+      "loss": 0.5395,
+      "step": 18177
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.521503591247612e-06,
+      "loss": 0.6131,
+      "step": 18178
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.520454771794722e-06,
+      "loss": 0.4904,
+      "step": 18179
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.519406013984312e-06,
+      "loss": 0.526,
+      "step": 18180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.518357317830807e-06,
+      "loss": 0.6001,
+      "step": 18181
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.517308683348648e-06,
+      "loss": 0.6233,
+      "step": 18182
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.516260110552258e-06,
+      "loss": 0.6004,
+      "step": 18183
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5152115994560675e-06,
+      "loss": 0.6554,
+      "step": 18184
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.514163150074509e-06,
+      "loss": 0.5371,
+      "step": 18185
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.513114762422007e-06,
+      "loss": 0.656,
+      "step": 18186
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.512066436512989e-06,
+      "loss": 0.6205,
+      "step": 18187
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.511018172361876e-06,
+      "loss": 0.6151,
+      "step": 18188
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.509969969983102e-06,
+      "loss": 0.5737,
+      "step": 18189
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5089218293910855e-06,
+      "loss": 0.6396,
+      "step": 18190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.507873750600247e-06,
+      "loss": 0.5624,
+      "step": 18191
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.506825733625017e-06,
+      "loss": 0.6042,
+      "step": 18192
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.505777778479812e-06,
+      "loss": 0.6047,
+      "step": 18193
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.504729885179054e-06,
+      "loss": 0.5563,
+      "step": 18194
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5036820537371604e-06,
+      "loss": 0.5948,
+      "step": 18195
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.502634284168555e-06,
+      "loss": 0.5959,
+      "step": 18196
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.501586576487654e-06,
+      "loss": 0.6136,
+      "step": 18197
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.500538930708874e-06,
+      "loss": 0.6251,
+      "step": 18198
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.499491346846628e-06,
+      "loss": 0.5734,
+      "step": 18199
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4984438249153405e-06,
+      "loss": 0.6044,
+      "step": 18200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.497396364929421e-06,
+      "loss": 0.5615,
+      "step": 18201
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4963489669032795e-06,
+      "loss": 0.5507,
+      "step": 18202
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4953016308513375e-06,
+      "loss": 0.5877,
+      "step": 18203
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4942543567880025e-06,
+      "loss": 0.5819,
+      "step": 18204
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.493207144727688e-06,
+      "loss": 0.6464,
+      "step": 18205
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.492159994684798e-06,
+      "loss": 0.5455,
+      "step": 18206
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.491112906673753e-06,
+      "loss": 0.5893,
+      "step": 18207
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.490065880708955e-06,
+      "loss": 0.6318,
+      "step": 18208
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.489018916804813e-06,
+      "loss": 0.5755,
+      "step": 18209
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.487972014975731e-06,
+      "loss": 0.5333,
+      "step": 18210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.486925175236123e-06,
+      "loss": 0.6267,
+      "step": 18211
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.48587839760039e-06,
+      "loss": 0.6504,
+      "step": 18212
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.484831682082932e-06,
+      "loss": 0.6174,
+      "step": 18213
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.483785028698162e-06,
+      "loss": 0.5596,
+      "step": 18214
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.482738437460477e-06,
+      "loss": 0.6254,
+      "step": 18215
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.481691908384282e-06,
+      "loss": 0.6095,
+      "step": 18216
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.480645441483973e-06,
+      "loss": 0.5935,
+      "step": 18217
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.479599036773957e-06,
+      "loss": 0.5885,
+      "step": 18218
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.478552694268628e-06,
+      "loss": 0.5852,
+      "step": 18219
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.47750641398239e-06,
+      "loss": 0.5748,
+      "step": 18220
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.476460195929632e-06,
+      "loss": 0.5666,
+      "step": 18221
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.475414040124761e-06,
+      "loss": 0.5466,
+      "step": 18222
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.474367946582168e-06,
+      "loss": 0.5754,
+      "step": 18223
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.473321915316245e-06,
+      "loss": 0.6623,
+      "step": 18224
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.472275946341394e-06,
+      "loss": 0.5704,
+      "step": 18225
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.471230039672004e-06,
+      "loss": 0.6217,
+      "step": 18226
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.470184195322468e-06,
+      "loss": 0.6044,
+      "step": 18227
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4691384133071775e-06,
+      "loss": 0.6344,
+      "step": 18228
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4680926936405245e-06,
+      "loss": 0.586,
+      "step": 18229
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4670470363368975e-06,
+      "loss": 0.5396,
+      "step": 18230
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4660014414106825e-06,
+      "loss": 0.6036,
+      "step": 18231
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.464955908876275e-06,
+      "loss": 0.5969,
+      "step": 18232
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.463910438748059e-06,
+      "loss": 0.5403,
+      "step": 18233
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4628650310404184e-06,
+      "loss": 0.6322,
+      "step": 18234
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.461819685767744e-06,
+      "loss": 0.5404,
+      "step": 18235
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.46077440294442e-06,
+      "loss": 0.6123,
+      "step": 18236
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.459729182584828e-06,
+      "loss": 0.5582,
+      "step": 18237
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.458684024703348e-06,
+      "loss": 0.6079,
+      "step": 18238
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45763892931437e-06,
+      "loss": 0.5166,
+      "step": 18239
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.456593896432272e-06,
+      "loss": 0.5544,
+      "step": 18240
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.455548926071432e-06,
+      "loss": 0.5751,
+      "step": 18241
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.454504018246231e-06,
+      "loss": 0.5568,
+      "step": 18242
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4534591729710505e-06,
+      "loss": 0.6801,
+      "step": 18243
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.452414390260267e-06,
+      "loss": 0.5953,
+      "step": 18244
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.451369670128255e-06,
+      "loss": 0.5608,
+      "step": 18245
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.450325012589396e-06,
+      "loss": 0.5819,
+      "step": 18246
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4492804176580625e-06,
+      "loss": 0.589,
+      "step": 18247
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.44823588534863e-06,
+      "loss": 0.5713,
+      "step": 18248
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.447191415675466e-06,
+      "loss": 0.6253,
+      "step": 18249
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.446147008652952e-06,
+      "loss": 0.6187,
+      "step": 18250
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.445102664295457e-06,
+      "loss": 0.6441,
+      "step": 18251
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4440583826173525e-06,
+      "loss": 0.6148,
+      "step": 18252
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.443014163633004e-06,
+      "loss": 0.6082,
+      "step": 18253
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.441970007356787e-06,
+      "loss": 0.5578,
+      "step": 18254
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.44092591380307e-06,
+      "loss": 0.6476,
+      "step": 18255
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4398818829862135e-06,
+      "loss": 0.6234,
+      "step": 18256
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4388379149205914e-06,
+      "loss": 0.6304,
+      "step": 18257
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.43779400962057e-06,
+      "loss": 0.6768,
+      "step": 18258
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.436750167100511e-06,
+      "loss": 0.4953,
+      "step": 18259
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.435706387374774e-06,
+      "loss": 0.5749,
+      "step": 18260
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.434662670457733e-06,
+      "loss": 0.5481,
+      "step": 18261
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.433619016363745e-06,
+      "loss": 0.6769,
+      "step": 18262
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.432575425107171e-06,
+      "loss": 0.5812,
+      "step": 18263
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.431531896702369e-06,
+      "loss": 0.556,
+      "step": 18264
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.430488431163705e-06,
+      "loss": 0.6434,
+      "step": 18265
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4294450285055376e-06,
+      "loss": 0.638,
+      "step": 18266
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4284016887422175e-06,
+      "loss": 0.5387,
+      "step": 18267
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.427358411888111e-06,
+      "loss": 0.5134,
+      "step": 18268
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.42631519795757e-06,
+      "loss": 0.5539,
+      "step": 18269
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.425272046964953e-06,
+      "loss": 0.6061,
+      "step": 18270
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.424228958924607e-06,
+      "loss": 0.6062,
+      "step": 18271
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.423185933850894e-06,
+      "loss": 0.5903,
+      "step": 18272
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.422142971758164e-06,
+      "loss": 0.6506,
+      "step": 18273
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.421100072660771e-06,
+      "loss": 0.6183,
+      "step": 18274
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.42005723657306e-06,
+      "loss": 0.537,
+      "step": 18275
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.419014463509389e-06,
+      "loss": 0.544,
+      "step": 18276
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.417971753484104e-06,
+      "loss": 0.5927,
+      "step": 18277
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.41692910651155e-06,
+      "loss": 0.5437,
+      "step": 18278
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.415886522606084e-06,
+      "loss": 0.6107,
+      "step": 18279
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.414844001782046e-06,
+      "loss": 0.6485,
+      "step": 18280
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.413801544053783e-06,
+      "loss": 0.5863,
+      "step": 18281
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.412759149435638e-06,
+      "loss": 0.6187,
+      "step": 18282
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.411716817941961e-06,
+      "loss": 0.6358,
+      "step": 18283
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.410674549587092e-06,
+      "loss": 0.6242,
+      "step": 18284
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.409632344385369e-06,
+      "loss": 0.6092,
+      "step": 18285
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4085902023511434e-06,
+      "loss": 0.6219,
+      "step": 18286
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.407548123498751e-06,
+      "loss": 0.6215,
+      "step": 18287
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.406506107842532e-06,
+      "loss": 0.6057,
+      "step": 18288
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.40546415539682e-06,
+      "loss": 0.5384,
+      "step": 18289
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.404422266175964e-06,
+      "loss": 0.5779,
+      "step": 18290
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.403380440194295e-06,
+      "loss": 0.5825,
+      "step": 18291
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.402338677466149e-06,
+      "loss": 0.5916,
+      "step": 18292
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.401296978005859e-06,
+      "loss": 0.5582,
+      "step": 18293
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.400255341827768e-06,
+      "loss": 0.6062,
+      "step": 18294
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.399213768946205e-06,
+      "loss": 0.6299,
+      "step": 18295
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3981722593754996e-06,
+      "loss": 0.6261,
+      "step": 18296
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.397130813129992e-06,
+      "loss": 0.5587,
+      "step": 18297
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.396089430224006e-06,
+      "loss": 0.5639,
+      "step": 18298
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.395048110671877e-06,
+      "loss": 0.5968,
+      "step": 18299
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3940068544879274e-06,
+      "loss": 0.5675,
+      "step": 18300
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.392965661686495e-06,
+      "loss": 0.6184,
+      "step": 18301
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.391924532281905e-06,
+      "loss": 0.5861,
+      "step": 18302
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.390883466288479e-06,
+      "loss": 0.5576,
+      "step": 18303
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.389842463720545e-06,
+      "loss": 0.6103,
+      "step": 18304
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.388801524592433e-06,
+      "loss": 0.6537,
+      "step": 18305
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.387760648918463e-06,
+      "loss": 0.6835,
+      "step": 18306
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.386719836712956e-06,
+      "loss": 0.6392,
+      "step": 18307
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.385679087990241e-06,
+      "loss": 0.5302,
+      "step": 18308
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.384638402764637e-06,
+      "loss": 0.5587,
+      "step": 18309
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.383597781050463e-06,
+      "loss": 0.5866,
+      "step": 18310
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.382557222862037e-06,
+      "loss": 0.6039,
+      "step": 18311
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.381516728213684e-06,
+      "loss": 0.5985,
+      "step": 18312
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.380476297119719e-06,
+      "loss": 0.5635,
+      "step": 18313
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.379435929594459e-06,
+      "loss": 0.5941,
+      "step": 18314
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.378395625652218e-06,
+      "loss": 0.6084,
+      "step": 18315
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.377355385307317e-06,
+      "loss": 0.609,
+      "step": 18316
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.376315208574068e-06,
+      "loss": 0.6386,
+      "step": 18317
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.375275095466779e-06,
+      "loss": 0.5225,
+      "step": 18318
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.374235045999774e-06,
+      "loss": 0.6548,
+      "step": 18319
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.373195060187359e-06,
+      "loss": 0.6055,
+      "step": 18320
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3721551380438455e-06,
+      "loss": 0.6548,
+      "step": 18321
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.371115279583539e-06,
+      "loss": 0.5727,
+      "step": 18322
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.370075484820758e-06,
+      "loss": 0.6614,
+      "step": 18323
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.369035753769807e-06,
+      "loss": 0.6211,
+      "step": 18324
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3679960864449935e-06,
+      "loss": 0.5588,
+      "step": 18325
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.366956482860619e-06,
+      "loss": 0.5899,
+      "step": 18326
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.365916943030998e-06,
+      "loss": 0.5677,
+      "step": 18327
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3648774669704315e-06,
+      "loss": 0.5903,
+      "step": 18328
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3638380546932215e-06,
+      "loss": 0.5809,
+      "step": 18329
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.362798706213677e-06,
+      "loss": 0.6102,
+      "step": 18330
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.361759421546096e-06,
+      "loss": 0.5633,
+      "step": 18331
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.360720200704781e-06,
+      "loss": 0.6194,
+      "step": 18332
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.359681043704028e-06,
+      "loss": 0.6586,
+      "step": 18333
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.358641950558145e-06,
+      "loss": 0.5743,
+      "step": 18334
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3576029212814285e-06,
+      "loss": 0.6094,
+      "step": 18335
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3565639558881745e-06,
+      "loss": 0.6027,
+      "step": 18336
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.355525054392677e-06,
+      "loss": 0.5752,
+      "step": 18337
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.354486216809239e-06,
+      "loss": 0.6009,
+      "step": 18338
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.353447443152153e-06,
+      "loss": 0.5614,
+      "step": 18339
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.352408733435709e-06,
+      "loss": 0.6575,
+      "step": 18340
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.351370087674208e-06,
+      "loss": 0.6067,
+      "step": 18341
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3503315058819405e-06,
+      "loss": 0.5945,
+      "step": 18342
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.349292988073196e-06,
+      "loss": 0.5367,
+      "step": 18343
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.348254534262262e-06,
+      "loss": 0.6681,
+      "step": 18344
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3472161444634375e-06,
+      "loss": 0.6093,
+      "step": 18345
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.346177818691007e-06,
+      "loss": 0.5627,
+      "step": 18346
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.345139556959255e-06,
+      "loss": 0.6038,
+      "step": 18347
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.344101359282477e-06,
+      "loss": 0.641,
+      "step": 18348
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.343063225674956e-06,
+      "loss": 0.5576,
+      "step": 18349
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.342025156150976e-06,
+      "loss": 0.6288,
+      "step": 18350
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.340987150724819e-06,
+      "loss": 0.5169,
+      "step": 18351
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.339949209410775e-06,
+      "loss": 0.6766,
+      "step": 18352
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.338911332223127e-06,
+      "loss": 0.6519,
+      "step": 18353
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3378735191761536e-06,
+      "loss": 0.5994,
+      "step": 18354
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.336835770284133e-06,
+      "loss": 0.5864,
+      "step": 18355
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3357980855613545e-06,
+      "loss": 0.5452,
+      "step": 18356
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.334760465022092e-06,
+      "loss": 0.6067,
+      "step": 18357
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3337229086806206e-06,
+      "loss": 0.5468,
+      "step": 18358
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.332685416551226e-06,
+      "loss": 0.6326,
+      "step": 18359
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.331647988648181e-06,
+      "loss": 0.6973,
+      "step": 18360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.330610624985762e-06,
+      "loss": 0.6079,
+      "step": 18361
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.32957332557824e-06,
+      "loss": 0.5997,
+      "step": 18362
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3285360904398965e-06,
+      "loss": 0.6225,
+      "step": 18363
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.327498919585001e-06,
+      "loss": 0.6433,
+      "step": 18364
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.326461813027826e-06,
+      "loss": 0.5913,
+      "step": 18365
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.325424770782639e-06,
+      "loss": 0.6229,
+      "step": 18366
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.324387792863719e-06,
+      "loss": 0.6148,
+      "step": 18367
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3233508792853315e-06,
+      "loss": 0.6029,
+      "step": 18368
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.32231403006174e-06,
+      "loss": 0.6408,
+      "step": 18369
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.321277245207223e-06,
+      "loss": 0.6378,
+      "step": 18370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3202405247360415e-06,
+      "loss": 0.5632,
+      "step": 18371
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.319203868662462e-06,
+      "loss": 0.6352,
+      "step": 18372
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.318167277000748e-06,
+      "loss": 0.6045,
+      "step": 18373
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.317130749765167e-06,
+      "loss": 0.5919,
+      "step": 18374
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.316094286969984e-06,
+      "loss": 0.5355,
+      "step": 18375
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.315057888629458e-06,
+      "loss": 0.5759,
+      "step": 18376
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.314021554757849e-06,
+      "loss": 0.5648,
+      "step": 18377
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.312985285369424e-06,
+      "loss": 0.545,
+      "step": 18378
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.31194908047844e-06,
+      "loss": 0.6306,
+      "step": 18379
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.310912940099153e-06,
+      "loss": 0.5747,
+      "step": 18380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3098768642458264e-06,
+      "loss": 0.5201,
+      "step": 18381
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3088408529327165e-06,
+      "loss": 0.6045,
+      "step": 18382
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.307804906174079e-06,
+      "loss": 0.5631,
+      "step": 18383
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.306769023984163e-06,
+      "loss": 0.7002,
+      "step": 18384
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.305733206377235e-06,
+      "loss": 0.6125,
+      "step": 18385
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.304697453367541e-06,
+      "loss": 0.561,
+      "step": 18386
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.303661764969336e-06,
+      "loss": 0.6213,
+      "step": 18387
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.30262614119687e-06,
+      "loss": 0.6046,
+      "step": 18388
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3015905820643975e-06,
+      "loss": 0.5839,
+      "step": 18389
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.300555087586168e-06,
+      "loss": 0.676,
+      "step": 18390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.299519657776424e-06,
+      "loss": 0.5808,
+      "step": 18391
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2984842926494265e-06,
+      "loss": 0.5732,
+      "step": 18392
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.297448992219414e-06,
+      "loss": 0.5871,
+      "step": 18393
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.296413756500636e-06,
+      "loss": 0.5369,
+      "step": 18394
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.295378585507334e-06,
+      "loss": 0.536,
+      "step": 18395
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.294343479253761e-06,
+      "loss": 0.6219,
+      "step": 18396
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.293308437754155e-06,
+      "loss": 0.5676,
+      "step": 18397
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.292273461022761e-06,
+      "loss": 0.6348,
+      "step": 18398
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.291238549073818e-06,
+      "loss": 0.5649,
+      "step": 18399
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.290203701921572e-06,
+      "loss": 0.6104,
+      "step": 18400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.289168919580263e-06,
+      "loss": 0.5772,
+      "step": 18401
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.288134202064126e-06,
+      "loss": 0.5935,
+      "step": 18402
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.287099549387405e-06,
+      "loss": 0.5526,
+      "step": 18403
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.286064961564334e-06,
+      "loss": 0.5887,
+      "step": 18404
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.285030438609153e-06,
+      "loss": 0.6249,
+      "step": 18405
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.283995980536092e-06,
+      "loss": 0.5393,
+      "step": 18406
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.282961587359394e-06,
+      "loss": 0.5994,
+      "step": 18407
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.281927259093289e-06,
+      "loss": 0.5666,
+      "step": 18408
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2808929957520075e-06,
+      "loss": 0.6211,
+      "step": 18409
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.279858797349789e-06,
+      "loss": 0.6908,
+      "step": 18410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.278824663900859e-06,
+      "loss": 0.5268,
+      "step": 18411
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.277790595419451e-06,
+      "loss": 0.5479,
+      "step": 18412
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2767565919197895e-06,
+      "loss": 0.5601,
+      "step": 18413
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.275722653416112e-06,
+      "loss": 0.6012,
+      "step": 18414
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.274688779922642e-06,
+      "loss": 0.6018,
+      "step": 18415
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.273654971453606e-06,
+      "loss": 0.6197,
+      "step": 18416
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.272621228023227e-06,
+      "loss": 0.6324,
+      "step": 18417
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.271587549645738e-06,
+      "loss": 0.6316,
+      "step": 18418
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.270553936335359e-06,
+      "loss": 0.585,
+      "step": 18419
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2695203881063085e-06,
+      "loss": 0.5863,
+      "step": 18420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.268486904972819e-06,
+      "loss": 0.5149,
+      "step": 18421
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2674534869491055e-06,
+      "loss": 0.6073,
+      "step": 18422
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.266420134049392e-06,
+      "loss": 0.6741,
+      "step": 18423
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2653868462878925e-06,
+      "loss": 0.5657,
+      "step": 18424
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2643536236788325e-06,
+      "loss": 0.5726,
+      "step": 18425
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.26332046623643e-06,
+      "loss": 0.5367,
+      "step": 18426
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2622873739749e-06,
+      "loss": 0.5661,
+      "step": 18427
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.261254346908453e-06,
+      "loss": 0.5824,
+      "step": 18428
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.260221385051315e-06,
+      "loss": 0.5714,
+      "step": 18429
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.259188488417695e-06,
+      "loss": 0.6158,
+      "step": 18430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.258155657021804e-06,
+      "loss": 0.6178,
+      "step": 18431
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.257122890877861e-06,
+      "loss": 0.5727,
+      "step": 18432
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2560901900000734e-06,
+      "loss": 0.5761,
+      "step": 18433
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.255057554402655e-06,
+      "loss": 0.604,
+      "step": 18434
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.254024984099809e-06,
+      "loss": 0.641,
+      "step": 18435
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2529924791057545e-06,
+      "loss": 0.5902,
+      "step": 18436
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.251960039434697e-06,
+      "loss": 0.6064,
+      "step": 18437
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.250927665100833e-06,
+      "loss": 0.6022,
+      "step": 18438
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.249895356118384e-06,
+      "loss": 0.6022,
+      "step": 18439
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.248863112501547e-06,
+      "loss": 0.5451,
+      "step": 18440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.247830934264528e-06,
+      "loss": 0.6137,
+      "step": 18441
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.246798821421527e-06,
+      "loss": 0.571,
+      "step": 18442
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.245766773986755e-06,
+      "loss": 0.5653,
+      "step": 18443
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.244734791974409e-06,
+      "loss": 0.61,
+      "step": 18444
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.243702875398687e-06,
+      "loss": 0.5841,
+      "step": 18445
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.242671024273798e-06,
+      "loss": 0.6276,
+      "step": 18446
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.241639238613934e-06,
+      "loss": 0.639,
+      "step": 18447
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.240607518433296e-06,
+      "loss": 0.693,
+      "step": 18448
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.239575863746076e-06,
+      "loss": 0.585,
+      "step": 18449
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.23854427456648e-06,
+      "loss": 0.5821,
+      "step": 18450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.237512750908698e-06,
+      "loss": 0.5695,
+      "step": 18451
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.236481292786922e-06,
+      "loss": 0.5883,
+      "step": 18452
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.235449900215351e-06,
+      "loss": 0.5948,
+      "step": 18453
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2344185732081776e-06,
+      "loss": 0.5674,
+      "step": 18454
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2333873117795905e-06,
+      "loss": 0.5647,
+      "step": 18455
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.23235611594378e-06,
+      "loss": 0.6406,
+      "step": 18456
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.231324985714942e-06,
+      "loss": 0.5888,
+      "step": 18457
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2302939211072615e-06,
+      "loss": 0.5112,
+      "step": 18458
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.229262922134929e-06,
+      "loss": 0.5703,
+      "step": 18459
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.228231988812126e-06,
+      "loss": 0.5323,
+      "step": 18460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.227201121153047e-06,
+      "loss": 0.5782,
+      "step": 18461
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.226170319171875e-06,
+      "loss": 0.6215,
+      "step": 18462
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.22513958288279e-06,
+      "loss": 0.5323,
+      "step": 18463
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.224108912299984e-06,
+      "loss": 0.572,
+      "step": 18464
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.223078307437637e-06,
+      "loss": 0.5801,
+      "step": 18465
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.222047768309928e-06,
+      "loss": 0.6056,
+      "step": 18466
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2210172949310365e-06,
+      "loss": 0.6148,
+      "step": 18467
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.219986887315151e-06,
+      "loss": 0.5509,
+      "step": 18468
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.218956545476445e-06,
+      "loss": 0.541,
+      "step": 18469
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2179262694290975e-06,
+      "loss": 0.6443,
+      "step": 18470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.216896059187284e-06,
+      "loss": 0.5463,
+      "step": 18471
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.215865914765186e-06,
+      "loss": 0.5763,
+      "step": 18472
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.214835836176978e-06,
+      "loss": 0.549,
+      "step": 18473
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.213805823436828e-06,
+      "loss": 0.6247,
+      "step": 18474
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.212775876558921e-06,
+      "loss": 0.5994,
+      "step": 18475
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2117459955574244e-06,
+      "loss": 0.6126,
+      "step": 18476
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.210716180446509e-06,
+      "loss": 0.5923,
+      "step": 18477
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.209686431240344e-06,
+      "loss": 0.5225,
+      "step": 18478
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2086567479531066e-06,
+      "loss": 0.613,
+      "step": 18479
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.207627130598962e-06,
+      "loss": 0.6185,
+      "step": 18480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.206597579192078e-06,
+      "loss": 0.5577,
+      "step": 18481
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.205568093746622e-06,
+      "loss": 0.5815,
+      "step": 18482
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.204538674276765e-06,
+      "loss": 0.633,
+      "step": 18483
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.203509320796668e-06,
+      "loss": 0.4877,
+      "step": 18484
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.202480033320494e-06,
+      "loss": 0.6407,
+      "step": 18485
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.201450811862414e-06,
+      "loss": 0.5912,
+      "step": 18486
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.200421656436589e-06,
+      "loss": 0.5963,
+      "step": 18487
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.199392567057178e-06,
+      "loss": 0.587,
+      "step": 18488
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.198363543738339e-06,
+      "loss": 0.5862,
+      "step": 18489
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.19733458649424e-06,
+      "loss": 0.6108,
+      "step": 18490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.196305695339039e-06,
+      "loss": 0.521,
+      "step": 18491
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.195276870286892e-06,
+      "loss": 0.5352,
+      "step": 18492
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.194248111351951e-06,
+      "loss": 0.6515,
+      "step": 18493
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.193219418548386e-06,
+      "loss": 0.5851,
+      "step": 18494
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.192190791890344e-06,
+      "loss": 0.4963,
+      "step": 18495
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.191162231391977e-06,
+      "loss": 0.6284,
+      "step": 18496
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.190133737067446e-06,
+      "loss": 0.5934,
+      "step": 18497
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.189105308930902e-06,
+      "loss": 0.5239,
+      "step": 18498
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1880769469964965e-06,
+      "loss": 0.6334,
+      "step": 18499
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1870486512783766e-06,
+      "loss": 0.5821,
+      "step": 18500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1860204217907004e-06,
+      "loss": 0.6021,
+      "step": 18501
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.184992258547612e-06,
+      "loss": 0.63,
+      "step": 18502
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.183964161563262e-06,
+      "loss": 0.5792,
+      "step": 18503
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.182936130851792e-06,
+      "loss": 0.5869,
+      "step": 18504
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.181908166427358e-06,
+      "loss": 0.5877,
+      "step": 18505
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.180880268304102e-06,
+      "loss": 0.5426,
+      "step": 18506
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.179852436496162e-06,
+      "loss": 0.6353,
+      "step": 18507
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.178824671017693e-06,
+      "loss": 0.5365,
+      "step": 18508
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1777969718828336e-06,
+      "loss": 0.6052,
+      "step": 18509
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.176769339105724e-06,
+      "loss": 0.571,
+      "step": 18510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.175741772700502e-06,
+      "loss": 0.5895,
+      "step": 18511
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.174714272681317e-06,
+      "loss": 0.5779,
+      "step": 18512
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.173686839062304e-06,
+      "loss": 0.5295,
+      "step": 18513
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.172659471857595e-06,
+      "loss": 0.6522,
+      "step": 18514
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1716321710813385e-06,
+      "loss": 0.6425,
+      "step": 18515
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.170604936747665e-06,
+      "loss": 0.6555,
+      "step": 18516
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.169577768870713e-06,
+      "loss": 0.6452,
+      "step": 18517
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.16855066746461e-06,
+      "loss": 0.6143,
+      "step": 18518
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.167523632543499e-06,
+      "loss": 0.5463,
+      "step": 18519
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.16649666412151e-06,
+      "loss": 0.6067,
+      "step": 18520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.165469762212772e-06,
+      "loss": 0.5646,
+      "step": 18521
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1644429268314155e-06,
+      "loss": 0.6209,
+      "step": 18522
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.163416157991576e-06,
+      "loss": 0.5962,
+      "step": 18523
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.162389455707382e-06,
+      "loss": 0.6105,
+      "step": 18524
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.161362819992954e-06,
+      "loss": 0.6004,
+      "step": 18525
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.16033625086243e-06,
+      "loss": 0.5835,
+      "step": 18526
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.159309748329933e-06,
+      "loss": 0.6222,
+      "step": 18527
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.158283312409584e-06,
+      "loss": 0.5894,
+      "step": 18528
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.15725694311551e-06,
+      "loss": 0.6628,
+      "step": 18529
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.156230640461839e-06,
+      "loss": 0.6114,
+      "step": 18530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.155204404462689e-06,
+      "loss": 0.5907,
+      "step": 18531
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.154178235132184e-06,
+      "loss": 0.5402,
+      "step": 18532
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1531521324844415e-06,
+      "loss": 0.5999,
+      "step": 18533
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1521260965335875e-06,
+      "loss": 0.5899,
+      "step": 18534
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.151100127293738e-06,
+      "loss": 0.5871,
+      "step": 18535
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.150074224779008e-06,
+      "loss": 0.6732,
+      "step": 18536
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.149048389003521e-06,
+      "loss": 0.6504,
+      "step": 18537
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.148022619981392e-06,
+      "loss": 0.6129,
+      "step": 18538
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.146996917726735e-06,
+      "loss": 0.6151,
+      "step": 18539
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.145971282253661e-06,
+      "loss": 0.6121,
+      "step": 18540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1449457135762905e-06,
+      "loss": 0.5574,
+      "step": 18541
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1439202117087325e-06,
+      "loss": 0.6218,
+      "step": 18542
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1428947766651e-06,
+      "loss": 0.6285,
+      "step": 18543
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.141869408459499e-06,
+      "loss": 0.6226,
+      "step": 18544
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.140844107106047e-06,
+      "loss": 0.57,
+      "step": 18545
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.139818872618852e-06,
+      "loss": 0.5661,
+      "step": 18546
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.138793705012013e-06,
+      "loss": 0.5982,
+      "step": 18547
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.137768604299649e-06,
+      "loss": 0.5997,
+      "step": 18548
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.136743570495861e-06,
+      "loss": 0.6115,
+      "step": 18549
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.135718603614757e-06,
+      "loss": 0.5446,
+      "step": 18550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.134693703670432e-06,
+      "loss": 0.6041,
+      "step": 18551
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1336688706770025e-06,
+      "loss": 0.5914,
+      "step": 18552
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.132644104648565e-06,
+      "loss": 0.627,
+      "step": 18553
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.131619405599222e-06,
+      "loss": 0.5567,
+      "step": 18554
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1305947735430684e-06,
+      "loss": 0.597,
+      "step": 18555
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.129570208494214e-06,
+      "loss": 0.5204,
+      "step": 18556
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1285457104667525e-06,
+      "loss": 0.5409,
+      "step": 18557
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.12752127947478e-06,
+      "loss": 0.6219,
+      "step": 18558
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.126496915532398e-06,
+      "loss": 0.5452,
+      "step": 18559
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.125472618653702e-06,
+      "loss": 0.5899,
+      "step": 18560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.124448388852785e-06,
+      "loss": 0.5859,
+      "step": 18561
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.12342422614374e-06,
+      "loss": 0.5983,
+      "step": 18562
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.122400130540666e-06,
+      "loss": 0.5835,
+      "step": 18563
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.121376102057651e-06,
+      "loss": 0.5818,
+      "step": 18564
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.120352140708788e-06,
+      "loss": 0.5578,
+      "step": 18565
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.119328246508163e-06,
+      "loss": 0.5808,
+      "step": 18566
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1183044194698735e-06,
+      "loss": 0.5682,
+      "step": 18567
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.117280659608006e-06,
+      "loss": 0.6175,
+      "step": 18568
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.116256966936641e-06,
+      "loss": 0.5559,
+      "step": 18569
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1152333414698774e-06,
+      "loss": 0.6557,
+      "step": 18570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.114209783221794e-06,
+      "loss": 0.5574,
+      "step": 18571
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.113186292206478e-06,
+      "loss": 0.6715,
+      "step": 18572
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.112162868438008e-06,
+      "loss": 0.6071,
+      "step": 18573
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.111139511930475e-06,
+      "loss": 0.5968,
+      "step": 18574
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.110116222697959e-06,
+      "loss": 0.62,
+      "step": 18575
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.109093000754536e-06,
+      "loss": 0.5221,
+      "step": 18576
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.108069846114294e-06,
+      "loss": 0.6029,
+      "step": 18577
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.107046758791311e-06,
+      "loss": 0.6064,
+      "step": 18578
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.106023738799664e-06,
+      "loss": 0.6292,
+      "step": 18579
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.105000786153424e-06,
+      "loss": 0.6063,
+      "step": 18580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.103977900866681e-06,
+      "loss": 0.6446,
+      "step": 18581
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.102955082953502e-06,
+      "loss": 0.6314,
+      "step": 18582
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.101932332427965e-06,
+      "loss": 0.6188,
+      "step": 18583
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.10090964930414e-06,
+      "loss": 0.5553,
+      "step": 18584
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.099887033596106e-06,
+      "loss": 0.5853,
+      "step": 18585
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0988644853179315e-06,
+      "loss": 0.6287,
+      "step": 18586
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0978420044836855e-06,
+      "loss": 0.6346,
+      "step": 18587
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.096819591107445e-06,
+      "loss": 0.4827,
+      "step": 18588
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.095797245203273e-06,
+      "loss": 0.5501,
+      "step": 18589
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.094774966785242e-06,
+      "loss": 0.6169,
+      "step": 18590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.093752755867413e-06,
+      "loss": 0.5957,
+      "step": 18591
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.092730612463861e-06,
+      "loss": 0.5631,
+      "step": 18592
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.091708536588648e-06,
+      "loss": 0.548,
+      "step": 18593
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.090686528255838e-06,
+      "loss": 0.5818,
+      "step": 18594
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.089664587479491e-06,
+      "loss": 0.6126,
+      "step": 18595
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0886427142736785e-06,
+      "loss": 0.5902,
+      "step": 18596
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.087620908652458e-06,
+      "loss": 0.6092,
+      "step": 18597
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.086599170629885e-06,
+      "loss": 0.5763,
+      "step": 18598
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.085577500220029e-06,
+      "loss": 0.5129,
+      "step": 18599
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0845558974369445e-06,
+      "loss": 0.5323,
+      "step": 18600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.083534362294691e-06,
+      "loss": 0.6015,
+      "step": 18601
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.082512894807321e-06,
+      "loss": 0.6407,
+      "step": 18602
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.081491494988897e-06,
+      "loss": 0.5851,
+      "step": 18603
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.080470162853473e-06,
+      "loss": 0.6309,
+      "step": 18604
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0794488984151005e-06,
+      "loss": 0.5892,
+      "step": 18605
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.078427701687832e-06,
+      "loss": 0.536,
+      "step": 18606
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.077406572685727e-06,
+      "loss": 0.6273,
+      "step": 18607
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.076385511422832e-06,
+      "loss": 0.6524,
+      "step": 18608
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.075364517913196e-06,
+      "loss": 0.6008,
+      "step": 18609
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.074343592170874e-06,
+      "loss": 0.635,
+      "step": 18610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.073322734209913e-06,
+      "loss": 0.6422,
+      "step": 18611
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.07230194404436e-06,
+      "loss": 0.5855,
+      "step": 18612
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.071281221688258e-06,
+      "loss": 0.5275,
+      "step": 18613
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.070260567155661e-06,
+      "loss": 0.5903,
+      "step": 18614
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.069239980460612e-06,
+      "loss": 0.6323,
+      "step": 18615
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.068219461617151e-06,
+      "loss": 0.534,
+      "step": 18616
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.067199010639321e-06,
+      "loss": 0.5952,
+      "step": 18617
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.06617862754117e-06,
+      "loss": 0.6177,
+      "step": 18618
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.065158312336738e-06,
+      "loss": 0.5449,
+      "step": 18619
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.064138065040058e-06,
+      "loss": 0.6045,
+      "step": 18620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.063117885665179e-06,
+      "loss": 0.5463,
+      "step": 18621
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.062097774226137e-06,
+      "loss": 0.5987,
+      "step": 18622
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.061077730736968e-06,
+      "loss": 0.6044,
+      "step": 18623
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.060057755211705e-06,
+      "loss": 0.6565,
+      "step": 18624
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.059037847664392e-06,
+      "loss": 0.5594,
+      "step": 18625
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.05801800810906e-06,
+      "loss": 0.5902,
+      "step": 18626
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.056998236559738e-06,
+      "loss": 0.6051,
+      "step": 18627
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.055978533030469e-06,
+      "loss": 0.5796,
+      "step": 18628
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.054958897535278e-06,
+      "loss": 0.5349,
+      "step": 18629
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.053939330088199e-06,
+      "loss": 0.5349,
+      "step": 18630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0529198307032555e-06,
+      "loss": 0.6286,
+      "step": 18631
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.051900399394487e-06,
+      "loss": 0.5303,
+      "step": 18632
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0508810361759165e-06,
+      "loss": 0.5447,
+      "step": 18633
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.049861741061571e-06,
+      "loss": 0.6299,
+      "step": 18634
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.048842514065475e-06,
+      "loss": 0.5397,
+      "step": 18635
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0478233552016585e-06,
+      "loss": 0.5674,
+      "step": 18636
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.046804264484145e-06,
+      "loss": 0.6161,
+      "step": 18637
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.045785241926954e-06,
+      "loss": 0.6362,
+      "step": 18638
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.044766287544114e-06,
+      "loss": 0.532,
+      "step": 18639
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.043747401349643e-06,
+      "loss": 0.5387,
+      "step": 18640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.042728583357566e-06,
+      "loss": 0.5986,
+      "step": 18641
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.041709833581893e-06,
+      "loss": 0.5583,
+      "step": 18642
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0406911520366545e-06,
+      "loss": 0.626,
+      "step": 18643
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.039672538735861e-06,
+      "loss": 0.5808,
+      "step": 18644
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0386539936935345e-06,
+      "loss": 0.5903,
+      "step": 18645
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.037635516923684e-06,
+      "loss": 0.5596,
+      "step": 18646
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.036617108440332e-06,
+      "loss": 0.5722,
+      "step": 18647
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.035598768257495e-06,
+      "loss": 0.6176,
+      "step": 18648
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.034580496389172e-06,
+      "loss": 0.6328,
+      "step": 18649
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.03356229284939e-06,
+      "loss": 0.5145,
+      "step": 18650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.032544157652152e-06,
+      "loss": 0.5568,
+      "step": 18651
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.031526090811469e-06,
+      "loss": 0.6154,
+      "step": 18652
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.030508092341357e-06,
+      "loss": 0.6243,
+      "step": 18653
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0294901622558204e-06,
+      "loss": 0.5984,
+      "step": 18654
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.028472300568865e-06,
+      "loss": 0.5396,
+      "step": 18655
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.027454507294497e-06,
+      "loss": 0.5633,
+      "step": 18656
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.026436782446727e-06,
+      "loss": 0.6249,
+      "step": 18657
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.025419126039557e-06,
+      "loss": 0.5543,
+      "step": 18658
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.024401538086991e-06,
+      "loss": 0.6254,
+      "step": 18659
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.023384018603029e-06,
+      "loss": 0.6122,
+      "step": 18660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.022366567601679e-06,
+      "loss": 0.5652,
+      "step": 18661
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0213491850969375e-06,
+      "loss": 0.6234,
+      "step": 18662
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.020331871102804e-06,
+      "loss": 0.4978,
+      "step": 18663
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.019314625633283e-06,
+      "loss": 0.6182,
+      "step": 18664
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.018297448702367e-06,
+      "loss": 0.5913,
+      "step": 18665
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0172803403240575e-06,
+      "loss": 0.5241,
+      "step": 18666
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.016263300512345e-06,
+      "loss": 0.5728,
+      "step": 18667
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0152463292812316e-06,
+      "loss": 0.5534,
+      "step": 18668
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.014229426644708e-06,
+      "loss": 0.5902,
+      "step": 18669
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.013212592616765e-06,
+      "loss": 0.585,
+      "step": 18670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.012195827211404e-06,
+      "loss": 0.5583,
+      "step": 18671
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.011179130442609e-06,
+      "loss": 0.5742,
+      "step": 18672
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0101625023243734e-06,
+      "loss": 0.5878,
+      "step": 18673
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.009145942870684e-06,
+      "loss": 0.5686,
+      "step": 18674
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.008129452095533e-06,
+      "loss": 0.6432,
+      "step": 18675
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.007113030012908e-06,
+      "loss": 0.5281,
+      "step": 18676
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.006096676636795e-06,
+      "loss": 0.6107,
+      "step": 18677
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.005080391981176e-06,
+      "loss": 0.588,
+      "step": 18678
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0040641760600425e-06,
+      "loss": 0.5653,
+      "step": 18679
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0030480288873775e-06,
+      "loss": 0.6424,
+      "step": 18680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.002031950477157e-06,
+      "loss": 0.5537,
+      "step": 18681
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.001015940843374e-06,
+      "loss": 0.6502,
+      "step": 18682
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.000000000000003e-06,
+      "loss": 0.5869,
+      "step": 18683
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.998984127961025e-06,
+      "loss": 0.5742,
+      "step": 18684
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.997968324740416e-06,
+      "loss": 0.5724,
+      "step": 18685
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.996952590352162e-06,
+      "loss": 0.5529,
+      "step": 18686
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.995936924810237e-06,
+      "loss": 0.6254,
+      "step": 18687
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9949213281286165e-06,
+      "loss": 0.5613,
+      "step": 18688
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.993905800321272e-06,
+      "loss": 0.5763,
+      "step": 18689
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.992890341402187e-06,
+      "loss": 0.5984,
+      "step": 18690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.99187495138533e-06,
+      "loss": 0.5663,
+      "step": 18691
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.990859630284671e-06,
+      "loss": 0.5907,
+      "step": 18692
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.989844378114188e-06,
+      "loss": 0.6018,
+      "step": 18693
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.98882919488785e-06,
+      "loss": 0.5915,
+      "step": 18694
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.987814080619624e-06,
+      "loss": 0.6104,
+      "step": 18695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.986799035323477e-06,
+      "loss": 0.6209,
+      "step": 18696
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9857840590133845e-06,
+      "loss": 0.5405,
+      "step": 18697
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.984769151703308e-06,
+      "loss": 0.5962,
+      "step": 18698
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.983754313407216e-06,
+      "loss": 0.6383,
+      "step": 18699
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.982739544139068e-06,
+      "loss": 0.548,
+      "step": 18700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.981724843912836e-06,
+      "loss": 0.6221,
+      "step": 18701
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.98071021274248e-06,
+      "loss": 0.5896,
+      "step": 18702
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.979695650641958e-06,
+      "loss": 0.647,
+      "step": 18703
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9786811576252395e-06,
+      "loss": 0.5452,
+      "step": 18704
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.977666733706279e-06,
+      "loss": 0.5918,
+      "step": 18705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.976652378899038e-06,
+      "loss": 0.5587,
+      "step": 18706
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.975638093217471e-06,
+      "loss": 0.5377,
+      "step": 18707
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.974623876675541e-06,
+      "loss": 0.6178,
+      "step": 18708
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.973609729287204e-06,
+      "loss": 0.591,
+      "step": 18709
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.972595651066412e-06,
+      "loss": 0.5107,
+      "step": 18710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.971581642027118e-06,
+      "loss": 0.5747,
+      "step": 18711
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.970567702183283e-06,
+      "loss": 0.604,
+      "step": 18712
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.969553831548856e-06,
+      "loss": 0.6769,
+      "step": 18713
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.968540030137783e-06,
+      "loss": 0.5825,
+      "step": 18714
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.967526297964025e-06,
+      "loss": 0.5522,
+      "step": 18715
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9665126350415275e-06,
+      "loss": 0.629,
+      "step": 18716
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.965499041384238e-06,
+      "loss": 0.632,
+      "step": 18717
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.964485517006102e-06,
+      "loss": 0.5753,
+      "step": 18718
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.963472061921075e-06,
+      "loss": 0.563,
+      "step": 18719
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.962458676143097e-06,
+      "loss": 0.5967,
+      "step": 18720
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.961445359686113e-06,
+      "loss": 0.5853,
+      "step": 18721
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.960432112564065e-06,
+      "loss": 0.6304,
+      "step": 18722
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.959418934790904e-06,
+      "loss": 0.6061,
+      "step": 18723
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.958405826380566e-06,
+      "loss": 0.6792,
+      "step": 18724
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.957392787346991e-06,
+      "loss": 0.5867,
+      "step": 18725
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.956379817704125e-06,
+      "loss": 0.5575,
+      "step": 18726
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.955366917465906e-06,
+      "loss": 0.5836,
+      "step": 18727
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.954354086646269e-06,
+      "loss": 0.5605,
+      "step": 18728
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.95334132525915e-06,
+      "loss": 0.5902,
+      "step": 18729
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.952328633318493e-06,
+      "loss": 0.6335,
+      "step": 18730
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9513160108382295e-06,
+      "loss": 0.5879,
+      "step": 18731
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9503034578322895e-06,
+      "loss": 0.5526,
+      "step": 18732
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.949290974314614e-06,
+      "loss": 0.655,
+      "step": 18733
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.948278560299135e-06,
+      "loss": 0.5194,
+      "step": 18734
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9472662157997795e-06,
+      "loss": 0.5386,
+      "step": 18735
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9462539408304774e-06,
+      "loss": 0.5796,
+      "step": 18736
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.945241735405166e-06,
+      "loss": 0.5517,
+      "step": 18737
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.94422959953777e-06,
+      "loss": 0.536,
+      "step": 18738
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.943217533242216e-06,
+      "loss": 0.5704,
+      "step": 18739
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.942205536532429e-06,
+      "loss": 0.5109,
+      "step": 18740
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.941193609422342e-06,
+      "loss": 0.619,
+      "step": 18741
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.940181751925876e-06,
+      "loss": 0.5782,
+      "step": 18742
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.93916996405695e-06,
+      "loss": 0.523,
+      "step": 18743
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.938158245829497e-06,
+      "loss": 0.6166,
+      "step": 18744
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.937146597257434e-06,
+      "loss": 0.5946,
+      "step": 18745
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.936135018354683e-06,
+      "loss": 0.6091,
+      "step": 18746
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.935123509135159e-06,
+      "loss": 0.5868,
+      "step": 18747
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.93411206961279e-06,
+      "loss": 0.6222,
+      "step": 18748
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.933100699801491e-06,
+      "loss": 0.5951,
+      "step": 18749
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.932089399715178e-06,
+      "loss": 0.6408,
+      "step": 18750
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.931078169367763e-06,
+      "loss": 0.5505,
+      "step": 18751
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.930067008773172e-06,
+      "loss": 0.5276,
+      "step": 18752
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9290559179453125e-06,
+      "loss": 0.5713,
+      "step": 18753
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.928044896898096e-06,
+      "loss": 0.5656,
+      "step": 18754
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.927033945645441e-06,
+      "loss": 0.6236,
+      "step": 18755
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.926023064201257e-06,
+      "loss": 0.5848,
+      "step": 18756
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9250122525794555e-06,
+      "loss": 0.5512,
+      "step": 18757
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.92400151079394e-06,
+      "loss": 0.5416,
+      "step": 18758
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9229908388586264e-06,
+      "loss": 0.5714,
+      "step": 18759
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.921980236787422e-06,
+      "loss": 0.6292,
+      "step": 18760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.920969704594229e-06,
+      "loss": 0.6038,
+      "step": 18761
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9199592422929545e-06,
+      "loss": 0.6091,
+      "step": 18762
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.918948849897507e-06,
+      "loss": 0.5996,
+      "step": 18763
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.917938527421788e-06,
+      "loss": 0.5727,
+      "step": 18764
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.916928274879696e-06,
+      "loss": 0.5811,
+      "step": 18765
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.915918092285142e-06,
+      "loss": 0.5845,
+      "step": 18766
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.914907979652023e-06,
+      "loss": 0.6821,
+      "step": 18767
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.913897936994236e-06,
+      "loss": 0.5693,
+      "step": 18768
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.912887964325681e-06,
+      "loss": 0.5461,
+      "step": 18769
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.91187806166026e-06,
+      "loss": 0.6193,
+      "step": 18770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.910868229011868e-06,
+      "loss": 0.523,
+      "step": 18771
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9098584663944015e-06,
+      "loss": 0.6063,
+      "step": 18772
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.908848773821751e-06,
+      "loss": 0.5605,
+      "step": 18773
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.907839151307819e-06,
+      "loss": 0.6624,
+      "step": 18774
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.906829598866493e-06,
+      "loss": 0.529,
+      "step": 18775
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.905820116511664e-06,
+      "loss": 0.5897,
+      "step": 18776
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.90481070425723e-06,
+      "loss": 0.5757,
+      "step": 18777
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.903801362117079e-06,
+      "loss": 0.6378,
+      "step": 18778
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.902792090105097e-06,
+      "loss": 0.5708,
+      "step": 18779
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.901782888235173e-06,
+      "loss": 0.5742,
+      "step": 18780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.900773756521199e-06,
+      "loss": 0.5942,
+      "step": 18781
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8997646949770586e-06,
+      "loss": 0.6091,
+      "step": 18782
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.898755703616638e-06,
+      "loss": 0.6206,
+      "step": 18783
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.897746782453818e-06,
+      "loss": 0.6032,
+      "step": 18784
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.896737931502487e-06,
+      "loss": 0.6007,
+      "step": 18785
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.895729150776528e-06,
+      "loss": 0.5669,
+      "step": 18786
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.894720440289816e-06,
+      "loss": 0.5478,
+      "step": 18787
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.893711800056241e-06,
+      "loss": 0.5853,
+      "step": 18788
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8927032300896794e-06,
+      "loss": 0.6349,
+      "step": 18789
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.891694730404007e-06,
+      "loss": 0.5705,
+      "step": 18790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.890686301013101e-06,
+      "loss": 0.5615,
+      "step": 18791
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.889677941930845e-06,
+      "loss": 0.5688,
+      "step": 18792
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.888669653171111e-06,
+      "loss": 0.5914,
+      "step": 18793
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8876614347477695e-06,
+      "loss": 0.588,
+      "step": 18794
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.886653286674702e-06,
+      "loss": 0.5715,
+      "step": 18795
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.885645208965779e-06,
+      "loss": 0.5918,
+      "step": 18796
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.884637201634872e-06,
+      "loss": 0.6015,
+      "step": 18797
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.883629264695847e-06,
+      "loss": 0.5933,
+      "step": 18798
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.882621398162583e-06,
+      "loss": 0.5731,
+      "step": 18799
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.881613602048946e-06,
+      "loss": 0.6012,
+      "step": 18800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8806058763688024e-06,
+      "loss": 0.5831,
+      "step": 18801
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8795982211360174e-06,
+      "loss": 0.5661,
+      "step": 18802
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.878590636364463e-06,
+      "loss": 0.6776,
+      "step": 18803
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.877583122068002e-06,
+      "loss": 0.516,
+      "step": 18804
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.876575678260493e-06,
+      "loss": 0.5893,
+      "step": 18805
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.87556830495581e-06,
+      "loss": 0.5765,
+      "step": 18806
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.874561002167809e-06,
+      "loss": 0.6172,
+      "step": 18807
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.873553769910353e-06,
+      "loss": 0.5973,
+      "step": 18808
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.872546608197297e-06,
+      "loss": 0.5937,
+      "step": 18809
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.871539517042509e-06,
+      "loss": 0.5423,
+      "step": 18810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.870532496459844e-06,
+      "loss": 0.568,
+      "step": 18811
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.869525546463158e-06,
+      "loss": 0.5954,
+      "step": 18812
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.868518667066307e-06,
+      "loss": 0.6239,
+      "step": 18813
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8675118582831496e-06,
+      "loss": 0.6215,
+      "step": 18814
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.866505120127539e-06,
+      "loss": 0.548,
+      "step": 18815
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.865498452613326e-06,
+      "loss": 0.5221,
+      "step": 18816
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.864491855754369e-06,
+      "loss": 0.5599,
+      "step": 18817
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.863485329564516e-06,
+      "loss": 0.5553,
+      "step": 18818
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8624788740576176e-06,
+      "loss": 0.5916,
+      "step": 18819
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.861472489247522e-06,
+      "loss": 0.5125,
+      "step": 18820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.860466175148083e-06,
+      "loss": 0.5947,
+      "step": 18821
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8594599317731465e-06,
+      "loss": 0.5943,
+      "step": 18822
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.858453759136557e-06,
+      "loss": 0.5264,
+      "step": 18823
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.857447657252158e-06,
+      "loss": 0.5738,
+      "step": 18824
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.856441626133802e-06,
+      "loss": 0.547,
+      "step": 18825
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.855435665795329e-06,
+      "loss": 0.6034,
+      "step": 18826
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.854429776250576e-06,
+      "loss": 0.5182,
+      "step": 18827
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.853423957513396e-06,
+      "loss": 0.5825,
+      "step": 18828
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.852418209597625e-06,
+      "loss": 0.5805,
+      "step": 18829
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.851412532517102e-06,
+      "loss": 0.6663,
+      "step": 18830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.850406926285661e-06,
+      "loss": 0.585,
+      "step": 18831
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.849401390917151e-06,
+      "loss": 0.6012,
+      "step": 18832
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.848395926425404e-06,
+      "loss": 0.5034,
+      "step": 18833
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.847390532824254e-06,
+      "loss": 0.6215,
+      "step": 18834
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8463852101275354e-06,
+      "loss": 0.5897,
+      "step": 18835
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.845379958349088e-06,
+      "loss": 0.5841,
+      "step": 18836
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.844374777502742e-06,
+      "loss": 0.5384,
+      "step": 18837
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8433696676023245e-06,
+      "loss": 0.5776,
+      "step": 18838
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.842364628661676e-06,
+      "loss": 0.6167,
+      "step": 18839
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.841359660694623e-06,
+      "loss": 0.586,
+      "step": 18840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.840354763714991e-06,
+      "loss": 0.5608,
+      "step": 18841
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.839349937736609e-06,
+      "loss": 0.5789,
+      "step": 18842
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.838345182773311e-06,
+      "loss": 0.5654,
+      "step": 18843
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.837340498838917e-06,
+      "loss": 0.6237,
+      "step": 18844
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.836335885947255e-06,
+      "loss": 0.6221,
+      "step": 18845
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.835331344112145e-06,
+      "loss": 0.6713,
+      "step": 18846
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8343268733474165e-06,
+      "loss": 0.5294,
+      "step": 18847
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.833322473666891e-06,
+      "loss": 0.5521,
+      "step": 18848
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.832318145084383e-06,
+      "loss": 0.6072,
+      "step": 18849
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.831313887613721e-06,
+      "loss": 0.5931,
+      "step": 18850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8303097012687226e-06,
+      "loss": 0.5847,
+      "step": 18851
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.829305586063206e-06,
+      "loss": 0.5987,
+      "step": 18852
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8283015420109825e-06,
+      "loss": 0.678,
+      "step": 18853
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.827297569125879e-06,
+      "loss": 0.5494,
+      "step": 18854
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.826293667421705e-06,
+      "loss": 0.6434,
+      "step": 18855
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.825289836912273e-06,
+      "loss": 0.5793,
+      "step": 18856
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.824286077611404e-06,
+      "loss": 0.5473,
+      "step": 18857
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.823282389532905e-06,
+      "loss": 0.6426,
+      "step": 18858
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.82227877269059e-06,
+      "loss": 0.5862,
+      "step": 18859
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8212752270982685e-06,
+      "loss": 0.5864,
+      "step": 18860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.820271752769751e-06,
+      "loss": 0.5715,
+      "step": 18861
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.819268349718844e-06,
+      "loss": 0.6012,
+      "step": 18862
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8182650179593535e-06,
+      "loss": 0.5492,
+      "step": 18863
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.817261757505094e-06,
+      "loss": 0.6572,
+      "step": 18864
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.816258568369866e-06,
+      "loss": 0.5735,
+      "step": 18865
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.815255450567474e-06,
+      "loss": 0.5629,
+      "step": 18866
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8142524041117215e-06,
+      "loss": 0.6114,
+      "step": 18867
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.813249429016413e-06,
+      "loss": 0.6264,
+      "step": 18868
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.812246525295352e-06,
+      "loss": 0.639,
+      "step": 18869
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.811243692962333e-06,
+      "loss": 0.5956,
+      "step": 18870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.810240932031164e-06,
+      "loss": 0.5504,
+      "step": 18871
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.809238242515639e-06,
+      "loss": 0.6669,
+      "step": 18872
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.808235624429558e-06,
+      "loss": 0.5695,
+      "step": 18873
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.807233077786712e-06,
+      "loss": 0.5674,
+      "step": 18874
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.806230602600907e-06,
+      "loss": 0.5727,
+      "step": 18875
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.805228198885931e-06,
+      "loss": 0.5553,
+      "step": 18876
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.804225866655578e-06,
+      "loss": 0.5899,
+      "step": 18877
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.803223605923641e-06,
+      "loss": 0.5514,
+      "step": 18878
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8022214167039175e-06,
+      "loss": 0.5864,
+      "step": 18879
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.801219299010194e-06,
+      "loss": 0.5774,
+      "step": 18880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.800217252856256e-06,
+      "loss": 0.6024,
+      "step": 18881
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.799215278255902e-06,
+      "loss": 0.5911,
+      "step": 18882
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.798213375222916e-06,
+      "loss": 0.6365,
+      "step": 18883
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.797211543771084e-06,
+      "loss": 0.5327,
+      "step": 18884
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7962097839141895e-06,
+      "loss": 0.6134,
+      "step": 18885
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.795208095666025e-06,
+      "loss": 0.5809,
+      "step": 18886
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.79420647904037e-06,
+      "loss": 0.622,
+      "step": 18887
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7932049340510045e-06,
+      "loss": 0.6165,
+      "step": 18888
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.792203460711718e-06,
+      "loss": 0.6937,
+      "step": 18889
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.791202059036288e-06,
+      "loss": 0.5308,
+      "step": 18890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.790200729038493e-06,
+      "loss": 0.5548,
+      "step": 18891
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.789199470732111e-06,
+      "loss": 0.5918,
+      "step": 18892
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.788198284130927e-06,
+      "loss": 0.5747,
+      "step": 18893
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.787197169248713e-06,
+      "loss": 0.5197,
+      "step": 18894
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7861961260992474e-06,
+      "loss": 0.6143,
+      "step": 18895
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7851951546963e-06,
+      "loss": 0.5677,
+      "step": 18896
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7841942550536535e-06,
+      "loss": 0.5705,
+      "step": 18897
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.783193427185078e-06,
+      "loss": 0.6467,
+      "step": 18898
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.78219267110434e-06,
+      "loss": 0.6016,
+      "step": 18899
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.781191986825219e-06,
+      "loss": 0.5622,
+      "step": 18900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.780191374361482e-06,
+      "loss": 0.603,
+      "step": 18901
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.779190833726899e-06,
+      "loss": 0.6367,
+      "step": 18902
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.778190364935232e-06,
+      "loss": 0.581,
+      "step": 18903
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.77718996800026e-06,
+      "loss": 0.5776,
+      "step": 18904
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.776189642935742e-06,
+      "loss": 0.5276,
+      "step": 18905
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.775189389755445e-06,
+      "loss": 0.5448,
+      "step": 18906
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.774189208473129e-06,
+      "loss": 0.6105,
+      "step": 18907
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7731890991025645e-06,
+      "loss": 0.5255,
+      "step": 18908
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.772189061657511e-06,
+      "loss": 0.5337,
+      "step": 18909
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.771189096151726e-06,
+      "loss": 0.6086,
+      "step": 18910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.770189202598977e-06,
+      "loss": 0.5921,
+      "step": 18911
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.769189381013019e-06,
+      "loss": 0.5439,
+      "step": 18912
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.768189631407614e-06,
+      "loss": 0.5685,
+      "step": 18913
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.76718995379651e-06,
+      "loss": 0.6094,
+      "step": 18914
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.766190348193475e-06,
+      "loss": 0.6517,
+      "step": 18915
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7651908146122605e-06,
+      "loss": 0.6009,
+      "step": 18916
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7641913530666195e-06,
+      "loss": 0.6352,
+      "step": 18917
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.763191963570302e-06,
+      "loss": 0.4952,
+      "step": 18918
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.762192646137069e-06,
+      "loss": 0.5782,
+      "step": 18919
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.761193400780667e-06,
+      "loss": 0.5654,
+      "step": 18920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.760194227514844e-06,
+      "loss": 0.5629,
+      "step": 18921
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7591951263533555e-06,
+      "loss": 0.6561,
+      "step": 18922
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.758196097309947e-06,
+      "loss": 0.5384,
+      "step": 18923
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7571971403983665e-06,
+      "loss": 0.4784,
+      "step": 18924
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.756198255632356e-06,
+      "loss": 0.5309,
+      "step": 18925
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.755199443025669e-06,
+      "loss": 0.584,
+      "step": 18926
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.754200702592046e-06,
+      "loss": 0.5773,
+      "step": 18927
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.753202034345231e-06,
+      "loss": 0.5788,
+      "step": 18928
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.752203438298962e-06,
+      "loss": 0.6003,
+      "step": 18929
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.751204914466989e-06,
+      "loss": 0.5986,
+      "step": 18930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7502064628630484e-06,
+      "loss": 0.6095,
+      "step": 18931
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.749208083500876e-06,
+      "loss": 0.636,
+      "step": 18932
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.748209776394217e-06,
+      "loss": 0.6194,
+      "step": 18933
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.747211541556807e-06,
+      "loss": 0.6132,
+      "step": 18934
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.746213379002381e-06,
+      "loss": 0.5459,
+      "step": 18935
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.745215288744673e-06,
+      "loss": 0.5716,
+      "step": 18936
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.744217270797422e-06,
+      "loss": 0.5627,
+      "step": 18937
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7432193251743595e-06,
+      "loss": 0.5696,
+      "step": 18938
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7422214518892195e-06,
+      "loss": 0.5765,
+      "step": 18939
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.741223650955727e-06,
+      "loss": 0.5578,
+      "step": 18940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.740225922387623e-06,
+      "loss": 0.5305,
+      "step": 18941
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.739228266198631e-06,
+      "loss": 0.6595,
+      "step": 18942
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.738230682402477e-06,
+      "loss": 0.6526,
+      "step": 18943
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.737233171012896e-06,
+      "loss": 0.6299,
+      "step": 18944
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.736235732043611e-06,
+      "loss": 0.604,
+      "step": 18945
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.735238365508348e-06,
+      "loss": 0.5734,
+      "step": 18946
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.734241071420826e-06,
+      "loss": 0.5472,
+      "step": 18947
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.733243849794779e-06,
+      "loss": 0.5341,
+      "step": 18948
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7322467006439235e-06,
+      "loss": 0.5841,
+      "step": 18949
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.731249623981979e-06,
+      "loss": 0.521,
+      "step": 18950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.73025261982267e-06,
+      "loss": 0.5954,
+      "step": 18951
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.729255688179719e-06,
+      "loss": 0.5524,
+      "step": 18952
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.72825882906684e-06,
+      "loss": 0.5875,
+      "step": 18953
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.727262042497747e-06,
+      "loss": 0.6831,
+      "step": 18954
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.726265328486166e-06,
+      "loss": 0.5491,
+      "step": 18955
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7252686870458075e-06,
+      "loss": 0.5911,
+      "step": 18956
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.724272118190387e-06,
+      "loss": 0.5621,
+      "step": 18957
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.723275621933614e-06,
+      "loss": 0.5912,
+      "step": 18958
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.722279198289209e-06,
+      "loss": 0.5946,
+      "step": 18959
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.72128284727088e-06,
+      "loss": 0.58,
+      "step": 18960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.720286568892331e-06,
+      "loss": 0.6121,
+      "step": 18961
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7192903631672845e-06,
+      "loss": 0.5909,
+      "step": 18962
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7182942301094416e-06,
+      "loss": 0.6513,
+      "step": 18963
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.717298169732512e-06,
+      "loss": 0.6247,
+      "step": 18964
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.716302182050195e-06,
+      "loss": 0.5394,
+      "step": 18965
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.71530626707621e-06,
+      "loss": 0.5844,
+      "step": 18966
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.714310424824251e-06,
+      "loss": 0.6223,
+      "step": 18967
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.713314655308027e-06,
+      "loss": 0.5966,
+      "step": 18968
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.712318958541235e-06,
+      "loss": 0.6191,
+      "step": 18969
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.711323334537582e-06,
+      "loss": 0.6064,
+      "step": 18970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.710327783310769e-06,
+      "loss": 0.5937,
+      "step": 18971
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.709332304874489e-06,
+      "loss": 0.6291,
+      "step": 18972
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.708336899242449e-06,
+      "loss": 0.5146,
+      "step": 18973
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.707341566428343e-06,
+      "loss": 0.5386,
+      "step": 18974
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.706346306445868e-06,
+      "loss": 0.6205,
+      "step": 18975
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.705351119308716e-06,
+      "loss": 0.6151,
+      "step": 18976
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.704356005030587e-06,
+      "loss": 0.5834,
+      "step": 18977
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7033609636251735e-06,
+      "loss": 0.5729,
+      "step": 18978
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.702365995106166e-06,
+      "loss": 0.5298,
+      "step": 18979
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.701371099487256e-06,
+      "loss": 0.6004,
+      "step": 18980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.700376276782137e-06,
+      "loss": 0.5483,
+      "step": 18981
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6993815270044985e-06,
+      "loss": 0.5666,
+      "step": 18982
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.698386850168023e-06,
+      "loss": 0.5639,
+      "step": 18983
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6973922462864066e-06,
+      "loss": 0.623,
+      "step": 18984
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.696397715373333e-06,
+      "loss": 0.5874,
+      "step": 18985
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.695403257442486e-06,
+      "loss": 0.569,
+      "step": 18986
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6944088725075475e-06,
+      "loss": 0.5043,
+      "step": 18987
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.69341456058221e-06,
+      "loss": 0.5361,
+      "step": 18988
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.69242032168015e-06,
+      "loss": 0.5149,
+      "step": 18989
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.69142615581505e-06,
+      "loss": 0.6121,
+      "step": 18990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.690432063000586e-06,
+      "loss": 0.543,
+      "step": 18991
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.689438043250447e-06,
+      "loss": 0.6295,
+      "step": 18992
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.688444096578307e-06,
+      "loss": 0.625,
+      "step": 18993
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.687450222997839e-06,
+      "loss": 0.5666,
+      "step": 18994
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.686456422522728e-06,
+      "loss": 0.5969,
+      "step": 18995
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.685462695166645e-06,
+      "loss": 0.5432,
+      "step": 18996
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.684469040943266e-06,
+      "loss": 0.5318,
+      "step": 18997
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.683475459866259e-06,
+      "loss": 0.6446,
+      "step": 18998
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.682481951949306e-06,
+      "loss": 0.5603,
+      "step": 18999
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.681488517206072e-06,
+      "loss": 0.5697,
+      "step": 19000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.680495155650231e-06,
+      "loss": 0.5691,
+      "step": 19001
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.679501867295446e-06,
+      "loss": 0.5963,
+      "step": 19002
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.678508652155395e-06,
+      "loss": 0.5768,
+      "step": 19003
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.67751551024374e-06,
+      "loss": 0.528,
+      "step": 19004
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.676522441574145e-06,
+      "loss": 0.6321,
+      "step": 19005
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.675529446160282e-06,
+      "loss": 0.6143,
+      "step": 19006
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.674536524015812e-06,
+      "loss": 0.5654,
+      "step": 19007
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.673543675154399e-06,
+      "loss": 0.6064,
+      "step": 19008
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6725508995897025e-06,
+      "loss": 0.5305,
+      "step": 19009
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6715581973353905e-06,
+      "loss": 0.6012,
+      "step": 19010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.670565568405119e-06,
+      "loss": 0.6311,
+      "step": 19011
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.669573012812545e-06,
+      "loss": 0.5954,
+      "step": 19012
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.668580530571333e-06,
+      "loss": 0.5427,
+      "step": 19013
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.667588121695138e-06,
+      "loss": 0.575,
+      "step": 19014
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6665957861976154e-06,
+      "loss": 0.568,
+      "step": 19015
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.665603524092418e-06,
+      "loss": 0.5224,
+      "step": 19016
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.664611335393207e-06,
+      "loss": 0.5924,
+      "step": 19017
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.663619220113631e-06,
+      "loss": 0.5122,
+      "step": 19018
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.662627178267345e-06,
+      "loss": 0.5803,
+      "step": 19019
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.661635209867994e-06,
+      "loss": 0.6097,
+      "step": 19020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.660643314929238e-06,
+      "loss": 0.578,
+      "step": 19021
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.659651493464721e-06,
+      "loss": 0.5779,
+      "step": 19022
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6586597454880874e-06,
+      "loss": 0.6422,
+      "step": 19023
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.657668071012994e-06,
+      "loss": 0.5838,
+      "step": 19024
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.656676470053081e-06,
+      "loss": 0.5737,
+      "step": 19025
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.655684942621995e-06,
+      "loss": 0.5649,
+      "step": 19026
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.654693488733377e-06,
+      "loss": 0.622,
+      "step": 19027
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6537021084008774e-06,
+      "loss": 0.5851,
+      "step": 19028
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.652710801638134e-06,
+      "loss": 0.5922,
+      "step": 19029
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.651719568458789e-06,
+      "loss": 0.5256,
+      "step": 19030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6507284088764774e-06,
+      "loss": 0.5842,
+      "step": 19031
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.649737322904847e-06,
+      "loss": 0.6382,
+      "step": 19032
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.648746310557533e-06,
+      "loss": 0.6316,
+      "step": 19033
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.647755371848167e-06,
+      "loss": 0.535,
+      "step": 19034
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.646764506790395e-06,
+      "loss": 0.6428,
+      "step": 19035
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.645773715397846e-06,
+      "loss": 0.6257,
+      "step": 19036
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.644782997684156e-06,
+      "loss": 0.5601,
+      "step": 19037
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.643792353662954e-06,
+      "loss": 0.5722,
+      "step": 19038
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.642801783347879e-06,
+      "loss": 0.5219,
+      "step": 19039
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.641811286752559e-06,
+      "loss": 0.5367,
+      "step": 19040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.640820863890623e-06,
+      "loss": 0.6105,
+      "step": 19041
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.639830514775697e-06,
+      "loss": 0.5897,
+      "step": 19042
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.638840239421418e-06,
+      "loss": 0.5785,
+      "step": 19043
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.637850037841407e-06,
+      "loss": 0.5495,
+      "step": 19044
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.636859910049288e-06,
+      "loss": 0.595,
+      "step": 19045
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6358698560586925e-06,
+      "loss": 0.6032,
+      "step": 19046
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.634879875883241e-06,
+      "loss": 0.5833,
+      "step": 19047
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6338899695365566e-06,
+      "loss": 0.6,
+      "step": 19048
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.632900137032258e-06,
+      "loss": 0.6038,
+      "step": 19049
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6319103783839726e-06,
+      "loss": 0.5935,
+      "step": 19050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.630920693605317e-06,
+      "loss": 0.521,
+      "step": 19051
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.629931082709912e-06,
+      "loss": 0.6085,
+      "step": 19052
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.628941545711369e-06,
+      "loss": 0.6335,
+      "step": 19053
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.627952082623313e-06,
+      "loss": 0.4845,
+      "step": 19054
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.626962693459357e-06,
+      "loss": 0.6101,
+      "step": 19055
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.625973378233112e-06,
+      "loss": 0.5478,
+      "step": 19056
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.624984136958199e-06,
+      "loss": 0.5439,
+      "step": 19057
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.623994969648227e-06,
+      "loss": 0.5491,
+      "step": 19058
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.623005876316808e-06,
+      "loss": 0.5695,
+      "step": 19059
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.622016856977549e-06,
+      "loss": 0.5901,
+      "step": 19060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.621027911644068e-06,
+      "loss": 0.5497,
+      "step": 19061
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.620039040329969e-06,
+      "loss": 0.6096,
+      "step": 19062
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.619050243048859e-06,
+      "loss": 0.5406,
+      "step": 19063
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.618061519814343e-06,
+      "loss": 0.6104,
+      "step": 19064
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.617072870640034e-06,
+      "loss": 0.5508,
+      "step": 19065
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.61608429553953e-06,
+      "loss": 0.621,
+      "step": 19066
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.615095794526435e-06,
+      "loss": 0.6097,
+      "step": 19067
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.614107367614357e-06,
+      "loss": 0.629,
+      "step": 19068
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.613119014816893e-06,
+      "loss": 0.5866,
+      "step": 19069
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6121307361476454e-06,
+      "loss": 0.6286,
+      "step": 19070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.611142531620213e-06,
+      "loss": 0.5961,
+      "step": 19071
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.610154401248194e-06,
+      "loss": 0.5887,
+      "step": 19072
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6091663450451876e-06,
+      "loss": 0.5835,
+      "step": 19073
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.608178363024783e-06,
+      "loss": 0.6196,
+      "step": 19074
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.607190455200588e-06,
+      "loss": 0.6346,
+      "step": 19075
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.60620262158619e-06,
+      "loss": 0.5759,
+      "step": 19076
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.605214862195179e-06,
+      "loss": 0.5427,
+      "step": 19077
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.604227177041156e-06,
+      "loss": 0.5644,
+      "step": 19078
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.603239566137708e-06,
+      "loss": 0.5921,
+      "step": 19079
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.602252029498427e-06,
+      "loss": 0.5874,
+      "step": 19080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.601264567136895e-06,
+      "loss": 0.6412,
+      "step": 19081
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.600277179066711e-06,
+      "loss": 0.5419,
+      "step": 19082
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5992898653014575e-06,
+      "loss": 0.5837,
+      "step": 19083
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.598302625854722e-06,
+      "loss": 0.6178,
+      "step": 19084
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.597315460740084e-06,
+      "loss": 0.5635,
+      "step": 19085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.596328369971136e-06,
+      "loss": 0.563,
+      "step": 19086
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.595341353561458e-06,
+      "loss": 0.5528,
+      "step": 19087
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.594354411524627e-06,
+      "loss": 0.526,
+      "step": 19088
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.593367543874234e-06,
+      "loss": 0.5692,
+      "step": 19089
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.592380750623855e-06,
+      "loss": 0.6224,
+      "step": 19090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.591394031787068e-06,
+      "loss": 0.5151,
+      "step": 19091
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.590407387377447e-06,
+      "loss": 0.6052,
+      "step": 19092
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.589420817408578e-06,
+      "loss": 0.5567,
+      "step": 19093
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.588434321894033e-06,
+      "loss": 0.6153,
+      "step": 19094
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.587447900847386e-06,
+      "loss": 0.5535,
+      "step": 19095
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.586461554282209e-06,
+      "loss": 0.5286,
+      "step": 19096
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.58547528221208e-06,
+      "loss": 0.5991,
+      "step": 19097
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.584489084650569e-06,
+      "loss": 0.679,
+      "step": 19098
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.583502961611245e-06,
+      "loss": 0.5383,
+      "step": 19099
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.582516913107682e-06,
+      "loss": 0.5897,
+      "step": 19100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.581530939153446e-06,
+      "loss": 0.5699,
+      "step": 19101
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.580545039762105e-06,
+      "loss": 0.6216,
+      "step": 19102
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.579559214947223e-06,
+      "loss": 0.5808,
+      "step": 19103
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.578573464722373e-06,
+      "loss": 0.5795,
+      "step": 19104
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.577587789101117e-06,
+      "loss": 0.6187,
+      "step": 19105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.576602188097016e-06,
+      "loss": 0.6104,
+      "step": 19106
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5756166617236305e-06,
+      "loss": 0.6378,
+      "step": 19107
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.57463120999453e-06,
+      "loss": 0.66,
+      "step": 19108
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.573645832923271e-06,
+      "loss": 0.5807,
+      "step": 19109
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5726605305234105e-06,
+      "loss": 0.5549,
+      "step": 19110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5716753028085135e-06,
+      "loss": 0.5714,
+      "step": 19111
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.570690149792134e-06,
+      "loss": 0.5452,
+      "step": 19112
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.569705071487827e-06,
+      "loss": 0.5679,
+      "step": 19113
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5687200679091484e-06,
+      "loss": 0.5305,
+      "step": 19114
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.567735139069657e-06,
+      "loss": 0.5318,
+      "step": 19115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.566750284982903e-06,
+      "loss": 0.5511,
+      "step": 19116
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.565765505662435e-06,
+      "loss": 0.5478,
+      "step": 19117
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.564780801121813e-06,
+      "loss": 0.6278,
+      "step": 19118
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5637961713745826e-06,
+      "loss": 0.6339,
+      "step": 19119
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.562811616434294e-06,
+      "loss": 0.5383,
+      "step": 19120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.561827136314491e-06,
+      "loss": 0.6301,
+      "step": 19121
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5608427310287285e-06,
+      "loss": 0.6276,
+      "step": 19122
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.55985840059055e-06,
+      "loss": 0.6277,
+      "step": 19123
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5588741450134985e-06,
+      "loss": 0.5862,
+      "step": 19124
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.557889964311117e-06,
+      "loss": 0.5624,
+      "step": 19125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.556905858496956e-06,
+      "loss": 0.6134,
+      "step": 19126
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.555921827584553e-06,
+      "loss": 0.5255,
+      "step": 19127
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5549378715874445e-06,
+      "loss": 0.5898,
+      "step": 19128
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.55395399051918e-06,
+      "loss": 0.6209,
+      "step": 19129
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.552970184393292e-06,
+      "loss": 0.62,
+      "step": 19130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5519864532233224e-06,
+      "loss": 0.5718,
+      "step": 19131
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.551002797022802e-06,
+      "loss": 0.545,
+      "step": 19132
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.550019215805275e-06,
+      "loss": 0.5978,
+      "step": 19133
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.549035709584272e-06,
+      "loss": 0.5914,
+      "step": 19134
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.548052278373327e-06,
+      "loss": 0.6378,
+      "step": 19135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.54706892218597e-06,
+      "loss": 0.6141,
+      "step": 19136
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.546085641035739e-06,
+      "loss": 0.6181,
+      "step": 19137
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.545102434936163e-06,
+      "loss": 0.6507,
+      "step": 19138
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.544119303900766e-06,
+      "loss": 0.6028,
+      "step": 19139
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.543136247943085e-06,
+      "loss": 0.5872,
+      "step": 19140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.542153267076644e-06,
+      "loss": 0.531,
+      "step": 19141
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.54117036131497e-06,
+      "loss": 0.5852,
+      "step": 19142
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.540187530671585e-06,
+      "loss": 0.617,
+      "step": 19143
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5392047751600195e-06,
+      "loss": 0.6049,
+      "step": 19144
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.538222094793796e-06,
+      "loss": 0.5959,
+      "step": 19145
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.537239489586435e-06,
+      "loss": 0.5458,
+      "step": 19146
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.536256959551455e-06,
+      "loss": 0.5868,
+      "step": 19147
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.535274504702384e-06,
+      "loss": 0.5798,
+      "step": 19148
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.534292125052737e-06,
+      "loss": 0.5869,
+      "step": 19149
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.53330982061603e-06,
+      "loss": 0.6095,
+      "step": 19150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.532327591405786e-06,
+      "loss": 0.5633,
+      "step": 19151
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5313454374355195e-06,
+      "loss": 0.5719,
+      "step": 19152
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.530363358718745e-06,
+      "loss": 0.5819,
+      "step": 19153
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.529381355268972e-06,
+      "loss": 0.57,
+      "step": 19154
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.528399427099723e-06,
+      "loss": 0.5656,
+      "step": 19155
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.527417574224505e-06,
+      "loss": 0.5783,
+      "step": 19156
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.52643579665683e-06,
+      "loss": 0.5914,
+      "step": 19157
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.525454094410205e-06,
+      "loss": 0.6188,
+      "step": 19158
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.524472467498146e-06,
+      "loss": 0.5933,
+      "step": 19159
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5234909159341555e-06,
+      "loss": 0.5912,
+      "step": 19160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.522509439731739e-06,
+      "loss": 0.5583,
+      "step": 19161
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.52152803890441e-06,
+      "loss": 0.6339,
+      "step": 19162
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.520546713465669e-06,
+      "loss": 0.5284,
+      "step": 19163
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5195654634290185e-06,
+      "loss": 0.5926,
+      "step": 19164
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.51858428880796e-06,
+      "loss": 0.7017,
+      "step": 19165
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.517603189616002e-06,
+      "loss": 0.5514,
+      "step": 19166
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.516622165866641e-06,
+      "loss": 0.5682,
+      "step": 19167
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.515641217573377e-06,
+      "loss": 0.6283,
+      "step": 19168
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.514660344749706e-06,
+      "loss": 0.5835,
+      "step": 19169
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5136795474091314e-06,
+      "loss": 0.5676,
+      "step": 19170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5126988255651475e-06,
+      "loss": 0.6309,
+      "step": 19171
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5117181792312446e-06,
+      "loss": 0.6056,
+      "step": 19172
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.510737608420926e-06,
+      "loss": 0.5871,
+      "step": 19173
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.509757113147681e-06,
+      "loss": 0.6443,
+      "step": 19174
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.508776693425003e-06,
+      "loss": 0.538,
+      "step": 19175
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.507796349266379e-06,
+      "loss": 0.5429,
+      "step": 19176
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.506816080685307e-06,
+      "loss": 0.6197,
+      "step": 19177
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.505835887695271e-06,
+      "loss": 0.6193,
+      "step": 19178
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.504855770309758e-06,
+      "loss": 0.5864,
+      "step": 19179
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.503875728542263e-06,
+      "loss": 0.5763,
+      "step": 19180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.502895762406265e-06,
+      "loss": 0.5825,
+      "step": 19181
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5019158719152535e-06,
+      "loss": 0.6321,
+      "step": 19182
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.500936057082705e-06,
+      "loss": 0.5506,
+      "step": 19183
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.499956317922114e-06,
+      "loss": 0.6172,
+      "step": 19184
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.498976654446956e-06,
+      "loss": 0.5713,
+      "step": 19185
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.497997066670712e-06,
+      "loss": 0.5646,
+      "step": 19186
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4970175546068595e-06,
+      "loss": 0.5424,
+      "step": 19187
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.496038118268885e-06,
+      "loss": 0.5855,
+      "step": 19188
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.495058757670262e-06,
+      "loss": 0.5581,
+      "step": 19189
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.494079472824464e-06,
+      "loss": 0.594,
+      "step": 19190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.493100263744974e-06,
+      "loss": 0.5527,
+      "step": 19191
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.492121130445264e-06,
+      "loss": 0.5484,
+      "step": 19192
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.491142072938806e-06,
+      "loss": 0.6415,
+      "step": 19193
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.490163091239069e-06,
+      "loss": 0.5803,
+      "step": 19194
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.489184185359534e-06,
+      "loss": 0.677,
+      "step": 19195
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.488205355313667e-06,
+      "loss": 0.5755,
+      "step": 19196
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.487226601114937e-06,
+      "loss": 0.562,
+      "step": 19197
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.48624792277681e-06,
+      "loss": 0.6496,
+      "step": 19198
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.485269320312758e-06,
+      "loss": 0.59,
+      "step": 19199
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.484290793736248e-06,
+      "loss": 0.5383,
+      "step": 19200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4833123430607385e-06,
+      "loss": 0.5884,
+      "step": 19201
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.482333968299702e-06,
+      "loss": 0.5715,
+      "step": 19202
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.481355669466599e-06,
+      "loss": 0.5769,
+      "step": 19203
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.480377446574891e-06,
+      "loss": 0.5348,
+      "step": 19204
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.479399299638035e-06,
+      "loss": 0.5132,
+      "step": 19205
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4784212286695005e-06,
+      "loss": 0.5744,
+      "step": 19206
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.47744323368274e-06,
+      "loss": 0.6095,
+      "step": 19207
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.476465314691214e-06,
+      "loss": 0.5334,
+      "step": 19208
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.475487471708375e-06,
+      "loss": 0.6427,
+      "step": 19209
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.474509704747686e-06,
+      "loss": 0.6378,
+      "step": 19210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.473532013822599e-06,
+      "loss": 0.6281,
+      "step": 19211
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.472554398946564e-06,
+      "loss": 0.6379,
+      "step": 19212
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.47157686013304e-06,
+      "loss": 0.6715,
+      "step": 19213
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4705993973954766e-06,
+      "loss": 0.6419,
+      "step": 19214
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.469622010747324e-06,
+      "loss": 0.6349,
+      "step": 19215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.468644700202028e-06,
+      "loss": 0.571,
+      "step": 19216
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.467667465773046e-06,
+      "loss": 0.5456,
+      "step": 19217
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.46669030747382e-06,
+      "loss": 0.6713,
+      "step": 19218
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.465713225317797e-06,
+      "loss": 0.5799,
+      "step": 19219
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.46473621931842e-06,
+      "loss": 0.6464,
+      "step": 19220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.463759289489139e-06,
+      "loss": 0.5894,
+      "step": 19221
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4627824358433956e-06,
+      "loss": 0.6738,
+      "step": 19222
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.461805658394628e-06,
+      "loss": 0.5068,
+      "step": 19223
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.460828957156283e-06,
+      "loss": 0.5384,
+      "step": 19224
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4598523321418e-06,
+      "loss": 0.5255,
+      "step": 19225
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.458875783364617e-06,
+      "loss": 0.56,
+      "step": 19226
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4578993108381674e-06,
+      "loss": 0.5998,
+      "step": 19227
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.456922914575897e-06,
+      "loss": 0.5757,
+      "step": 19228
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4559465945912385e-06,
+      "loss": 0.57,
+      "step": 19229
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.454970350897625e-06,
+      "loss": 0.5826,
+      "step": 19230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.453994183508489e-06,
+      "loss": 0.538,
+      "step": 19231
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.45301809243727e-06,
+      "loss": 0.5922,
+      "step": 19232
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.452042077697396e-06,
+      "loss": 0.5711,
+      "step": 19233
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.451066139302294e-06,
+      "loss": 0.6305,
+      "step": 19234
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.450090277265401e-06,
+      "loss": 0.6151,
+      "step": 19235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.449114491600143e-06,
+      "loss": 0.5425,
+      "step": 19236
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4481387823199465e-06,
+      "loss": 0.6312,
+      "step": 19237
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.447163149438235e-06,
+      "loss": 0.6392,
+      "step": 19238
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.446187592968441e-06,
+      "loss": 0.5888,
+      "step": 19239
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.445212112923988e-06,
+      "loss": 0.5134,
+      "step": 19240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.444236709318291e-06,
+      "loss": 0.5706,
+      "step": 19241
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.443261382164783e-06,
+      "loss": 0.61,
+      "step": 19242
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.442286131476882e-06,
+      "loss": 0.5591,
+      "step": 19243
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4413109572680054e-06,
+      "loss": 0.5584,
+      "step": 19244
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.440335859551572e-06,
+      "loss": 0.5944,
+      "step": 19245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.439360838341006e-06,
+      "loss": 0.5664,
+      "step": 19246
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.438385893649719e-06,
+      "loss": 0.6426,
+      "step": 19247
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.437411025491131e-06,
+      "loss": 0.6511,
+      "step": 19248
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.43643623387865e-06,
+      "loss": 0.6389,
+      "step": 19249
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4354615188257e-06,
+      "loss": 0.6391,
+      "step": 19250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.434486880345689e-06,
+      "loss": 0.5671,
+      "step": 19251
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.433512318452024e-06,
+      "loss": 0.6359,
+      "step": 19252
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.432537833158125e-06,
+      "loss": 0.5932,
+      "step": 19253
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4315634244773974e-06,
+      "loss": 0.5562,
+      "step": 19254
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.43058909242325e-06,
+      "loss": 0.6139,
+      "step": 19255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.429614837009086e-06,
+      "loss": 0.6593,
+      "step": 19256
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.428640658248321e-06,
+      "loss": 0.5729,
+      "step": 19257
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.427666556154355e-06,
+      "loss": 0.5916,
+      "step": 19258
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.426692530740594e-06,
+      "loss": 0.5848,
+      "step": 19259
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.425718582020438e-06,
+      "loss": 0.5858,
+      "step": 19260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.424744710007295e-06,
+      "loss": 0.5938,
+      "step": 19261
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.423770914714564e-06,
+      "loss": 0.5844,
+      "step": 19262
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.42279719615564e-06,
+      "loss": 0.6018,
+      "step": 19263
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.421823554343933e-06,
+      "loss": 0.5241,
+      "step": 19264
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.420849989292834e-06,
+      "loss": 0.5862,
+      "step": 19265
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.419876501015742e-06,
+      "loss": 0.6189,
+      "step": 19266
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.418903089526048e-06,
+      "loss": 0.5642,
+      "step": 19267
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4179297548371555e-06,
+      "loss": 0.5949,
+      "step": 19268
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.416956496962455e-06,
+      "loss": 0.6152,
+      "step": 19269
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.415983315915338e-06,
+      "loss": 0.5681,
+      "step": 19270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.415010211709193e-06,
+      "loss": 0.5418,
+      "step": 19271
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.414037184357419e-06,
+      "loss": 0.5867,
+      "step": 19272
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.413064233873403e-06,
+      "loss": 0.5886,
+      "step": 19273
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4120913602705276e-06,
+      "loss": 0.5542,
+      "step": 19274
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.411118563562188e-06,
+      "loss": 0.5693,
+      "step": 19275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.410145843761768e-06,
+      "loss": 0.5562,
+      "step": 19276
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4091732008826535e-06,
+      "loss": 0.5906,
+      "step": 19277
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4082006349382245e-06,
+      "loss": 0.5778,
+      "step": 19278
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.407228145941872e-06,
+      "loss": 0.5956,
+      "step": 19279
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.406255733906975e-06,
+      "loss": 0.6065,
+      "step": 19280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.405283398846913e-06,
+      "loss": 0.6046,
+      "step": 19281
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.404311140775067e-06,
+      "loss": 0.5928,
+      "step": 19282
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.403338959704817e-06,
+      "loss": 0.5423,
+      "step": 19283
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.402366855649537e-06,
+      "loss": 0.5912,
+      "step": 19284
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.401394828622612e-06,
+      "loss": 0.6067,
+      "step": 19285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.400422878637412e-06,
+      "loss": 0.5418,
+      "step": 19286
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.399451005707315e-06,
+      "loss": 0.5949,
+      "step": 19287
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3984792098456885e-06,
+      "loss": 0.6322,
+      "step": 19288
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.397507491065914e-06,
+      "loss": 0.5369,
+      "step": 19289
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.396535849381359e-06,
+      "loss": 0.5731,
+      "step": 19290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.395564284805395e-06,
+      "loss": 0.5758,
+      "step": 19291
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.394592797351387e-06,
+      "loss": 0.6095,
+      "step": 19292
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.393621387032712e-06,
+      "loss": 0.567,
+      "step": 19293
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.392650053862731e-06,
+      "loss": 0.587,
+      "step": 19294
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3916787978548094e-06,
+      "loss": 0.6178,
+      "step": 19295
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.390707619022319e-06,
+      "loss": 0.5791,
+      "step": 19296
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.389736517378621e-06,
+      "loss": 0.5291,
+      "step": 19297
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3887654929370785e-06,
+      "loss": 0.5621,
+      "step": 19298
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.387794545711048e-06,
+      "loss": 0.6123,
+      "step": 19299
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3868236757139e-06,
+      "loss": 0.5903,
+      "step": 19300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.385852882958992e-06,
+      "loss": 0.4895,
+      "step": 19301
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.38488216745968e-06,
+      "loss": 0.5544,
+      "step": 19302
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.38391152922932e-06,
+      "loss": 0.5593,
+      "step": 19303
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3829409682812755e-06,
+      "loss": 0.5927,
+      "step": 19304
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.381970484628899e-06,
+      "loss": 0.5796,
+      "step": 19305
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.381000078285541e-06,
+      "loss": 0.643,
+      "step": 19306
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.380029749264564e-06,
+      "loss": 0.6322,
+      "step": 19307
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.379059497579315e-06,
+      "loss": 0.5453,
+      "step": 19308
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.378089323243147e-06,
+      "loss": 0.6857,
+      "step": 19309
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.377119226269404e-06,
+      "loss": 0.5463,
+      "step": 19310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.376149206671446e-06,
+      "loss": 0.6414,
+      "step": 19311
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.375179264462618e-06,
+      "loss": 0.6502,
+      "step": 19312
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.374209399656264e-06,
+      "loss": 0.5494,
+      "step": 19313
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.373239612265729e-06,
+      "loss": 0.5887,
+      "step": 19314
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3722699023043634e-06,
+      "loss": 0.5882,
+      "step": 19315
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.37130026978551e-06,
+      "loss": 0.5232,
+      "step": 19316
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.370330714722507e-06,
+      "loss": 0.6151,
+      "step": 19317
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.369361237128704e-06,
+      "loss": 0.5499,
+      "step": 19318
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.368391837017437e-06,
+      "loss": 0.5764,
+      "step": 19319
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.367422514402048e-06,
+      "loss": 0.5725,
+      "step": 19320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.36645326929587e-06,
+      "loss": 0.6029,
+      "step": 19321
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.365484101712249e-06,
+      "loss": 0.581,
+      "step": 19322
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.364515011664517e-06,
+      "loss": 0.5746,
+      "step": 19323
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.36354599916601e-06,
+      "loss": 0.6339,
+      "step": 19324
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.362577064230059e-06,
+      "loss": 0.5758,
+      "step": 19325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.361608206870006e-06,
+      "loss": 0.6155,
+      "step": 19326
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.360639427099177e-06,
+      "loss": 0.6497,
+      "step": 19327
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.359670724930902e-06,
+      "loss": 0.6305,
+      "step": 19328
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.358702100378517e-06,
+      "loss": 0.5531,
+      "step": 19329
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.357733553455349e-06,
+      "loss": 0.5577,
+      "step": 19330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.356765084174723e-06,
+      "loss": 0.6644,
+      "step": 19331
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.355796692549966e-06,
+      "loss": 0.5782,
+      "step": 19332
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.354828378594409e-06,
+      "loss": 0.5628,
+      "step": 19333
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.353860142321374e-06,
+      "loss": 0.5385,
+      "step": 19334
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.352891983744182e-06,
+      "loss": 0.5918,
+      "step": 19335
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.351923902876161e-06,
+      "loss": 0.5832,
+      "step": 19336
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.350955899730629e-06,
+      "loss": 0.5918,
+      "step": 19337
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.349987974320908e-06,
+      "loss": 0.5505,
+      "step": 19338
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.349020126660314e-06,
+      "loss": 0.62,
+      "step": 19339
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3480523567621716e-06,
+      "loss": 0.5525,
+      "step": 19340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.347084664639795e-06,
+      "loss": 0.6281,
+      "step": 19341
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.346117050306501e-06,
+      "loss": 0.5867,
+      "step": 19342
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3451495137756e-06,
+      "loss": 0.6158,
+      "step": 19343
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.344182055060414e-06,
+      "loss": 0.5691,
+      "step": 19344
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.343214674174253e-06,
+      "loss": 0.5789,
+      "step": 19345
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.342247371130426e-06,
+      "loss": 0.5978,
+      "step": 19346
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3412801459422485e-06,
+      "loss": 0.5827,
+      "step": 19347
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.340312998623029e-06,
+      "loss": 0.6043,
+      "step": 19348
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3393459291860755e-06,
+      "loss": 0.5656,
+      "step": 19349
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.338378937644693e-06,
+      "loss": 0.663,
+      "step": 19350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.337412024012193e-06,
+      "loss": 0.6044,
+      "step": 19351
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.336445188301881e-06,
+      "loss": 0.6367,
+      "step": 19352
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.335478430527058e-06,
+      "loss": 0.6413,
+      "step": 19353
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.334511750701027e-06,
+      "loss": 0.5527,
+      "step": 19354
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.333545148837095e-06,
+      "loss": 0.5764,
+      "step": 19355
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.332578624948561e-06,
+      "loss": 0.5756,
+      "step": 19356
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.331612179048722e-06,
+      "loss": 0.6513,
+      "step": 19357
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.330645811150883e-06,
+      "loss": 0.5994,
+      "step": 19358
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.329679521268339e-06,
+      "loss": 0.6432,
+      "step": 19359
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.328713309414388e-06,
+      "loss": 0.512,
+      "step": 19360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.327747175602321e-06,
+      "loss": 0.5682,
+      "step": 19361
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.32678111984544e-06,
+      "loss": 0.6928,
+      "step": 19362
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.325815142157036e-06,
+      "loss": 0.6164,
+      "step": 19363
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.324849242550402e-06,
+      "loss": 0.6293,
+      "step": 19364
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.323883421038825e-06,
+      "loss": 0.5748,
+      "step": 19365
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.322917677635604e-06,
+      "loss": 0.5373,
+      "step": 19366
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.321952012354023e-06,
+      "loss": 0.5475,
+      "step": 19367
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.320986425207369e-06,
+      "loss": 0.5881,
+      "step": 19368
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.320020916208934e-06,
+      "loss": 0.6298,
+      "step": 19369
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3190554853720025e-06,
+      "loss": 0.5602,
+      "step": 19370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.31809013270986e-06,
+      "loss": 0.5483,
+      "step": 19371
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.317124858235785e-06,
+      "loss": 0.5543,
+      "step": 19372
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.31615966196307e-06,
+      "loss": 0.5968,
+      "step": 19373
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.315194543904993e-06,
+      "loss": 0.5655,
+      "step": 19374
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3142295040748316e-06,
+      "loss": 0.5728,
+      "step": 19375
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.313264542485867e-06,
+      "loss": 0.529,
+      "step": 19376
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.312299659151381e-06,
+      "loss": 0.6286,
+      "step": 19377
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.311334854084651e-06,
+      "loss": 0.6555,
+      "step": 19378
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.310370127298948e-06,
+      "loss": 0.6327,
+      "step": 19379
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.309405478807553e-06,
+      "loss": 0.5769,
+      "step": 19380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.308440908623741e-06,
+      "loss": 0.5496,
+      "step": 19381
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.307476416760783e-06,
+      "loss": 0.5799,
+      "step": 19382
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.306512003231946e-06,
+      "loss": 0.6427,
+      "step": 19383
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.305547668050511e-06,
+      "loss": 0.5609,
+      "step": 19384
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3045834112297445e-06,
+      "loss": 0.6009,
+      "step": 19385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.303619232782915e-06,
+      "loss": 0.6203,
+      "step": 19386
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3026551327232854e-06,
+      "loss": 0.6023,
+      "step": 19387
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.301691111064132e-06,
+      "loss": 0.6102,
+      "step": 19388
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.300727167818714e-06,
+      "loss": 0.5592,
+      "step": 19389
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.299763303000296e-06,
+      "loss": 0.5946,
+      "step": 19390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.298799516622147e-06,
+      "loss": 0.5453,
+      "step": 19391
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.297835808697526e-06,
+      "loss": 0.6276,
+      "step": 19392
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.296872179239695e-06,
+      "loss": 0.5966,
+      "step": 19393
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.29590862826191e-06,
+      "loss": 0.5896,
+      "step": 19394
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.294945155777438e-06,
+      "loss": 0.6426,
+      "step": 19395
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.293981761799533e-06,
+      "loss": 0.5606,
+      "step": 19396
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.29301844634145e-06,
+      "loss": 0.6127,
+      "step": 19397
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.29205520941645e-06,
+      "loss": 0.5135,
+      "step": 19398
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.291092051037787e-06,
+      "loss": 0.5411,
+      "step": 19399
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.290128971218713e-06,
+      "loss": 0.5896,
+      "step": 19400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.289165969972479e-06,
+      "loss": 0.5371,
+      "step": 19401
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.288203047312343e-06,
+      "loss": 0.5566,
+      "step": 19402
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.287240203251551e-06,
+      "loss": 0.5879,
+      "step": 19403
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2862774378033525e-06,
+      "loss": 0.6392,
+      "step": 19404
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.285314750980994e-06,
+      "loss": 0.5637,
+      "step": 19405
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.284352142797731e-06,
+      "loss": 0.5538,
+      "step": 19406
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2833896132668025e-06,
+      "loss": 0.5842,
+      "step": 19407
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.282427162401454e-06,
+      "loss": 0.6307,
+      "step": 19408
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.281464790214935e-06,
+      "loss": 0.6072,
+      "step": 19409
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.280502496720484e-06,
+      "loss": 0.5859,
+      "step": 19410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.279540281931346e-06,
+      "loss": 0.5883,
+      "step": 19411
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2785781458607555e-06,
+      "loss": 0.5349,
+      "step": 19412
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.277616088521961e-06,
+      "loss": 0.6256,
+      "step": 19413
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.276654109928197e-06,
+      "loss": 0.6028,
+      "step": 19414
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.275692210092701e-06,
+      "loss": 0.5698,
+      "step": 19415
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2747303890287076e-06,
+      "loss": 0.5925,
+      "step": 19416
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.273768646749457e-06,
+      "loss": 0.6057,
+      "step": 19417
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.272806983268182e-06,
+      "loss": 0.5747,
+      "step": 19418
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.271845398598111e-06,
+      "loss": 0.5413,
+      "step": 19419
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.270883892752485e-06,
+      "loss": 0.6181,
+      "step": 19420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.26992246574453e-06,
+      "loss": 0.5605,
+      "step": 19421
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.268961117587478e-06,
+      "loss": 0.6381,
+      "step": 19422
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.267999848294551e-06,
+      "loss": 0.6272,
+      "step": 19423
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.267038657878988e-06,
+      "loss": 0.6379,
+      "step": 19424
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.266077546354011e-06,
+      "loss": 0.5698,
+      "step": 19425
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2651165137328446e-06,
+      "loss": 0.5758,
+      "step": 19426
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.264155560028709e-06,
+      "loss": 0.5747,
+      "step": 19427
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.263194685254838e-06,
+      "loss": 0.5239,
+      "step": 19428
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.262233889424448e-06,
+      "loss": 0.6446,
+      "step": 19429
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2612731725507585e-06,
+      "loss": 0.5702,
+      "step": 19430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.260312534646996e-06,
+      "loss": 0.5509,
+      "step": 19431
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.259351975726377e-06,
+      "loss": 0.5247,
+      "step": 19432
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.258391495802118e-06,
+      "loss": 0.5891,
+      "step": 19433
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2574310948874335e-06,
+      "loss": 0.5621,
+      "step": 19434
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.256470772995547e-06,
+      "loss": 0.5454,
+      "step": 19435
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.255510530139669e-06,
+      "loss": 0.5809,
+      "step": 19436
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.254550366333013e-06,
+      "loss": 0.5087,
+      "step": 19437
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.25359028158879e-06,
+      "loss": 0.578,
+      "step": 19438
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.252630275920218e-06,
+      "loss": 0.5383,
+      "step": 19439
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.251670349340502e-06,
+      "loss": 0.5679,
+      "step": 19440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.25071050186285e-06,
+      "loss": 0.5662,
+      "step": 19441
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.249750733500477e-06,
+      "loss": 0.6028,
+      "step": 19442
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.248791044266587e-06,
+      "loss": 0.601,
+      "step": 19443
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.247831434174384e-06,
+      "loss": 0.6325,
+      "step": 19444
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.246871903237073e-06,
+      "loss": 0.5258,
+      "step": 19445
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.245912451467863e-06,
+      "loss": 0.5646,
+      "step": 19446
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.244953078879953e-06,
+      "loss": 0.507,
+      "step": 19447
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.243993785486545e-06,
+      "loss": 0.5676,
+      "step": 19448
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.243034571300838e-06,
+      "loss": 0.5401,
+      "step": 19449
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2420754363360364e-06,
+      "loss": 0.5758,
+      "step": 19450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.241116380605337e-06,
+      "loss": 0.633,
+      "step": 19451
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.240157404121932e-06,
+      "loss": 0.6085,
+      "step": 19452
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.239198506899026e-06,
+      "loss": 0.5574,
+      "step": 19453
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.23823968894981e-06,
+      "loss": 0.5533,
+      "step": 19454
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.237280950287479e-06,
+      "loss": 0.6182,
+      "step": 19455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.236322290925222e-06,
+      "loss": 0.6224,
+      "step": 19456
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2353637108762395e-06,
+      "loss": 0.614,
+      "step": 19457
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.234405210153716e-06,
+      "loss": 0.5688,
+      "step": 19458
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.23344678877084e-06,
+      "loss": 0.6057,
+      "step": 19459
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.232488446740805e-06,
+      "loss": 0.6338,
+      "step": 19460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2315301840768e-06,
+      "loss": 0.5957,
+      "step": 19461
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.230572000792006e-06,
+      "loss": 0.6356,
+      "step": 19462
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.229613896899608e-06,
+      "loss": 0.5554,
+      "step": 19463
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2286558724127954e-06,
+      "loss": 0.665,
+      "step": 19464
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.227697927344751e-06,
+      "loss": 0.5715,
+      "step": 19465
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.226740061708654e-06,
+      "loss": 0.5529,
+      "step": 19466
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.225782275517683e-06,
+      "loss": 0.563,
+      "step": 19467
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.224824568785025e-06,
+      "loss": 0.5756,
+      "step": 19468
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.223866941523857e-06,
+      "loss": 0.5818,
+      "step": 19469
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.222909393747349e-06,
+      "loss": 0.6053,
+      "step": 19470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.22195192546869e-06,
+      "loss": 0.5745,
+      "step": 19471
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.220994536701049e-06,
+      "loss": 0.6452,
+      "step": 19472
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.220037227457601e-06,
+      "loss": 0.5871,
+      "step": 19473
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.219079997751515e-06,
+      "loss": 0.6232,
+      "step": 19474
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.218122847595971e-06,
+      "loss": 0.5993,
+      "step": 19475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2171657770041385e-06,
+      "loss": 0.5017,
+      "step": 19476
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.216208785989186e-06,
+      "loss": 0.5645,
+      "step": 19477
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.215251874564278e-06,
+      "loss": 0.5484,
+      "step": 19478
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.214295042742591e-06,
+      "loss": 0.5816,
+      "step": 19479
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.213338290537289e-06,
+      "loss": 0.5618,
+      "step": 19480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.212381617961532e-06,
+      "loss": 0.5917,
+      "step": 19481
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.211425025028494e-06,
+      "loss": 0.6097,
+      "step": 19482
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.210468511751333e-06,
+      "loss": 0.5549,
+      "step": 19483
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.209512078143213e-06,
+      "loss": 0.6614,
+      "step": 19484
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.208555724217292e-06,
+      "loss": 0.5461,
+      "step": 19485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.207599449986734e-06,
+      "loss": 0.579,
+      "step": 19486
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2066432554647e-06,
+      "loss": 0.6128,
+      "step": 19487
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.205687140664343e-06,
+      "loss": 0.601,
+      "step": 19488
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.20473110559882e-06,
+      "loss": 0.6054,
+      "step": 19489
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.203775150281293e-06,
+      "loss": 0.559,
+      "step": 19490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.202819274724912e-06,
+      "loss": 0.5362,
+      "step": 19491
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.201863478942833e-06,
+      "loss": 0.6727,
+      "step": 19492
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.200907762948205e-06,
+      "loss": 0.5965,
+      "step": 19493
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.199952126754184e-06,
+      "loss": 0.5396,
+      "step": 19494
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.198996570373913e-06,
+      "loss": 0.5764,
+      "step": 19495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.198041093820551e-06,
+      "loss": 0.6261,
+      "step": 19496
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.197085697107241e-06,
+      "loss": 0.5757,
+      "step": 19497
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.196130380247131e-06,
+      "loss": 0.6128,
+      "step": 19498
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.195175143253363e-06,
+      "loss": 0.5978,
+      "step": 19499
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.194219986139089e-06,
+      "loss": 0.5586,
+      "step": 19500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.19326490891745e-06,
+      "loss": 0.6473,
+      "step": 19501
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1923099116015845e-06,
+      "loss": 0.6574,
+      "step": 19502
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.191354994204641e-06,
+      "loss": 0.5454,
+      "step": 19503
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.190400156739755e-06,
+      "loss": 0.4891,
+      "step": 19504
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.18944539922007e-06,
+      "loss": 0.5959,
+      "step": 19505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.188490721658717e-06,
+      "loss": 0.5774,
+      "step": 19506
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.187536124068843e-06,
+      "loss": 0.5318,
+      "step": 19507
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.186581606463578e-06,
+      "loss": 0.5935,
+      "step": 19508
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.185627168856058e-06,
+      "loss": 0.609,
+      "step": 19509
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1846728112594145e-06,
+      "loss": 0.604,
+      "step": 19510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.183718533686786e-06,
+      "loss": 0.5561,
+      "step": 19511
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1827643361513015e-06,
+      "loss": 0.5954,
+      "step": 19512
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.181810218666087e-06,
+      "loss": 0.5593,
+      "step": 19513
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.18085618124428e-06,
+      "loss": 0.5754,
+      "step": 19514
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.179902223899005e-06,
+      "loss": 0.5859,
+      "step": 19515
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.17894834664339e-06,
+      "loss": 0.6061,
+      "step": 19516
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.177994549490556e-06,
+      "loss": 0.5671,
+      "step": 19517
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.177040832453637e-06,
+      "loss": 0.6155,
+      "step": 19518
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.176087195545753e-06,
+      "loss": 0.5488,
+      "step": 19519
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.175133638780025e-06,
+      "loss": 0.5852,
+      "step": 19520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.174180162169575e-06,
+      "loss": 0.5126,
+      "step": 19521
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.173226765727527e-06,
+      "loss": 0.5391,
+      "step": 19522
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.172273449466999e-06,
+      "loss": 0.6035,
+      "step": 19523
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.171320213401105e-06,
+      "loss": 0.6079,
+      "step": 19524
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.17036705754297e-06,
+      "loss": 0.5624,
+      "step": 19525
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.169413981905707e-06,
+      "loss": 0.6296,
+      "step": 19526
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.168460986502431e-06,
+      "loss": 0.6193,
+      "step": 19527
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.167508071346252e-06,
+      "loss": 0.6643,
+      "step": 19528
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.16655523645029e-06,
+      "loss": 0.5454,
+      "step": 19529
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.165602481827654e-06,
+      "loss": 0.5773,
+      "step": 19530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.164649807491454e-06,
+      "loss": 0.6189,
+      "step": 19531
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.163697213454798e-06,
+      "loss": 0.5686,
+      "step": 19532
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1627446997307984e-06,
+      "loss": 0.617,
+      "step": 19533
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1617922663325615e-06,
+      "loss": 0.6204,
+      "step": 19534
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.160839913273189e-06,
+      "loss": 0.5581,
+      "step": 19535
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.159887640565794e-06,
+      "loss": 0.6632,
+      "step": 19536
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.158935448223477e-06,
+      "loss": 0.5713,
+      "step": 19537
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.15798333625934e-06,
+      "loss": 0.6083,
+      "step": 19538
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1570313046864815e-06,
+      "loss": 0.6268,
+      "step": 19539
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.156079353518011e-06,
+      "loss": 0.6017,
+      "step": 19540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.155127482767024e-06,
+      "loss": 0.6034,
+      "step": 19541
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.154175692446617e-06,
+      "loss": 0.5641,
+      "step": 19542
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.153223982569886e-06,
+      "loss": 0.6055,
+      "step": 19543
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.152272353149934e-06,
+      "loss": 0.5181,
+      "step": 19544
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.151320804199853e-06,
+      "loss": 0.6421,
+      "step": 19545
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.150369335732734e-06,
+      "loss": 0.5832,
+      "step": 19546
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1494179477616745e-06,
+      "loss": 0.5424,
+      "step": 19547
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.148466640299766e-06,
+      "loss": 0.6061,
+      "step": 19548
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.147515413360098e-06,
+      "loss": 0.6279,
+      "step": 19549
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.146564266955757e-06,
+      "loss": 0.5667,
+      "step": 19550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.145613201099839e-06,
+      "loss": 0.5707,
+      "step": 19551
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.144662215805426e-06,
+      "loss": 0.5801,
+      "step": 19552
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.143711311085606e-06,
+      "loss": 0.5987,
+      "step": 19553
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.142760486953462e-06,
+      "loss": 0.5164,
+      "step": 19554
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.141809743422083e-06,
+      "loss": 0.5721,
+      "step": 19555
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.140859080504549e-06,
+      "loss": 0.5674,
+      "step": 19556
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.139908498213939e-06,
+      "loss": 0.5525,
+      "step": 19557
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.138957996563342e-06,
+      "loss": 0.634,
+      "step": 19558
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1380075755658325e-06,
+      "loss": 0.6364,
+      "step": 19559
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.137057235234488e-06,
+      "loss": 0.6245,
+      "step": 19560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.136106975582386e-06,
+      "loss": 0.6004,
+      "step": 19561
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.135156796622606e-06,
+      "loss": 0.5994,
+      "step": 19562
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.134206698368223e-06,
+      "loss": 0.5777,
+      "step": 19563
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.133256680832306e-06,
+      "loss": 0.6162,
+      "step": 19564
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.132306744027936e-06,
+      "loss": 0.5984,
+      "step": 19565
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.131356887968181e-06,
+      "loss": 0.5613,
+      "step": 19566
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.130407112666112e-06,
+      "loss": 0.626,
+      "step": 19567
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.129457418134795e-06,
+      "loss": 0.5633,
+      "step": 19568
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.128507804387305e-06,
+      "loss": 0.5626,
+      "step": 19569
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.127558271436708e-06,
+      "loss": 0.561,
+      "step": 19570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.126608819296069e-06,
+      "loss": 0.6005,
+      "step": 19571
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1256594479784485e-06,
+      "loss": 0.5982,
+      "step": 19572
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.12471015749692e-06,
+      "loss": 0.5356,
+      "step": 19573
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.123760947864542e-06,
+      "loss": 0.6282,
+      "step": 19574
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.122811819094373e-06,
+      "loss": 0.5738,
+      "step": 19575
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.121862771199482e-06,
+      "loss": 0.5466,
+      "step": 19576
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.120913804192924e-06,
+      "loss": 0.5676,
+      "step": 19577
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.119964918087758e-06,
+      "loss": 0.5494,
+      "step": 19578
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.119016112897037e-06,
+      "loss": 0.5685,
+      "step": 19579
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.118067388633827e-06,
+      "loss": 0.6866,
+      "step": 19580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.117118745311176e-06,
+      "loss": 0.5495,
+      "step": 19581
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1161701829421405e-06,
+      "loss": 0.5677,
+      "step": 19582
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.115221701539771e-06,
+      "loss": 0.6073,
+      "step": 19583
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1142733011171245e-06,
+      "loss": 0.5767,
+      "step": 19584
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.113324981687249e-06,
+      "loss": 0.6763,
+      "step": 19585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.11237674326319e-06,
+      "loss": 0.6474,
+      "step": 19586
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.111428585858005e-06,
+      "loss": 0.5566,
+      "step": 19587
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.110480509484737e-06,
+      "loss": 0.5946,
+      "step": 19588
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.109532514156431e-06,
+      "loss": 0.5692,
+      "step": 19589
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.108584599886131e-06,
+      "loss": 0.5475,
+      "step": 19590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.107636766686887e-06,
+      "loss": 0.5831,
+      "step": 19591
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.106689014571737e-06,
+      "loss": 0.5816,
+      "step": 19592
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.105741343553727e-06,
+      "loss": 0.561,
+      "step": 19593
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1047937536458895e-06,
+      "loss": 0.5466,
+      "step": 19594
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.103846244861273e-06,
+      "loss": 0.5999,
+      "step": 19595
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.102898817212916e-06,
+      "loss": 0.5686,
+      "step": 19596
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.101951470713847e-06,
+      "loss": 0.5974,
+      "step": 19597
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.101004205377113e-06,
+      "loss": 0.5988,
+      "step": 19598
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.100057021215743e-06,
+      "loss": 0.5839,
+      "step": 19599
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.099109918242774e-06,
+      "loss": 0.5593,
+      "step": 19600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.098162896471234e-06,
+      "loss": 0.5922,
+      "step": 19601
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.097215955914161e-06,
+      "loss": 0.5887,
+      "step": 19602
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.096269096584583e-06,
+      "loss": 0.5474,
+      "step": 19603
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.095322318495531e-06,
+      "loss": 0.5486,
+      "step": 19604
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.094375621660027e-06,
+      "loss": 0.5745,
+      "step": 19605
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.093429006091108e-06,
+      "loss": 0.6317,
+      "step": 19606
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.092482471801795e-06,
+      "loss": 0.6078,
+      "step": 19607
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.091536018805111e-06,
+      "loss": 0.6386,
+      "step": 19608
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.090589647114087e-06,
+      "loss": 0.5787,
+      "step": 19609
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0896433567417424e-06,
+      "loss": 0.5947,
+      "step": 19610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.088697147701097e-06,
+      "loss": 0.588,
+      "step": 19611
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.087751020005171e-06,
+      "loss": 0.5257,
+      "step": 19612
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.086804973666988e-06,
+      "loss": 0.5899,
+      "step": 19613
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.085859008699565e-06,
+      "loss": 0.5416,
+      "step": 19614
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.084913125115919e-06,
+      "loss": 0.5987,
+      "step": 19615
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.083967322929062e-06,
+      "loss": 0.6128,
+      "step": 19616
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.083021602152016e-06,
+      "loss": 0.5827,
+      "step": 19617
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0820759627977925e-06,
+      "loss": 0.5995,
+      "step": 19618
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0811304048794e-06,
+      "loss": 0.571,
+      "step": 19619
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.080184928409857e-06,
+      "loss": 0.6229,
+      "step": 19620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.079239533402172e-06,
+      "loss": 0.5722,
+      "step": 19621
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.078294219869353e-06,
+      "loss": 0.6097,
+      "step": 19622
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.077348987824404e-06,
+      "loss": 0.5203,
+      "step": 19623
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.076403837280342e-06,
+      "loss": 0.6144,
+      "step": 19624
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0754587682501675e-06,
+      "loss": 0.6367,
+      "step": 19625
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.074513780746884e-06,
+      "loss": 0.5948,
+      "step": 19626
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.073568874783499e-06,
+      "loss": 0.5615,
+      "step": 19627
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.072624050373015e-06,
+      "loss": 0.5496,
+      "step": 19628
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0716793075284324e-06,
+      "loss": 0.6189,
+      "step": 19629
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0707346462627474e-06,
+      "loss": 0.5613,
+      "step": 19630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.069790066588966e-06,
+      "loss": 0.5366,
+      "step": 19631
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.068845568520086e-06,
+      "loss": 0.5991,
+      "step": 19632
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.067901152069102e-06,
+      "loss": 0.5422,
+      "step": 19633
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.066956817249007e-06,
+      "loss": 0.5805,
+      "step": 19634
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0660125640728024e-06,
+      "loss": 0.6113,
+      "step": 19635
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.065068392553479e-06,
+      "loss": 0.5348,
+      "step": 19636
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.064124302704027e-06,
+      "loss": 0.5607,
+      "step": 19637
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0631802945374435e-06,
+      "loss": 0.5527,
+      "step": 19638
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0622363680667145e-06,
+      "loss": 0.5875,
+      "step": 19639
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0612925233048316e-06,
+      "loss": 0.5615,
+      "step": 19640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.060348760264777e-06,
+      "loss": 0.5755,
+      "step": 19641
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.059405078959547e-06,
+      "loss": 0.5725,
+      "step": 19642
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.058461479402124e-06,
+      "loss": 0.5435,
+      "step": 19643
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0575179616054904e-06,
+      "loss": 0.5425,
+      "step": 19644
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0565745255826274e-06,
+      "loss": 0.6232,
+      "step": 19645
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.055631171346526e-06,
+      "loss": 0.6054,
+      "step": 19646
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.054687898910161e-06,
+      "loss": 0.579,
+      "step": 19647
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.053744708286512e-06,
+      "loss": 0.5928,
+      "step": 19648
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.052801599488565e-06,
+      "loss": 0.5332,
+      "step": 19649
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0518585725292924e-06,
+      "loss": 0.5454,
+      "step": 19650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.050915627421673e-06,
+      "loss": 0.5713,
+      "step": 19651
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.049972764178679e-06,
+      "loss": 0.6028,
+      "step": 19652
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.049029982813291e-06,
+      "loss": 0.6069,
+      "step": 19653
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0480872833384786e-06,
+      "loss": 0.5698,
+      "step": 19654
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.047144665767216e-06,
+      "loss": 0.5853,
+      "step": 19655
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.046202130112468e-06,
+      "loss": 0.5733,
+      "step": 19656
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.045259676387217e-06,
+      "loss": 0.5614,
+      "step": 19657
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.044317304604423e-06,
+      "loss": 0.6417,
+      "step": 19658
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.043375014777052e-06,
+      "loss": 0.5594,
+      "step": 19659
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.042432806918078e-06,
+      "loss": 0.5973,
+      "step": 19660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.041490681040465e-06,
+      "loss": 0.5712,
+      "step": 19661
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.040548637157175e-06,
+      "loss": 0.647,
+      "step": 19662
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.039606675281169e-06,
+      "loss": 0.4796,
+      "step": 19663
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.038664795425417e-06,
+      "loss": 0.5484,
+      "step": 19664
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.037722997602873e-06,
+      "loss": 0.684,
+      "step": 19665
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.036781281826502e-06,
+      "loss": 0.5768,
+      "step": 19666
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.035839648109254e-06,
+      "loss": 0.5904,
+      "step": 19667
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0348980964640995e-06,
+      "loss": 0.573,
+      "step": 19668
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0339566269039864e-06,
+      "loss": 0.578,
+      "step": 19669
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.03301523944187e-06,
+      "loss": 0.5061,
+      "step": 19670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.03207393409071e-06,
+      "loss": 0.6199,
+      "step": 19671
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.031132710863457e-06,
+      "loss": 0.5771,
+      "step": 19672
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.030191569773062e-06,
+      "loss": 0.5471,
+      "step": 19673
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.029250510832473e-06,
+      "loss": 0.5525,
+      "step": 19674
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.028309534054647e-06,
+      "loss": 0.5403,
+      "step": 19675
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.027368639452529e-06,
+      "loss": 0.611,
+      "step": 19676
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.026427827039066e-06,
+      "loss": 0.6076,
+      "step": 19677
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.025487096827203e-06,
+      "loss": 0.5997,
+      "step": 19678
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.024546448829889e-06,
+      "loss": 0.6071,
+      "step": 19679
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0236058830600675e-06,
+      "loss": 0.5816,
+      "step": 19680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.022665399530676e-06,
+      "loss": 0.5647,
+      "step": 19681
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.021724998254665e-06,
+      "loss": 0.5818,
+      "step": 19682
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.020784679244971e-06,
+      "loss": 0.6149,
+      "step": 19683
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.019844442514534e-06,
+      "loss": 0.5559,
+      "step": 19684
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.018904288076287e-06,
+      "loss": 0.6444,
+      "step": 19685
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.017964215943178e-06,
+      "loss": 0.5496,
+      "step": 19686
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.017024226128136e-06,
+      "loss": 0.5615,
+      "step": 19687
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0160843186440945e-06,
+      "loss": 0.6378,
+      "step": 19688
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.015144493503994e-06,
+      "loss": 0.5741,
+      "step": 19689
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.014204750720765e-06,
+      "loss": 0.5478,
+      "step": 19690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.013265090307338e-06,
+      "loss": 0.529,
+      "step": 19691
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.01232551227664e-06,
+      "loss": 0.6225,
+      "step": 19692
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0113860166416075e-06,
+      "loss": 0.6025,
+      "step": 19693
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.010446603415166e-06,
+      "loss": 0.5258,
+      "step": 19694
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.009507272610241e-06,
+      "loss": 0.5645,
+      "step": 19695
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.008568024239757e-06,
+      "loss": 0.6282,
+      "step": 19696
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.007628858316645e-06,
+      "loss": 0.6195,
+      "step": 19697
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.006689774853826e-06,
+      "loss": 0.5506,
+      "step": 19698
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.005750773864217e-06,
+      "loss": 0.6075,
+      "step": 19699
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0048118553607485e-06,
+      "loss": 0.5093,
+      "step": 19700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.003873019356337e-06,
+      "loss": 0.6073,
+      "step": 19701
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0029342658639e-06,
+      "loss": 0.5531,
+      "step": 19702
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.001995594896358e-06,
+      "loss": 0.5928,
+      "step": 19703
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.001057006466628e-06,
+      "loss": 0.5534,
+      "step": 19704
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.000118500587623e-06,
+      "loss": 0.6228,
+      "step": 19705
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.999180077272257e-06,
+      "loss": 0.5395,
+      "step": 19706
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.998241736533449e-06,
+      "loss": 0.5726,
+      "step": 19707
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.997303478384108e-06,
+      "loss": 0.5843,
+      "step": 19708
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9963653028371465e-06,
+      "loss": 0.5611,
+      "step": 19709
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9954272099054696e-06,
+      "loss": 0.6036,
+      "step": 19710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.994489199601993e-06,
+      "loss": 0.5896,
+      "step": 19711
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.993551271939623e-06,
+      "loss": 0.5327,
+      "step": 19712
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.992613426931261e-06,
+      "loss": 0.6339,
+      "step": 19713
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.991675664589821e-06,
+      "loss": 0.5987,
+      "step": 19714
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.990737984928204e-06,
+      "loss": 0.5586,
+      "step": 19715
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.989800387959311e-06,
+      "loss": 0.5126,
+      "step": 19716
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.988862873696042e-06,
+      "loss": 0.5724,
+      "step": 19717
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.987925442151307e-06,
+      "loss": 0.6025,
+      "step": 19718
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.986988093337999e-06,
+      "loss": 0.5479,
+      "step": 19719
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.986050827269015e-06,
+      "loss": 0.5571,
+      "step": 19720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9851136439572605e-06,
+      "loss": 0.5663,
+      "step": 19721
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.984176543415626e-06,
+      "loss": 0.5327,
+      "step": 19722
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.983239525657009e-06,
+      "loss": 0.5727,
+      "step": 19723
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.982302590694299e-06,
+      "loss": 0.5711,
+      "step": 19724
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9813657385403966e-06,
+      "loss": 0.5936,
+      "step": 19725
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.980428969208191e-06,
+      "loss": 0.5297,
+      "step": 19726
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.979492282710572e-06,
+      "loss": 0.5394,
+      "step": 19727
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.978555679060424e-06,
+      "loss": 0.634,
+      "step": 19728
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.977619158270646e-06,
+      "loss": 0.6149,
+      "step": 19729
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9766827203541195e-06,
+      "loss": 0.5645,
+      "step": 19730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.975746365323727e-06,
+      "loss": 0.5677,
+      "step": 19731
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.974810093192363e-06,
+      "loss": 0.5853,
+      "step": 19732
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.973873903972905e-06,
+      "loss": 0.6083,
+      "step": 19733
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.972937797678238e-06,
+      "loss": 0.6016,
+      "step": 19734
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.972001774321238e-06,
+      "loss": 0.5437,
+      "step": 19735
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.971065833914796e-06,
+      "loss": 0.5965,
+      "step": 19736
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.970129976471784e-06,
+      "loss": 0.5649,
+      "step": 19737
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.969194202005081e-06,
+      "loss": 0.5615,
+      "step": 19738
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.968258510527563e-06,
+      "loss": 0.5693,
+      "step": 19739
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.967322902052111e-06,
+      "loss": 0.5559,
+      "step": 19740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.966387376591596e-06,
+      "loss": 0.5943,
+      "step": 19741
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.965451934158889e-06,
+      "loss": 0.608,
+      "step": 19742
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.964516574766869e-06,
+      "loss": 0.6539,
+      "step": 19743
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.963581298428405e-06,
+      "loss": 0.6073,
+      "step": 19744
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9626461051563655e-06,
+      "loss": 0.576,
+      "step": 19745
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.961710994963618e-06,
+      "loss": 0.635,
+      "step": 19746
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.960775967863037e-06,
+      "loss": 0.6013,
+      "step": 19747
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.959841023867484e-06,
+      "loss": 0.6074,
+      "step": 19748
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.958906162989825e-06,
+      "loss": 0.5458,
+      "step": 19749
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.957971385242924e-06,
+      "loss": 0.6089,
+      "step": 19750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.957036690639647e-06,
+      "loss": 0.5736,
+      "step": 19751
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.956102079192856e-06,
+      "loss": 0.5738,
+      "step": 19752
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.955167550915409e-06,
+      "loss": 0.6011,
+      "step": 19753
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.954233105820169e-06,
+      "loss": 0.5549,
+      "step": 19754
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.953298743919996e-06,
+      "loss": 0.6071,
+      "step": 19755
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.952364465227742e-06,
+      "loss": 0.591,
+      "step": 19756
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.951430269756267e-06,
+      "loss": 0.5534,
+      "step": 19757
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.950496157518427e-06,
+      "loss": 0.6776,
+      "step": 19758
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.949562128527078e-06,
+      "loss": 0.6503,
+      "step": 19759
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.948628182795069e-06,
+      "loss": 0.5202,
+      "step": 19760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.947694320335251e-06,
+      "loss": 0.6203,
+      "step": 19761
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.946760541160479e-06,
+      "loss": 0.5915,
+      "step": 19762
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.945826845283601e-06,
+      "loss": 0.5613,
+      "step": 19763
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9448932327174625e-06,
+      "loss": 0.6036,
+      "step": 19764
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.943959703474917e-06,
+      "loss": 0.5747,
+      "step": 19765
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.943026257568807e-06,
+      "loss": 0.5749,
+      "step": 19766
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.942092895011979e-06,
+      "loss": 0.584,
+      "step": 19767
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.94115961581727e-06,
+      "loss": 0.6255,
+      "step": 19768
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.940226419997534e-06,
+      "loss": 0.5428,
+      "step": 19769
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9392933075656055e-06,
+      "loss": 0.5773,
+      "step": 19770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.938360278534327e-06,
+      "loss": 0.6441,
+      "step": 19771
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.937427332916535e-06,
+      "loss": 0.5775,
+      "step": 19772
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.936494470725072e-06,
+      "loss": 0.5722,
+      "step": 19773
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.935561691972773e-06,
+      "loss": 0.5902,
+      "step": 19774
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9346289966724715e-06,
+      "loss": 0.5856,
+      "step": 19775
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.933696384837007e-06,
+      "loss": 0.5651,
+      "step": 19776
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9327638564792115e-06,
+      "loss": 0.6067,
+      "step": 19777
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.931831411611915e-06,
+      "loss": 0.5511,
+      "step": 19778
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.930899050247948e-06,
+      "loss": 0.573,
+      "step": 19779
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.929966772400146e-06,
+      "loss": 0.576,
+      "step": 19780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9290345780813345e-06,
+      "loss": 0.5122,
+      "step": 19781
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9281024673043375e-06,
+      "loss": 0.6219,
+      "step": 19782
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.927170440081991e-06,
+      "loss": 0.5628,
+      "step": 19783
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9262384964271135e-06,
+      "loss": 0.605,
+      "step": 19784
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.925306636352533e-06,
+      "loss": 0.5918,
+      "step": 19785
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.924374859871066e-06,
+      "loss": 0.5739,
+      "step": 19786
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.923443166995543e-06,
+      "loss": 0.4891,
+      "step": 19787
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.922511557738782e-06,
+      "loss": 0.5334,
+      "step": 19788
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.921580032113602e-06,
+      "loss": 0.5625,
+      "step": 19789
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.920648590132819e-06,
+      "loss": 0.6021,
+      "step": 19790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.919717231809255e-06,
+      "loss": 0.6434,
+      "step": 19791
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.918785957155726e-06,
+      "loss": 0.5654,
+      "step": 19792
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.917854766185041e-06,
+      "loss": 0.5838,
+      "step": 19793
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.916923658910022e-06,
+      "loss": 0.5672,
+      "step": 19794
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9159926353434786e-06,
+      "loss": 0.572,
+      "step": 19795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.915061695498224e-06,
+      "loss": 0.6305,
+      "step": 19796
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.914130839387061e-06,
+      "loss": 0.645,
+      "step": 19797
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.913200067022811e-06,
+      "loss": 0.5815,
+      "step": 19798
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.912269378418275e-06,
+      "loss": 0.5517,
+      "step": 19799
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.911338773586263e-06,
+      "loss": 0.5881,
+      "step": 19800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.910408252539575e-06,
+      "loss": 0.5445,
+      "step": 19801
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.909477815291023e-06,
+      "loss": 0.5625,
+      "step": 19802
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9085474618534095e-06,
+      "loss": 0.5999,
+      "step": 19803
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.90761719223953e-06,
+      "loss": 0.6248,
+      "step": 19804
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.906687006462196e-06,
+      "loss": 0.604,
+      "step": 19805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.905756904534203e-06,
+      "loss": 0.6022,
+      "step": 19806
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.904826886468349e-06,
+      "loss": 0.5631,
+      "step": 19807
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9038969522774295e-06,
+      "loss": 0.5905,
+      "step": 19808
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.902967101974248e-06,
+      "loss": 0.6089,
+      "step": 19809
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.902037335571597e-06,
+      "loss": 0.5451,
+      "step": 19810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.90110765308227e-06,
+      "loss": 0.578,
+      "step": 19811
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.900178054519057e-06,
+      "loss": 0.6486,
+      "step": 19812
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.899248539894756e-06,
+      "loss": 0.5292,
+      "step": 19813
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.898319109222158e-06,
+      "loss": 0.5263,
+      "step": 19814
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.897389762514046e-06,
+      "loss": 0.5982,
+      "step": 19815
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8964604997832165e-06,
+      "loss": 0.5632,
+      "step": 19816
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.895531321042454e-06,
+      "loss": 0.5186,
+      "step": 19817
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.894602226304543e-06,
+      "loss": 0.5843,
+      "step": 19818
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.893673215582269e-06,
+      "loss": 0.6228,
+      "step": 19819
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.892744288888418e-06,
+      "loss": 0.5439,
+      "step": 19820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.891815446235773e-06,
+      "loss": 0.544,
+      "step": 19821
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8908866876371144e-06,
+      "loss": 0.6297,
+      "step": 19822
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.889958013105219e-06,
+      "loss": 0.5959,
+      "step": 19823
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.889029422652875e-06,
+      "loss": 0.606,
+      "step": 19824
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.888100916292855e-06,
+      "loss": 0.5812,
+      "step": 19825
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.887172494037933e-06,
+      "loss": 0.569,
+      "step": 19826
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.886244155900893e-06,
+      "loss": 0.5925,
+      "step": 19827
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.885315901894504e-06,
+      "loss": 0.5211,
+      "step": 19828
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.884387732031542e-06,
+      "loss": 0.5814,
+      "step": 19829
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.883459646324775e-06,
+      "loss": 0.6518,
+      "step": 19830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.882531644786981e-06,
+      "loss": 0.5661,
+      "step": 19831
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.881603727430927e-06,
+      "loss": 0.5367,
+      "step": 19832
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.880675894269381e-06,
+      "loss": 0.5873,
+      "step": 19833
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.879748145315109e-06,
+      "loss": 0.5614,
+      "step": 19834
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.878820480580884e-06,
+      "loss": 0.6504,
+      "step": 19835
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.877892900079466e-06,
+      "loss": 0.6252,
+      "step": 19836
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8769654038236195e-06,
+      "loss": 0.5861,
+      "step": 19837
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.876037991826111e-06,
+      "loss": 0.6236,
+      "step": 19838
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.875110664099703e-06,
+      "loss": 0.5603,
+      "step": 19839
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.874183420657152e-06,
+      "loss": 0.5981,
+      "step": 19840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.873256261511217e-06,
+      "loss": 0.6011,
+      "step": 19841
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8723291866746625e-06,
+      "loss": 0.5794,
+      "step": 19842
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.871402196160243e-06,
+      "loss": 0.571,
+      "step": 19843
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.870475289980711e-06,
+      "loss": 0.5754,
+      "step": 19844
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.869548468148828e-06,
+      "loss": 0.5584,
+      "step": 19845
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8686217306773455e-06,
+      "loss": 0.6228,
+      "step": 19846
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.867695077579015e-06,
+      "loss": 0.5346,
+      "step": 19847
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.866768508866584e-06,
+      "loss": 0.6344,
+      "step": 19848
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.865842024552813e-06,
+      "loss": 0.603,
+      "step": 19849
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.864915624650444e-06,
+      "loss": 0.5976,
+      "step": 19850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.863989309172227e-06,
+      "loss": 0.5688,
+      "step": 19851
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.863063078130905e-06,
+      "loss": 0.6292,
+      "step": 19852
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.862136931539232e-06,
+      "loss": 0.6233,
+      "step": 19853
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.861210869409947e-06,
+      "loss": 0.5631,
+      "step": 19854
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.860284891755792e-06,
+      "loss": 0.5556,
+      "step": 19855
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.859358998589514e-06,
+      "loss": 0.5664,
+      "step": 19856
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.858433189923852e-06,
+      "loss": 0.5083,
+      "step": 19857
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.857507465771547e-06,
+      "loss": 0.6275,
+      "step": 19858
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.856581826145333e-06,
+      "loss": 0.6179,
+      "step": 19859
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.855656271057953e-06,
+      "loss": 0.4966,
+      "step": 19860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.854730800522143e-06,
+      "loss": 0.5851,
+      "step": 19861
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.853805414550636e-06,
+      "loss": 0.6262,
+      "step": 19862
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.852880113156163e-06,
+      "loss": 0.5522,
+      "step": 19863
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.851954896351467e-06,
+      "loss": 0.5821,
+      "step": 19864
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.851029764149271e-06,
+      "loss": 0.5891,
+      "step": 19865
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.850104716562307e-06,
+      "loss": 0.6643,
+      "step": 19866
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.849179753603309e-06,
+      "loss": 0.601,
+      "step": 19867
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.848254875285e-06,
+      "loss": 0.537,
+      "step": 19868
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.847330081620111e-06,
+      "loss": 0.5892,
+      "step": 19869
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.846405372621363e-06,
+      "loss": 0.568,
+      "step": 19870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.845480748301487e-06,
+      "loss": 0.6187,
+      "step": 19871
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.844556208673204e-06,
+      "loss": 0.6617,
+      "step": 19872
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8436317537492364e-06,
+      "loss": 0.6104,
+      "step": 19873
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.842707383542301e-06,
+      "loss": 0.5678,
+      "step": 19874
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.841783098065127e-06,
+      "loss": 0.6243,
+      "step": 19875
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.840858897330429e-06,
+      "loss": 0.6106,
+      "step": 19876
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.839934781350919e-06,
+      "loss": 0.5764,
+      "step": 19877
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.839010750139324e-06,
+      "loss": 0.5669,
+      "step": 19878
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.838086803708354e-06,
+      "loss": 0.585,
+      "step": 19879
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.837162942070725e-06,
+      "loss": 0.5842,
+      "step": 19880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.836239165239144e-06,
+      "loss": 0.5863,
+      "step": 19881
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.835315473226332e-06,
+      "loss": 0.5945,
+      "step": 19882
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.834391866044996e-06,
+      "loss": 0.5976,
+      "step": 19883
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.833468343707846e-06,
+      "loss": 0.5607,
+      "step": 19884
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.832544906227585e-06,
+      "loss": 0.5914,
+      "step": 19885
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.831621553616931e-06,
+      "loss": 0.5819,
+      "step": 19886
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.830698285888583e-06,
+      "loss": 0.6216,
+      "step": 19887
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8297751030552445e-06,
+      "loss": 0.6248,
+      "step": 19888
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.828852005129625e-06,
+      "loss": 0.601,
+      "step": 19889
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.827928992124425e-06,
+      "loss": 0.5696,
+      "step": 19890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.827006064052346e-06,
+      "loss": 0.5318,
+      "step": 19891
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8260832209260825e-06,
+      "loss": 0.637,
+      "step": 19892
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.825160462758343e-06,
+      "loss": 0.5998,
+      "step": 19893
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.824237789561821e-06,
+      "loss": 0.5269,
+      "step": 19894
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.823315201349214e-06,
+      "loss": 0.5555,
+      "step": 19895
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.822392698133214e-06,
+      "loss": 0.6536,
+      "step": 19896
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.821470279926521e-06,
+      "loss": 0.5808,
+      "step": 19897
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.820547946741826e-06,
+      "loss": 0.5287,
+      "step": 19898
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.819625698591818e-06,
+      "loss": 0.5942,
+      "step": 19899
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.818703535489195e-06,
+      "loss": 0.6487,
+      "step": 19900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.817781457446642e-06,
+      "loss": 0.6226,
+      "step": 19901
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8168594644768475e-06,
+      "loss": 0.6553,
+      "step": 19902
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8159375565924975e-06,
+      "loss": 0.5937,
+      "step": 19903
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.815015733806283e-06,
+      "loss": 0.5661,
+      "step": 19904
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8140939961308876e-06,
+      "loss": 0.5582,
+      "step": 19905
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.81317234357899e-06,
+      "loss": 0.5752,
+      "step": 19906
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.812250776163282e-06,
+      "loss": 0.6171,
+      "step": 19907
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8113292938964384e-06,
+      "loss": 0.586,
+      "step": 19908
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.810407896791143e-06,
+      "loss": 0.5643,
+      "step": 19909
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8094865848600694e-06,
+      "loss": 0.577,
+      "step": 19910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8085653581159043e-06,
+      "loss": 0.5803,
+      "step": 19911
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.807644216571319e-06,
+      "loss": 0.5624,
+      "step": 19912
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.806723160238991e-06,
+      "loss": 0.5818,
+      "step": 19913
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.805802189131593e-06,
+      "loss": 0.5358,
+      "step": 19914
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8048813032618016e-06,
+      "loss": 0.595,
+      "step": 19915
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8039605026422856e-06,
+      "loss": 0.5918,
+      "step": 19916
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.803039787285714e-06,
+      "loss": 0.5368,
+      "step": 19917
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8021191572047646e-06,
+      "loss": 0.5323,
+      "step": 19918
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8011986124121003e-06,
+      "loss": 0.5366,
+      "step": 19919
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.800278152920387e-06,
+      "loss": 0.5864,
+      "step": 19920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7993577787422964e-06,
+      "loss": 0.6485,
+      "step": 19921
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.798437489890492e-06,
+      "loss": 0.5411,
+      "step": 19922
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.797517286377637e-06,
+      "loss": 0.5793,
+      "step": 19923
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7965971682163903e-06,
+      "loss": 0.6051,
+      "step": 19924
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.795677135419421e-06,
+      "loss": 0.5886,
+      "step": 19925
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.794757187999386e-06,
+      "loss": 0.5571,
+      "step": 19926
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7938373259689445e-06,
+      "loss": 0.5543,
+      "step": 19927
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.792917549340751e-06,
+      "loss": 0.5584,
+      "step": 19928
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7919978581274686e-06,
+      "loss": 0.6023,
+      "step": 19929
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7910782523417523e-06,
+      "loss": 0.569,
+      "step": 19930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7901587319962497e-06,
+      "loss": 0.598,
+      "step": 19931
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.789239297103624e-06,
+      "loss": 0.6086,
+      "step": 19932
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.788319947676522e-06,
+      "loss": 0.5622,
+      "step": 19933
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7874006837275966e-06,
+      "loss": 0.5421,
+      "step": 19934
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7864815052694927e-06,
+      "loss": 0.538,
+      "step": 19935
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.785562412314866e-06,
+      "loss": 0.6342,
+      "step": 19936
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7846434048763615e-06,
+      "loss": 0.6627,
+      "step": 19937
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7837244829666255e-06,
+      "loss": 0.5989,
+      "step": 19938
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7828056465982974e-06,
+      "loss": 0.6029,
+      "step": 19939
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7818868957840316e-06,
+      "loss": 0.601,
+      "step": 19940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.780968230536466e-06,
+      "loss": 0.6511,
+      "step": 19941
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.780049650868238e-06,
+      "loss": 0.6092,
+      "step": 19942
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.779131156791995e-06,
+      "loss": 0.6339,
+      "step": 19943
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7782127483203747e-06,
+      "loss": 0.6359,
+      "step": 19944
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7772944254660136e-06,
+      "loss": 0.5935,
+      "step": 19945
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.776376188241545e-06,
+      "loss": 0.5137,
+      "step": 19946
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7754580366596116e-06,
+      "loss": 0.5395,
+      "step": 19947
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7745399707328457e-06,
+      "loss": 0.5774,
+      "step": 19948
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.773621990473877e-06,
+      "loss": 0.5896,
+      "step": 19949
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.772704095895343e-06,
+      "loss": 0.6725,
+      "step": 19950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.771786287009873e-06,
+      "loss": 0.5867,
+      "step": 19951
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7708685638300967e-06,
+      "loss": 0.5228,
+      "step": 19952
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7699509263686385e-06,
+      "loss": 0.6342,
+      "step": 19953
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7690333746381334e-06,
+      "loss": 0.5936,
+      "step": 19954
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7681159086512032e-06,
+      "loss": 0.586,
+      "step": 19955
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7671985284204746e-06,
+      "loss": 0.5567,
+      "step": 19956
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7662812339585675e-06,
+      "loss": 0.5385,
+      "step": 19957
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.765364025278112e-06,
+      "loss": 0.5978,
+      "step": 19958
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.764446902391725e-06,
+      "loss": 0.5798,
+      "step": 19959
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7635298653120244e-06,
+      "loss": 0.6104,
+      "step": 19960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7626129140516354e-06,
+      "loss": 0.5661,
+      "step": 19961
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.761696048623175e-06,
+      "loss": 0.5527,
+      "step": 19962
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7607792690392576e-06,
+      "loss": 0.6449,
+      "step": 19963
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.759862575312495e-06,
+      "loss": 0.5516,
+      "step": 19964
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7589459674555116e-06,
+      "loss": 0.6063,
+      "step": 19965
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.758029445480915e-06,
+      "loss": 0.5524,
+      "step": 19966
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7571130094013175e-06,
+      "loss": 0.616,
+      "step": 19967
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7561966592293274e-06,
+      "loss": 0.533,
+      "step": 19968
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7552803949775617e-06,
+      "loss": 0.6122,
+      "step": 19969
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.754364216658624e-06,
+      "loss": 0.6195,
+      "step": 19970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7534481242851193e-06,
+      "loss": 0.5866,
+      "step": 19971
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7525321178696606e-06,
+      "loss": 0.5597,
+      "step": 19972
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.751616197424849e-06,
+      "loss": 0.5415,
+      "step": 19973
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.750700362963289e-06,
+      "loss": 0.6405,
+      "step": 19974
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.74978461449758e-06,
+      "loss": 0.5594,
+      "step": 19975
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.74886895204033e-06,
+      "loss": 0.5393,
+      "step": 19976
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7479533756041364e-06,
+      "loss": 0.5766,
+      "step": 19977
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7470378852015976e-06,
+      "loss": 0.6403,
+      "step": 19978
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7461224808453078e-06,
+      "loss": 0.6096,
+      "step": 19979
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7452071625478716e-06,
+      "loss": 0.5487,
+      "step": 19980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7442919303218806e-06,
+      "loss": 0.6393,
+      "step": 19981
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7433767841799264e-06,
+      "loss": 0.6245,
+      "step": 19982
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7424617241346086e-06,
+      "loss": 0.5443,
+      "step": 19983
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7415467501985158e-06,
+      "loss": 0.5563,
+      "step": 19984
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7406318623842387e-06,
+      "loss": 0.5996,
+      "step": 19985
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.739717060704363e-06,
+      "loss": 0.5866,
+      "step": 19986
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.738802345171486e-06,
+      "loss": 0.5723,
+      "step": 19987
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7378877157981895e-06,
+      "loss": 0.5949,
+      "step": 19988
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7369731725970605e-06,
+      "loss": 0.5599,
+      "step": 19989
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7360587155806804e-06,
+      "loss": 0.4997,
+      "step": 19990
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.735144344761641e-06,
+      "loss": 0.6131,
+      "step": 19991
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7342300601525184e-06,
+      "loss": 0.5513,
+      "step": 19992
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7333158617658937e-06,
+      "loss": 0.6047,
+      "step": 19993
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7324017496143517e-06,
+      "loss": 0.5769,
+      "step": 19994
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7314877237104686e-06,
+      "loss": 0.5981,
+      "step": 19995
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7305737840668234e-06,
+      "loss": 0.5296,
+      "step": 19996
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.729659930695988e-06,
+      "loss": 0.5939,
+      "step": 19997
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.728746163610545e-06,
+      "loss": 0.6076,
+      "step": 19998
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.727832482823064e-06,
+      "loss": 0.5964,
+      "step": 19999
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7269188883461203e-06,
+      "loss": 0.5985,
+      "step": 20000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.726005380192281e-06,
+      "loss": 0.636,
+      "step": 20001
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7250919583741243e-06,
+      "loss": 0.5767,
+      "step": 20002
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7241786229042153e-06,
+      "loss": 0.6284,
+      "step": 20003
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.723265373795121e-06,
+      "loss": 0.6262,
+      "step": 20004
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7223522110594123e-06,
+      "loss": 0.5635,
+      "step": 20005
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7214391347096535e-06,
+      "loss": 0.5909,
+      "step": 20006
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7205261447584085e-06,
+      "loss": 0.6159,
+      "step": 20007
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7196132412182385e-06,
+      "loss": 0.5585,
+      "step": 20008
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.718700424101712e-06,
+      "loss": 0.5299,
+      "step": 20009
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7177876934213874e-06,
+      "loss": 0.6094,
+      "step": 20010
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.716875049189821e-06,
+      "loss": 0.5849,
+      "step": 20011
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.715962491419577e-06,
+      "loss": 0.5919,
+      "step": 20012
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.715050020123212e-06,
+      "loss": 0.5491,
+      "step": 20013
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7141376353132798e-06,
+      "loss": 0.6104,
+      "step": 20014
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.713225337002334e-06,
+      "loss": 0.5071,
+      "step": 20015
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.712313125202935e-06,
+      "loss": 0.5872,
+      "step": 20016
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7114009999276333e-06,
+      "loss": 0.512,
+      "step": 20017
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.710488961188978e-06,
+      "loss": 0.5668,
+      "step": 20018
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.709577008999519e-06,
+      "loss": 0.6047,
+      "step": 20019
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7086651433718102e-06,
+      "loss": 0.5724,
+      "step": 20020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7077533643183973e-06,
+      "loss": 0.5585,
+      "step": 20021
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.706841671851824e-06,
+      "loss": 0.5407,
+      "step": 20022
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.705930065984642e-06,
+      "loss": 0.5924,
+      "step": 20023
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.705018546729392e-06,
+      "loss": 0.6701,
+      "step": 20024
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7041071140986184e-06,
+      "loss": 0.5258,
+      "step": 20025
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7031957681048604e-06,
+      "loss": 0.5187,
+      "step": 20026
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.702284508760664e-06,
+      "loss": 0.6169,
+      "step": 20027
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.701373336078566e-06,
+      "loss": 0.5878,
+      "step": 20028
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.700462250071107e-06,
+      "loss": 0.5487,
+      "step": 20029
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.699551250750818e-06,
+      "loss": 0.5536,
+      "step": 20030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6986403381302436e-06,
+      "loss": 0.6004,
+      "step": 20031
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.697729512221916e-06,
+      "loss": 0.6174,
+      "step": 20032
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.696818773038363e-06,
+      "loss": 0.5695,
+      "step": 20033
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6959081205921267e-06,
+      "loss": 0.5433,
+      "step": 20034
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6949975548957327e-06,
+      "loss": 0.5949,
+      "step": 20035
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6940870759617133e-06,
+      "loss": 0.577,
+      "step": 20036
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.693176683802593e-06,
+      "loss": 0.5656,
+      "step": 20037
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6922663784309064e-06,
+      "loss": 0.5309,
+      "step": 20038
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6913561598591775e-06,
+      "loss": 0.5375,
+      "step": 20039
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.69044602809993e-06,
+      "loss": 0.5351,
+      "step": 20040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.689535983165686e-06,
+      "loss": 0.5631,
+      "step": 20041
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6886260250689764e-06,
+      "loss": 0.616,
+      "step": 20042
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6877161538223173e-06,
+      "loss": 0.5975,
+      "step": 20043
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.686806369438227e-06,
+      "loss": 0.6345,
+      "step": 20044
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.685896671929232e-06,
+      "loss": 0.5466,
+      "step": 20045
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6849870613078476e-06,
+      "loss": 0.5155,
+      "step": 20046
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.684077537586589e-06,
+      "loss": 0.566,
+      "step": 20047
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6831681007779698e-06,
+      "loss": 0.5949,
+      "step": 20048
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6822587508945116e-06,
+      "loss": 0.611,
+      "step": 20049
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.681349487948724e-06,
+      "loss": 0.6204,
+      "step": 20050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6804403119531206e-06,
+      "loss": 0.621,
+      "step": 20051
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.679531222920206e-06,
+      "loss": 0.531,
+      "step": 20052
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.678622220862501e-06,
+      "loss": 0.6546,
+      "step": 20053
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6777133057925076e-06,
+      "loss": 0.5813,
+      "step": 20054
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6768044777227306e-06,
+      "loss": 0.572,
+      "step": 20055
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6758957366656832e-06,
+      "loss": 0.5491,
+      "step": 20056
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6749870826338685e-06,
+      "loss": 0.6546,
+      "step": 20057
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6740785156397883e-06,
+      "loss": 0.6084,
+      "step": 20058
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.673170035695943e-06,
+      "loss": 0.5737,
+      "step": 20059
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.672261642814842e-06,
+      "loss": 0.5803,
+      "step": 20060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.671353337008979e-06,
+      "loss": 0.5856,
+      "step": 20061
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6704451182908517e-06,
+      "loss": 0.5784,
+      "step": 20062
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6695369866729646e-06,
+      "loss": 0.5794,
+      "step": 20063
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.668628942167812e-06,
+      "loss": 0.5771,
+      "step": 20064
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.667720984787888e-06,
+      "loss": 0.5644,
+      "step": 20065
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6668131145456833e-06,
+      "loss": 0.5568,
+      "step": 20066
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6659053314536987e-06,
+      "loss": 0.55,
+      "step": 20067
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6649976355244223e-06,
+      "loss": 0.5495,
+      "step": 20068
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.664090026770345e-06,
+      "loss": 0.6703,
+      "step": 20069
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6631825052039527e-06,
+      "loss": 0.5238,
+      "step": 20070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6622750708377406e-06,
+      "loss": 0.5568,
+      "step": 20071
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6613677236841927e-06,
+      "loss": 0.6025,
+      "step": 20072
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6604604637557908e-06,
+      "loss": 0.6214,
+      "step": 20073
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6595532910650265e-06,
+      "loss": 0.6116,
+      "step": 20074
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.65864620562438e-06,
+      "loss": 0.6341,
+      "step": 20075
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6577392074463336e-06,
+      "loss": 0.6103,
+      "step": 20076
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6568322965433654e-06,
+      "loss": 0.5632,
+      "step": 20077
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6559254729279624e-06,
+      "loss": 0.5621,
+      "step": 20078
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6550187366125987e-06,
+      "loss": 0.5602,
+      "step": 20079
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6541120876097525e-06,
+      "loss": 0.5235,
+      "step": 20080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6532055259318964e-06,
+      "loss": 0.6014,
+      "step": 20081
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6522990515915125e-06,
+      "loss": 0.6056,
+      "step": 20082
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.651392664601071e-06,
+      "loss": 0.5879,
+      "step": 20083
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6504863649730415e-06,
+      "loss": 0.5572,
+      "step": 20084
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.649580152719903e-06,
+      "loss": 0.5798,
+      "step": 20085
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.648674027854121e-06,
+      "loss": 0.6059,
+      "step": 20086
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6477679903881645e-06,
+      "loss": 0.6212,
+      "step": 20087
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.646862040334499e-06,
+      "loss": 0.5919,
+      "step": 20088
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6459561777055984e-06,
+      "loss": 0.5201,
+      "step": 20089
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.645050402513923e-06,
+      "loss": 0.5984,
+      "step": 20090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6441447147719377e-06,
+      "loss": 0.5846,
+      "step": 20091
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.643239114492103e-06,
+      "loss": 0.5865,
+      "step": 20092
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6423336016868883e-06,
+      "loss": 0.6212,
+      "step": 20093
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6414281763687474e-06,
+      "loss": 0.5368,
+      "step": 20094
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.64052283855014e-06,
+      "loss": 0.5082,
+      "step": 20095
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6396175882435293e-06,
+      "loss": 0.56,
+      "step": 20096
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.638712425461369e-06,
+      "loss": 0.6513,
+      "step": 20097
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.637807350216115e-06,
+      "loss": 0.6389,
+      "step": 20098
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6369023625202194e-06,
+      "loss": 0.533,
+      "step": 20099
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6359974623861415e-06,
+      "loss": 0.6082,
+      "step": 20100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6350926498263307e-06,
+      "loss": 0.5437,
+      "step": 20101
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6341879248532376e-06,
+      "loss": 0.6118,
+      "step": 20102
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.633283287479309e-06,
+      "loss": 0.6383,
+      "step": 20103
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6323787377169994e-06,
+      "loss": 0.5708,
+      "step": 20104
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6314742755787537e-06,
+      "loss": 0.6079,
+      "step": 20105
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6305699010770148e-06,
+      "loss": 0.5788,
+      "step": 20106
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6296656142242335e-06,
+      "loss": 0.6258,
+      "step": 20107
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.628761415032851e-06,
+      "loss": 0.533,
+      "step": 20108
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6278573035153097e-06,
+      "loss": 0.5724,
+      "step": 20109
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.626953279684048e-06,
+      "loss": 0.5371,
+      "step": 20110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.626049343551512e-06,
+      "loss": 0.4889,
+      "step": 20111
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.625145495130138e-06,
+      "loss": 0.5946,
+      "step": 20112
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6242417344323633e-06,
+      "loss": 0.5453,
+      "step": 20113
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6233380614706217e-06,
+      "loss": 0.5913,
+      "step": 20114
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.622434476257354e-06,
+      "loss": 0.6328,
+      "step": 20115
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6215309788049912e-06,
+      "loss": 0.575,
+      "step": 20116
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6206275691259653e-06,
+      "loss": 0.5412,
+      "step": 20117
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.619724247232711e-06,
+      "loss": 0.542,
+      "step": 20118
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.618821013137658e-06,
+      "loss": 0.568,
+      "step": 20119
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6179178668532344e-06,
+      "loss": 0.5884,
+      "step": 20120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6170148083918656e-06,
+      "loss": 0.6048,
+      "step": 20121
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6161118377659855e-06,
+      "loss": 0.5665,
+      "step": 20122
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6152089549880154e-06,
+      "loss": 0.6003,
+      "step": 20123
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.61430616007038e-06,
+      "loss": 0.5869,
+      "step": 20124
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6134034530255034e-06,
+      "loss": 0.6125,
+      "step": 20125
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6125008338658075e-06,
+      "loss": 0.5083,
+      "step": 20126
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.611598302603708e-06,
+      "loss": 0.586,
+      "step": 20127
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6106958592516326e-06,
+      "loss": 0.6164,
+      "step": 20128
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6097935038219976e-06,
+      "loss": 0.5742,
+      "step": 20129
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6088912363272176e-06,
+      "loss": 0.6366,
+      "step": 20130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.607989056779707e-06,
+      "loss": 0.6581,
+      "step": 20131
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.607086965191887e-06,
+      "loss": 0.6358,
+      "step": 20132
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6061849615761678e-06,
+      "loss": 0.5595,
+      "step": 20133
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6052830459449616e-06,
+      "loss": 0.5533,
+      "step": 20134
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6043812183106764e-06,
+      "loss": 0.6281,
+      "step": 20135
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6034794786857284e-06,
+      "loss": 0.6844,
+      "step": 20136
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6025778270825228e-06,
+      "loss": 0.5158,
+      "step": 20137
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.601676263513465e-06,
+      "loss": 0.6144,
+      "step": 20138
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.600774787990967e-06,
+      "loss": 0.5753,
+      "step": 20139
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5998734005274306e-06,
+      "loss": 0.5489,
+      "step": 20140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.59897210113526e-06,
+      "loss": 0.5677,
+      "step": 20141
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5980708898268533e-06,
+      "loss": 0.5385,
+      "step": 20142
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.597169766614621e-06,
+      "loss": 0.5901,
+      "step": 20143
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.596268731510959e-06,
+      "loss": 0.6377,
+      "step": 20144
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.595367784528265e-06,
+      "loss": 0.5902,
+      "step": 20145
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5944669256789357e-06,
+      "loss": 0.5849,
+      "step": 20146
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.593566154975372e-06,
+      "loss": 0.6309,
+      "step": 20147
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.592665472429968e-06,
+      "loss": 0.6327,
+      "step": 20148
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.591764878055113e-06,
+      "loss": 0.5439,
+      "step": 20149
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5908643718632086e-06,
+      "loss": 0.5406,
+      "step": 20150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.589963953866641e-06,
+      "loss": 0.5333,
+      "step": 20151
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.589063624077802e-06,
+      "loss": 0.5946,
+      "step": 20152
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.588163382509078e-06,
+      "loss": 0.5883,
+      "step": 20153
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5872632291728615e-06,
+      "loss": 0.5387,
+      "step": 20154
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5863631640815387e-06,
+      "loss": 0.5422,
+      "step": 20155
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5854631872474943e-06,
+      "loss": 0.6122,
+      "step": 20156
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5845632986831092e-06,
+      "loss": 0.6004,
+      "step": 20157
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5836634984007745e-06,
+      "loss": 0.6477,
+      "step": 20158
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.582763786412867e-06,
+      "loss": 0.5417,
+      "step": 20159
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.581864162731765e-06,
+      "loss": 0.6128,
+      "step": 20160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.580964627369855e-06,
+      "loss": 0.5758,
+      "step": 20161
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5800651803395126e-06,
+      "loss": 0.6854,
+      "step": 20162
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5791658216531145e-06,
+      "loss": 0.5809,
+      "step": 20163
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.578266551323032e-06,
+      "loss": 0.5503,
+      "step": 20164
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5773673693616483e-06,
+      "loss": 0.6105,
+      "step": 20165
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.576468275781333e-06,
+      "loss": 0.5585,
+      "step": 20166
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5755692705944556e-06,
+      "loss": 0.5566,
+      "step": 20167
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5746703538133932e-06,
+      "loss": 0.5834,
+      "step": 20168
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.573771525450512e-06,
+      "loss": 0.5878,
+      "step": 20169
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.572872785518182e-06,
+      "loss": 0.5954,
+      "step": 20170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5719741340287663e-06,
+      "loss": 0.5964,
+      "step": 20171
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5710755709946366e-06,
+      "loss": 0.5353,
+      "step": 20172
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.570177096428158e-06,
+      "loss": 0.5469,
+      "step": 20173
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5692787103416913e-06,
+      "loss": 0.6403,
+      "step": 20174
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5683804127475965e-06,
+      "loss": 0.5856,
+      "step": 20175
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5674822036582425e-06,
+      "loss": 0.5149,
+      "step": 20176
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5665840830859843e-06,
+      "loss": 0.5575,
+      "step": 20177
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.56568605104318e-06,
+      "loss": 0.5906,
+      "step": 20178
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.564788107542192e-06,
+      "loss": 0.5427,
+      "step": 20179
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5638902525953745e-06,
+      "loss": 0.6268,
+      "step": 20180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.562992486215082e-06,
+      "loss": 0.5945,
+      "step": 20181
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5620948084136663e-06,
+      "loss": 0.5608,
+      "step": 20182
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.561197219203486e-06,
+      "loss": 0.5653,
+      "step": 20183
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.560299718596889e-06,
+      "loss": 0.5569,
+      "step": 20184
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.559402306606227e-06,
+      "loss": 0.5144,
+      "step": 20185
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5585049832438446e-06,
+      "loss": 0.593,
+      "step": 20186
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5576077485220973e-06,
+      "loss": 0.6671,
+      "step": 20187
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5567106024533273e-06,
+      "loss": 0.5464,
+      "step": 20188
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5558135450498786e-06,
+      "loss": 0.6161,
+      "step": 20189
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.554916576324101e-06,
+      "loss": 0.5716,
+      "step": 20190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5540196962883343e-06,
+      "loss": 0.6142,
+      "step": 20191
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.553122904954921e-06,
+      "loss": 0.546,
+      "step": 20192
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5522262023361963e-06,
+      "loss": 0.5597,
+      "step": 20193
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5513295884445097e-06,
+      "loss": 0.6143,
+      "step": 20194
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5504330632921934e-06,
+      "loss": 0.5896,
+      "step": 20195
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5495366268915853e-06,
+      "loss": 0.5807,
+      "step": 20196
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.548640279255019e-06,
+      "loss": 0.5154,
+      "step": 20197
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5477440203948333e-06,
+      "loss": 0.6151,
+      "step": 20198
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5468478503233593e-06,
+      "loss": 0.5372,
+      "step": 20199
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5459517690529254e-06,
+      "loss": 0.5635,
+      "step": 20200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5450557765958706e-06,
+      "loss": 0.5962,
+      "step": 20201
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.544159872964519e-06,
+      "loss": 0.6139,
+      "step": 20202
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.543264058171201e-06,
+      "loss": 0.5463,
+      "step": 20203
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5423683322282386e-06,
+      "loss": 0.6524,
+      "step": 20204
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5414726951479674e-06,
+      "loss": 0.5855,
+      "step": 20205
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5405771469427052e-06,
+      "loss": 0.5786,
+      "step": 20206
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5396816876247774e-06,
+      "loss": 0.5708,
+      "step": 20207
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.538786317206503e-06,
+      "loss": 0.6202,
+      "step": 20208
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5378910357002093e-06,
+      "loss": 0.5707,
+      "step": 20209
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.536995843118214e-06,
+      "loss": 0.5876,
+      "step": 20210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.536100739472831e-06,
+      "loss": 0.6404,
+      "step": 20211
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5352057247763837e-06,
+      "loss": 0.6168,
+      "step": 20212
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5343107990411875e-06,
+      "loss": 0.5434,
+      "step": 20213
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5334159622795548e-06,
+      "loss": 0.6014,
+      "step": 20214
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.532521214503798e-06,
+      "loss": 0.6181,
+      "step": 20215
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.531626555726235e-06,
+      "loss": 0.5865,
+      "step": 20216
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.530731985959175e-06,
+      "loss": 0.5667,
+      "step": 20217
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.529837505214927e-06,
+      "loss": 0.6088,
+      "step": 20218
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.528943113505796e-06,
+      "loss": 0.5507,
+      "step": 20219
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.528048810844098e-06,
+      "loss": 0.5494,
+      "step": 20220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.527154597242135e-06,
+      "loss": 0.5758,
+      "step": 20221
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5262604727122084e-06,
+      "loss": 0.5752,
+      "step": 20222
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5253664372666297e-06,
+      "loss": 0.588,
+      "step": 20223
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5244724909176973e-06,
+      "loss": 0.6542,
+      "step": 20224
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5235786336777144e-06,
+      "loss": 0.5904,
+      "step": 20225
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5226848655589763e-06,
+      "loss": 0.6512,
+      "step": 20226
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5217911865737885e-06,
+      "loss": 0.5632,
+      "step": 20227
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5208975967344472e-06,
+      "loss": 0.5068,
+      "step": 20228
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5200040960532443e-06,
+      "loss": 0.5857,
+      "step": 20229
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.519110684542483e-06,
+      "loss": 0.6402,
+      "step": 20230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5182173622144524e-06,
+      "loss": 0.6613,
+      "step": 20231
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5173241290814464e-06,
+      "loss": 0.64,
+      "step": 20232
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5164309851557532e-06,
+      "loss": 0.5106,
+      "step": 20233
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5155379304496705e-06,
+      "loss": 0.5712,
+      "step": 20234
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.514644964975483e-06,
+      "loss": 0.6098,
+      "step": 20235
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.513752088745479e-06,
+      "loss": 0.6112,
+      "step": 20236
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.512859301771944e-06,
+      "loss": 0.5393,
+      "step": 20237
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5119666040671663e-06,
+      "loss": 0.6121,
+      "step": 20238
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5110739956434303e-06,
+      "loss": 0.6556,
+      "step": 20239
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.510181476513014e-06,
+      "loss": 0.6083,
+      "step": 20240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.509289046688207e-06,
+      "loss": 0.598,
+      "step": 20241
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5083967061812852e-06,
+      "loss": 0.5924,
+      "step": 20242
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.507504455004529e-06,
+      "loss": 0.5618,
+      "step": 20243
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.506612293170213e-06,
+      "loss": 0.5622,
+      "step": 20244
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5057202206906205e-06,
+      "loss": 0.5663,
+      "step": 20245
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.504828237578024e-06,
+      "loss": 0.5474,
+      "step": 20246
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5039363438446984e-06,
+      "loss": 0.5424,
+      "step": 20247
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.503044539502913e-06,
+      "loss": 0.6024,
+      "step": 20248
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5021528245649473e-06,
+      "loss": 0.64,
+      "step": 20249
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.501261199043069e-06,
+      "loss": 0.6024,
+      "step": 20250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.500369662949542e-06,
+      "loss": 0.6133,
+      "step": 20251
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4994782162966435e-06,
+      "loss": 0.6247,
+      "step": 20252
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4985868590966364e-06,
+      "loss": 0.5763,
+      "step": 20253
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.497695591361787e-06,
+      "loss": 0.6267,
+      "step": 20254
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4968044131043565e-06,
+      "loss": 0.5163,
+      "step": 20255
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.495913324336615e-06,
+      "loss": 0.6057,
+      "step": 20256
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4950223250708204e-06,
+      "loss": 0.5566,
+      "step": 20257
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.494131415319235e-06,
+      "loss": 0.6557,
+      "step": 20258
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.493240595094114e-06,
+      "loss": 0.6299,
+      "step": 20259
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4923498644077236e-06,
+      "loss": 0.564,
+      "step": 20260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4914592232723164e-06,
+      "loss": 0.5806,
+      "step": 20261
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4905686717001473e-06,
+      "loss": 0.5921,
+      "step": 20262
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4896782097034755e-06,
+      "loss": 0.6169,
+      "step": 20263
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4887878372945517e-06,
+      "loss": 0.6021,
+      "step": 20264
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4878975544856285e-06,
+      "loss": 0.5331,
+      "step": 20265
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4870073612889544e-06,
+      "loss": 0.5997,
+      "step": 20266
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4861172577167845e-06,
+      "loss": 0.6148,
+      "step": 20267
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.485227243781365e-06,
+      "loss": 0.5821,
+      "step": 20268
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4843373194949427e-06,
+      "loss": 0.5541,
+      "step": 20269
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.483447484869761e-06,
+      "loss": 0.5857,
+      "step": 20270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.482557739918071e-06,
+      "loss": 0.5645,
+      "step": 20271
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.481668084652113e-06,
+      "loss": 0.5874,
+      "step": 20272
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.480778519084127e-06,
+      "loss": 0.605,
+      "step": 20273
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.47988904322636e-06,
+      "loss": 0.5261,
+      "step": 20274
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.478999657091049e-06,
+      "loss": 0.5895,
+      "step": 20275
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.478110360690432e-06,
+      "loss": 0.5976,
+      "step": 20276
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.477221154036744e-06,
+      "loss": 0.6501,
+      "step": 20277
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.476332037142227e-06,
+      "loss": 0.595,
+      "step": 20278
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.475443010019115e-06,
+      "loss": 0.6715,
+      "step": 20279
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.474554072679639e-06,
+      "loss": 0.5435,
+      "step": 20280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.473665225136029e-06,
+      "loss": 0.4821,
+      "step": 20281
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.472776467400525e-06,
+      "loss": 0.6143,
+      "step": 20282
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4718877994853508e-06,
+      "loss": 0.5891,
+      "step": 20283
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.470999221402733e-06,
+      "loss": 0.5667,
+      "step": 20284
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4701107331649066e-06,
+      "loss": 0.5386,
+      "step": 20285
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4692223347840947e-06,
+      "loss": 0.5594,
+      "step": 20286
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4683340262725217e-06,
+      "loss": 0.5474,
+      "step": 20287
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.467445807642408e-06,
+      "loss": 0.5497,
+      "step": 20288
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.466557678905983e-06,
+      "loss": 0.6076,
+      "step": 20289
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4656696400754652e-06,
+      "loss": 0.6119,
+      "step": 20290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.464781691163073e-06,
+      "loss": 0.5755,
+      "step": 20291
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.463893832181029e-06,
+      "loss": 0.6065,
+      "step": 20292
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4630060631415487e-06,
+      "loss": 0.605,
+      "step": 20293
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4621183840568495e-06,
+      "loss": 0.5921,
+      "step": 20294
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.461230794939142e-06,
+      "loss": 0.5817,
+      "step": 20295
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4603432958006488e-06,
+      "loss": 0.5443,
+      "step": 20296
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4594558866535778e-06,
+      "loss": 0.5377,
+      "step": 20297
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.45856856751014e-06,
+      "loss": 0.624,
+      "step": 20298
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4576813383825436e-06,
+      "loss": 0.6425,
+      "step": 20299
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4567941992830034e-06,
+      "loss": 0.5752,
+      "step": 20300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4559071502237252e-06,
+      "loss": 0.5676,
+      "step": 20301
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.455020191216911e-06,
+      "loss": 0.6158,
+      "step": 20302
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.454133322274773e-06,
+      "loss": 0.6319,
+      "step": 20303
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.453246543409513e-06,
+      "loss": 0.612,
+      "step": 20304
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4523598546333325e-06,
+      "loss": 0.5961,
+      "step": 20305
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.451473255958431e-06,
+      "loss": 0.5819,
+      "step": 20306
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.450586747397013e-06,
+      "loss": 0.5585,
+      "step": 20307
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4497003289612783e-06,
+      "loss": 0.6216,
+      "step": 20308
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.448814000663422e-06,
+      "loss": 0.6681,
+      "step": 20309
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4479277625156383e-06,
+      "loss": 0.5542,
+      "step": 20310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4470416145301276e-06,
+      "loss": 0.6717,
+      "step": 20311
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4461555567190842e-06,
+      "loss": 0.5849,
+      "step": 20312
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.445269589094694e-06,
+      "loss": 0.5709,
+      "step": 20313
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.444383711669158e-06,
+      "loss": 0.5638,
+      "step": 20314
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.443497924454663e-06,
+      "loss": 0.6069,
+      "step": 20315
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4426122274633968e-06,
+      "loss": 0.5091,
+      "step": 20316
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.441726620707545e-06,
+      "loss": 0.5745,
+      "step": 20317
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4408411041993007e-06,
+      "loss": 0.602,
+      "step": 20318
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.439955677950846e-06,
+      "loss": 0.5975,
+      "step": 20319
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4390703419743654e-06,
+      "loss": 0.5843,
+      "step": 20320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4381850962820396e-06,
+      "loss": 0.5574,
+      "step": 20321
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.437299940886054e-06,
+      "loss": 0.5889,
+      "step": 20322
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4364148757985895e-06,
+      "loss": 0.5994,
+      "step": 20323
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4355299010318187e-06,
+      "loss": 0.5377,
+      "step": 20324
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4346450165979294e-06,
+      "loss": 0.5374,
+      "step": 20325
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4337602225090926e-06,
+      "loss": 0.591,
+      "step": 20326
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4328755187774853e-06,
+      "loss": 0.6082,
+      "step": 20327
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.431990905415277e-06,
+      "loss": 0.5229,
+      "step": 20328
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.431106382434649e-06,
+      "loss": 0.5949,
+      "step": 20329
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4302219498477683e-06,
+      "loss": 0.5878,
+      "step": 20330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4293376076668073e-06,
+      "loss": 0.605,
+      "step": 20331
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.42845335590393e-06,
+      "loss": 0.6272,
+      "step": 20332
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.427569194571313e-06,
+      "loss": 0.5993,
+      "step": 20333
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.426685123681119e-06,
+      "loss": 0.5909,
+      "step": 20334
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.425801143245513e-06,
+      "loss": 0.5832,
+      "step": 20335
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4249172532766607e-06,
+      "loss": 0.5649,
+      "step": 20336
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4240334537867236e-06,
+      "loss": 0.6294,
+      "step": 20337
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4231497447878604e-06,
+      "loss": 0.6228,
+      "step": 20338
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4222661262922395e-06,
+      "loss": 0.6335,
+      "step": 20339
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4213825983120175e-06,
+      "loss": 0.5265,
+      "step": 20340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4204991608593497e-06,
+      "loss": 0.5553,
+      "step": 20341
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.419615813946392e-06,
+      "loss": 0.57,
+      "step": 20342
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4187325575853068e-06,
+      "loss": 0.6002,
+      "step": 20343
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.417849391788244e-06,
+      "loss": 0.6094,
+      "step": 20344
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.416966316567354e-06,
+      "loss": 0.5637,
+      "step": 20345
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.416083331934794e-06,
+      "loss": 0.612,
+      "step": 20346
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4152004379027136e-06,
+      "loss": 0.5472,
+      "step": 20347
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4143176344832605e-06,
+      "loss": 0.6005,
+      "step": 20348
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4134349216885808e-06,
+      "loss": 0.5397,
+      "step": 20349
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.412552299530827e-06,
+      "loss": 0.6155,
+      "step": 20350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.411669768022141e-06,
+      "loss": 0.5395,
+      "step": 20351
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4107873271746695e-06,
+      "loss": 0.6064,
+      "step": 20352
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.40990497700055e-06,
+      "loss": 0.5467,
+      "step": 20353
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4090227175119315e-06,
+      "loss": 0.486,
+      "step": 20354
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.408140548720953e-06,
+      "loss": 0.6222,
+      "step": 20355
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.407258470639748e-06,
+      "loss": 0.6205,
+      "step": 20356
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4063764832804637e-06,
+      "loss": 0.5999,
+      "step": 20357
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4054945866552324e-06,
+      "loss": 0.6259,
+      "step": 20358
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.40461278077619e-06,
+      "loss": 0.5504,
+      "step": 20359
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4037310656554668e-06,
+      "loss": 0.6327,
+      "step": 20360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.402849441305205e-06,
+      "loss": 0.6218,
+      "step": 20361
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.40196790773753e-06,
+      "loss": 0.5788,
+      "step": 20362
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.401086464964575e-06,
+      "loss": 0.545,
+      "step": 20363
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.400205112998466e-06,
+      "loss": 0.5813,
+      "step": 20364
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.399323851851336e-06,
+      "loss": 0.5487,
+      "step": 20365
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.39844268153531e-06,
+      "loss": 0.5505,
+      "step": 20366
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.397561602062509e-06,
+      "loss": 0.5189,
+      "step": 20367
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3966806134450657e-06,
+      "loss": 0.609,
+      "step": 20368
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3957997156951007e-06,
+      "loss": 0.577,
+      "step": 20369
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3949189088247326e-06,
+      "loss": 0.6228,
+      "step": 20370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.394038192846082e-06,
+      "loss": 0.4928,
+      "step": 20371
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3931575677712738e-06,
+      "loss": 0.6631,
+      "step": 20372
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3922770336124223e-06,
+      "loss": 0.6053,
+      "step": 20373
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3913965903816446e-06,
+      "loss": 0.5906,
+      "step": 20374
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.390516238091053e-06,
+      "loss": 0.5713,
+      "step": 20375
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.38963597675277e-06,
+      "loss": 0.5638,
+      "step": 20376
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3887558063789037e-06,
+      "loss": 0.6476,
+      "step": 20377
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.387875726981563e-06,
+      "loss": 0.5082,
+      "step": 20378
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3869957385728657e-06,
+      "loss": 0.5619,
+      "step": 20379
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.386115841164919e-06,
+      "loss": 0.5624,
+      "step": 20380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3852360347698274e-06,
+      "loss": 0.5739,
+      "step": 20381
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3843563193996985e-06,
+      "loss": 0.5498,
+      "step": 20382
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.383476695066643e-06,
+      "loss": 0.5359,
+      "step": 20383
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.382597161782761e-06,
+      "loss": 0.5505,
+      "step": 20384
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.381717719560155e-06,
+      "loss": 0.5186,
+      "step": 20385
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.38083836841093e-06,
+      "loss": 0.5667,
+      "step": 20386
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3799591083471862e-06,
+      "loss": 0.6323,
+      "step": 20387
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3790799393810227e-06,
+      "loss": 0.5806,
+      "step": 20388
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.378200861524532e-06,
+      "loss": 0.5838,
+      "step": 20389
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3773218747898195e-06,
+      "loss": 0.5033,
+      "step": 20390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3764429791889774e-06,
+      "loss": 0.5262,
+      "step": 20391
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3755641747341008e-06,
+      "loss": 0.6198,
+      "step": 20392
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.374685461437277e-06,
+      "loss": 0.5712,
+      "step": 20393
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.373806839310607e-06,
+      "loss": 0.561,
+      "step": 20394
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3729283083661767e-06,
+      "loss": 0.5676,
+      "step": 20395
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.372049868616073e-06,
+      "loss": 0.5806,
+      "step": 20396
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3711715200723904e-06,
+      "loss": 0.6049,
+      "step": 20397
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.370293262747212e-06,
+      "loss": 0.5867,
+      "step": 20398
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.369415096652624e-06,
+      "loss": 0.5588,
+      "step": 20399
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.368537021800706e-06,
+      "loss": 0.6249,
+      "step": 20400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3676590382035503e-06,
+      "loss": 0.5284,
+      "step": 20401
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3667811458732323e-06,
+      "loss": 0.5803,
+      "step": 20402
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.365903344821836e-06,
+      "loss": 0.5932,
+      "step": 20403
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.365025635061434e-06,
+      "loss": 0.5426,
+      "step": 20404
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.364148016604114e-06,
+      "loss": 0.6179,
+      "step": 20405
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3632704894619473e-06,
+      "loss": 0.5134,
+      "step": 20406
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.362393053647006e-06,
+      "loss": 0.5839,
+      "step": 20407
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3615157091713734e-06,
+      "loss": 0.5848,
+      "step": 20408
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3606384560471173e-06,
+      "loss": 0.4974,
+      "step": 20409
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.35976129428631e-06,
+      "loss": 0.6373,
+      "step": 20410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.358884223901018e-06,
+      "loss": 0.5564,
+      "step": 20411
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3580072449033186e-06,
+      "loss": 0.5632,
+      "step": 20412
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.357130357305275e-06,
+      "loss": 0.4941,
+      "step": 20413
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.356253561118955e-06,
+      "loss": 0.5753,
+      "step": 20414
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.355376856356419e-06,
+      "loss": 0.5844,
+      "step": 20415
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3545002430297414e-06,
+      "loss": 0.6028,
+      "step": 20416
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.353623721150978e-06,
+      "loss": 0.6117,
+      "step": 20417
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.35274729073219e-06,
+      "loss": 0.5998,
+      "step": 20418
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3518709517854433e-06,
+      "loss": 0.577,
+      "step": 20419
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.350994704322794e-06,
+      "loss": 0.59,
+      "step": 20420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.350118548356299e-06,
+      "loss": 0.5719,
+      "step": 20421
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.349242483898013e-06,
+      "loss": 0.6444,
+      "step": 20422
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.348366510959998e-06,
+      "loss": 0.6263,
+      "step": 20423
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.347490629554304e-06,
+      "loss": 0.4931,
+      "step": 20424
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3466148396929843e-06,
+      "loss": 0.5744,
+      "step": 20425
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.345739141388088e-06,
+      "loss": 0.5577,
+      "step": 20426
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.344863534651671e-06,
+      "loss": 0.6177,
+      "step": 20427
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3439880194957785e-06,
+      "loss": 0.5727,
+      "step": 20428
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3431125959324574e-06,
+      "loss": 0.5463,
+      "step": 20429
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.342237263973759e-06,
+      "loss": 0.642,
+      "step": 20430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3413620236317255e-06,
+      "loss": 0.53,
+      "step": 20431
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3404868749184015e-06,
+      "loss": 0.6003,
+      "step": 20432
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.339611817845827e-06,
+      "loss": 0.6073,
+      "step": 20433
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.338736852426049e-06,
+      "loss": 0.5911,
+      "step": 20434
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3378619786711052e-06,
+      "loss": 0.5908,
+      "step": 20435
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3369871965930346e-06,
+      "loss": 0.5085,
+      "step": 20436
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.33611250620387e-06,
+      "loss": 0.5445,
+      "step": 20437
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3352379075156582e-06,
+      "loss": 0.5993,
+      "step": 20438
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3343634005404278e-06,
+      "loss": 0.571,
+      "step": 20439
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.333488985290211e-06,
+      "loss": 0.5468,
+      "step": 20440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3326146617770462e-06,
+      "loss": 0.5776,
+      "step": 20441
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.331740430012963e-06,
+      "loss": 0.6766,
+      "step": 20442
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3308662900099895e-06,
+      "loss": 0.521,
+      "step": 20443
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3299922417801533e-06,
+      "loss": 0.5966,
+      "step": 20444
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3291182853354888e-06,
+      "loss": 0.6099,
+      "step": 20445
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3282444206880173e-06,
+      "loss": 0.5649,
+      "step": 20446
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3273706478497613e-06,
+      "loss": 0.5384,
+      "step": 20447
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3264969668327516e-06,
+      "loss": 0.5365,
+      "step": 20448
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.325623377649009e-06,
+      "loss": 0.5134,
+      "step": 20449
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3247498803105517e-06,
+      "loss": 0.6094,
+      "step": 20450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.323876474829398e-06,
+      "loss": 0.6168,
+      "step": 20451
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.323003161217575e-06,
+      "loss": 0.6536,
+      "step": 20452
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.322129939487094e-06,
+      "loss": 0.6118,
+      "step": 20453
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.321256809649973e-06,
+      "loss": 0.6364,
+      "step": 20454
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3203837717182243e-06,
+      "loss": 0.6427,
+      "step": 20455
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3195108257038665e-06,
+      "loss": 0.5553,
+      "step": 20456
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.318637971618911e-06,
+      "loss": 0.5997,
+      "step": 20457
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.317765209475363e-06,
+      "loss": 0.5333,
+      "step": 20458
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3168925392852414e-06,
+      "loss": 0.667,
+      "step": 20459
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3160199610605506e-06,
+      "loss": 0.6067,
+      "step": 20460
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.315147474813297e-06,
+      "loss": 0.56,
+      "step": 20461
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3142750805554867e-06,
+      "loss": 0.5586,
+      "step": 20462
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.313402778299127e-06,
+      "loss": 0.6,
+      "step": 20463
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3125305680562226e-06,
+      "loss": 0.6047,
+      "step": 20464
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3116584498387726e-06,
+      "loss": 0.5685,
+      "step": 20465
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.310786423658776e-06,
+      "loss": 0.5731,
+      "step": 20466
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3099144895282396e-06,
+      "loss": 0.5395,
+      "step": 20467
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3090426474591573e-06,
+      "loss": 0.5176,
+      "step": 20468
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.308170897463525e-06,
+      "loss": 0.6384,
+      "step": 20469
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.307299239553344e-06,
+      "loss": 0.6191,
+      "step": 20470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3064276737406064e-06,
+      "loss": 0.6187,
+      "step": 20471
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3055562000373055e-06,
+      "loss": 0.59,
+      "step": 20472
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3046848184554294e-06,
+      "loss": 0.5968,
+      "step": 20473
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3038135290069773e-06,
+      "loss": 0.5893,
+      "step": 20474
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3029423317039345e-06,
+      "loss": 0.5655,
+      "step": 20475
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.30207122655829e-06,
+      "loss": 0.604,
+      "step": 20476
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3012002135820278e-06,
+      "loss": 0.5124,
+      "step": 20477
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.30032929278714e-06,
+      "loss": 0.5391,
+      "step": 20478
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2994584641856074e-06,
+      "loss": 0.5994,
+      "step": 20479
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2985877277894107e-06,
+      "loss": 0.574,
+      "step": 20480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.297717083610539e-06,
+      "loss": 0.5486,
+      "step": 20481
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.296846531660969e-06,
+      "loss": 0.5352,
+      "step": 20482
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.29597607195268e-06,
+      "loss": 0.6185,
+      "step": 20483
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2951057044976487e-06,
+      "loss": 0.5987,
+      "step": 20484
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.294235429307857e-06,
+      "loss": 0.5788,
+      "step": 20485
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.293365246395278e-06,
+      "loss": 0.5588,
+      "step": 20486
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2924951557718877e-06,
+      "loss": 0.5577,
+      "step": 20487
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2916251574496526e-06,
+      "loss": 0.5943,
+      "step": 20488
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.290755251440554e-06,
+      "loss": 0.5819,
+      "step": 20489
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.289885437756558e-06,
+      "loss": 0.5801,
+      "step": 20490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2890157164096315e-06,
+      "loss": 0.6411,
+      "step": 20491
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.288146087411749e-06,
+      "loss": 0.5548,
+      "step": 20492
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2872765507748726e-06,
+      "loss": 0.5503,
+      "step": 20493
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2864071065109702e-06,
+      "loss": 0.5889,
+      "step": 20494
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2855377546320023e-06,
+      "loss": 0.5965,
+      "step": 20495
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2846684951499365e-06,
+      "loss": 0.5826,
+      "step": 20496
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2837993280767343e-06,
+      "loss": 0.6083,
+      "step": 20497
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2829302534243537e-06,
+      "loss": 0.6027,
+      "step": 20498
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2820612712047517e-06,
+      "loss": 0.6228,
+      "step": 20499
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.281192381429894e-06,
+      "loss": 0.5864,
+      "step": 20500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.280323584111732e-06,
+      "loss": 0.5593,
+      "step": 20501
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.279454879262218e-06,
+      "loss": 0.5717,
+      "step": 20502
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.278586266893314e-06,
+      "loss": 0.5521,
+      "step": 20503
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2777177470169687e-06,
+      "loss": 0.5381,
+      "step": 20504
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2768493196451346e-06,
+      "loss": 0.5537,
+      "step": 20505
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.275980984789757e-06,
+      "loss": 0.5752,
+      "step": 20506
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.275112742462794e-06,
+      "loss": 0.5965,
+      "step": 20507
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2742445926761867e-06,
+      "loss": 0.5261,
+      "step": 20508
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2733765354418822e-06,
+      "loss": 0.5705,
+      "step": 20509
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2725085707718295e-06,
+      "loss": 0.6037,
+      "step": 20510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2716406986779704e-06,
+      "loss": 0.5716,
+      "step": 20511
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.270772919172247e-06,
+      "loss": 0.5904,
+      "step": 20512
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.269905232266597e-06,
+      "loss": 0.6077,
+      "step": 20513
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2690376379729694e-06,
+      "loss": 0.5951,
+      "step": 20514
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.268170136303297e-06,
+      "loss": 0.5663,
+      "step": 20515
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.267302727269519e-06,
+      "loss": 0.5203,
+      "step": 20516
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.266435410883568e-06,
+      "loss": 0.6127,
+      "step": 20517
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2655681871573864e-06,
+      "loss": 0.5292,
+      "step": 20518
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2647010561029034e-06,
+      "loss": 0.6239,
+      "step": 20519
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2638340177320495e-06,
+      "loss": 0.5633,
+      "step": 20520
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.262967072056761e-06,
+      "loss": 0.5567,
+      "step": 20521
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.262100219088965e-06,
+      "loss": 0.5955,
+      "step": 20522
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2612334588405902e-06,
+      "loss": 0.5501,
+      "step": 20523
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2603667913235616e-06,
+      "loss": 0.5025,
+      "step": 20524
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2595002165498114e-06,
+      "loss": 0.5886,
+      "step": 20525
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.25863373453126e-06,
+      "loss": 0.6124,
+      "step": 20526
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.257767345279833e-06,
+      "loss": 0.617,
+      "step": 20527
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2569010488074482e-06,
+      "loss": 0.5819,
+      "step": 20528
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2560348451260325e-06,
+      "loss": 0.5892,
+      "step": 20529
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2551687342475035e-06,
+      "loss": 0.5506,
+      "step": 20530
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.254302716183776e-06,
+      "loss": 0.6322,
+      "step": 20531
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2534367909467735e-06,
+      "loss": 0.5573,
+      "step": 20532
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.252570958548409e-06,
+      "loss": 0.6406,
+      "step": 20533
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.251705219000596e-06,
+      "loss": 0.6247,
+      "step": 20534
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2508395723152464e-06,
+      "loss": 0.5441,
+      "step": 20535
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2499740185042783e-06,
+      "loss": 0.582,
+      "step": 20536
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.249108557579598e-06,
+      "loss": 0.6529,
+      "step": 20537
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2482431895531163e-06,
+      "loss": 0.6115,
+      "step": 20538
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2473779144367367e-06,
+      "loss": 0.5294,
+      "step": 20539
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.246512732242375e-06,
+      "loss": 0.583,
+      "step": 20540
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2456476429819315e-06,
+      "loss": 0.5258,
+      "step": 20541
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2447826466673084e-06,
+      "loss": 0.5693,
+      "step": 20542
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.243917743310414e-06,
+      "loss": 0.5241,
+      "step": 20543
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2430529329231497e-06,
+      "loss": 0.5983,
+      "step": 20544
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2421882155174124e-06,
+      "loss": 0.5113,
+      "step": 20545
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2413235911051044e-06,
+      "loss": 0.5751,
+      "step": 20546
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2404590596981224e-06,
+      "loss": 0.5637,
+      "step": 20547
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2395946213083617e-06,
+      "loss": 0.5965,
+      "step": 20548
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.238730275947718e-06,
+      "loss": 0.6543,
+      "step": 20549
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.237866023628089e-06,
+      "loss": 0.54,
+      "step": 20550
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2370018643613642e-06,
+      "loss": 0.5588,
+      "step": 20551
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2361377981594344e-06,
+      "loss": 0.5687,
+      "step": 20552
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.235273825034195e-06,
+      "loss": 0.555,
+      "step": 20553
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2344099449975308e-06,
+      "loss": 0.598,
+      "step": 20554
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2335461580613315e-06,
+      "loss": 0.57,
+      "step": 20555
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2326824642374787e-06,
+      "loss": 0.6606,
+      "step": 20556
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2318188635378655e-06,
+      "loss": 0.5085,
+      "step": 20557
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2309553559743712e-06,
+      "loss": 0.5534,
+      "step": 20558
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.230091941558879e-06,
+      "loss": 0.5813,
+      "step": 20559
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2292286203032674e-06,
+      "loss": 0.5638,
+      "step": 20560
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2283653922194223e-06,
+      "loss": 0.5691,
+      "step": 20561
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2275022573192192e-06,
+      "loss": 0.5528,
+      "step": 20562
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.226639215614533e-06,
+      "loss": 0.5557,
+      "step": 20563
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2257762671172455e-06,
+      "loss": 0.5637,
+      "step": 20564
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.224913411839229e-06,
+      "loss": 0.6451,
+      "step": 20565
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.224050649792357e-06,
+      "loss": 0.5398,
+      "step": 20566
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2231879809885e-06,
+      "loss": 0.5796,
+      "step": 20567
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2223254054395336e-06,
+      "loss": 0.6519,
+      "step": 20568
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2214629231573248e-06,
+      "loss": 0.5346,
+      "step": 20569
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2206005341537417e-06,
+      "loss": 0.5857,
+      "step": 20570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2197382384406493e-06,
+      "loss": 0.5991,
+      "step": 20571
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.21887603602992e-06,
+      "loss": 0.6485,
+      "step": 20572
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2180139269334145e-06,
+      "loss": 0.5799,
+      "step": 20573
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2171519111629933e-06,
+      "loss": 0.6246,
+      "step": 20574
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2162899887305254e-06,
+      "loss": 0.5312,
+      "step": 20575
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2154281596478675e-06,
+      "loss": 0.5403,
+      "step": 20576
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.21456642392688e-06,
+      "loss": 0.5524,
+      "step": 20577
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2137047815794177e-06,
+      "loss": 0.6288,
+      "step": 20578
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.212843232617343e-06,
+      "loss": 0.5523,
+      "step": 20579
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2119817770525086e-06,
+      "loss": 0.5858,
+      "step": 20580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2111204148967713e-06,
+      "loss": 0.4925,
+      "step": 20581
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.210259146161978e-06,
+      "loss": 0.5759,
+      "step": 20582
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2093979708599877e-06,
+      "loss": 0.5982,
+      "step": 20583
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.208536889002648e-06,
+      "loss": 0.5708,
+      "step": 20584
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2076759006018055e-06,
+      "loss": 0.549,
+      "step": 20585
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.206815005669315e-06,
+      "loss": 0.5627,
+      "step": 20586
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2059542042170177e-06,
+      "loss": 0.557,
+      "step": 20587
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2050934962567604e-06,
+      "loss": 0.5845,
+      "step": 20588
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.204232881800384e-06,
+      "loss": 0.5652,
+      "step": 20589
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.203372360859739e-06,
+      "loss": 0.6281,
+      "step": 20590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.20251193344666e-06,
+      "loss": 0.5567,
+      "step": 20591
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.201651599572991e-06,
+      "loss": 0.5402,
+      "step": 20592
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2007913592505668e-06,
+      "loss": 0.5841,
+      "step": 20593
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1999312124912306e-06,
+      "loss": 0.5406,
+      "step": 20594
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1990711593068158e-06,
+      "loss": 0.561,
+      "step": 20595
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1982111997091536e-06,
+      "loss": 0.5723,
+      "step": 20596
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.197351333710086e-06,
+      "loss": 0.5794,
+      "step": 20597
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.196491561321442e-06,
+      "loss": 0.5319,
+      "step": 20598
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1956318825550516e-06,
+      "loss": 0.5943,
+      "step": 20599
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.194772297422741e-06,
+      "loss": 0.5958,
+      "step": 20600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1939128059363477e-06,
+      "loss": 0.5805,
+      "step": 20601
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1930534081076947e-06,
+      "loss": 0.6177,
+      "step": 20602
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.192194103948607e-06,
+      "loss": 0.6344,
+      "step": 20603
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1913348934709076e-06,
+      "loss": 0.5114,
+      "step": 20604
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1904757766864258e-06,
+      "loss": 0.6273,
+      "step": 20605
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1896167536069813e-06,
+      "loss": 0.5709,
+      "step": 20606
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1887578242443895e-06,
+      "loss": 0.567,
+      "step": 20607
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1878989886104792e-06,
+      "loss": 0.5717,
+      "step": 20608
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.187040246717065e-06,
+      "loss": 0.5152,
+      "step": 20609
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1861815985759625e-06,
+      "loss": 0.5353,
+      "step": 20610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.185323044198986e-06,
+      "loss": 0.5626,
+      "step": 20611
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.184464583597955e-06,
+      "loss": 0.5733,
+      "step": 20612
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1836062167846803e-06,
+      "loss": 0.5933,
+      "step": 20613
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.18274794377097e-06,
+      "loss": 0.5563,
+      "step": 20614
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.181889764568641e-06,
+      "loss": 0.5742,
+      "step": 20615
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1810316791895003e-06,
+      "loss": 0.5911,
+      "step": 20616
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.180173687645355e-06,
+      "loss": 0.5987,
+      "step": 20617
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.179315789948009e-06,
+      "loss": 0.5282,
+      "step": 20618
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1784579861092744e-06,
+      "loss": 0.6255,
+      "step": 20619
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.177600276140952e-06,
+      "loss": 0.6401,
+      "step": 20620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1767426600548434e-06,
+      "loss": 0.5632,
+      "step": 20621
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1758851378627485e-06,
+      "loss": 0.5583,
+      "step": 20622
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1750277095764747e-06,
+      "loss": 0.5727,
+      "step": 20623
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1741703752078155e-06,
+      "loss": 0.6568,
+      "step": 20624
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1733131347685665e-06,
+      "loss": 0.6407,
+      "step": 20625
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1724559882705306e-06,
+      "loss": 0.5533,
+      "step": 20626
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1715989357254993e-06,
+      "loss": 0.5595,
+      "step": 20627
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1707419771452663e-06,
+      "loss": 0.6179,
+      "step": 20628
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.169885112541622e-06,
+      "loss": 0.5514,
+      "step": 20629
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1690283419263634e-06,
+      "loss": 0.5995,
+      "step": 20630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1681716653112757e-06,
+      "loss": 0.5674,
+      "step": 20631
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.16731508270815e-06,
+      "loss": 0.557,
+      "step": 20632
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1664585941287685e-06,
+      "loss": 0.6196,
+      "step": 20633
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1656021995849252e-06,
+      "loss": 0.5655,
+      "step": 20634
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1647458990884007e-06,
+      "loss": 0.5521,
+      "step": 20635
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1638896926509755e-06,
+      "loss": 0.532,
+      "step": 20636
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.163033580284437e-06,
+      "loss": 0.5715,
+      "step": 20637
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1621775620005647e-06,
+      "loss": 0.534,
+      "step": 20638
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.161321637811138e-06,
+      "loss": 0.629,
+      "step": 20639
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1604658077279306e-06,
+      "loss": 0.5437,
+      "step": 20640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1596100717627265e-06,
+      "loss": 0.587,
+      "step": 20641
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1587544299272985e-06,
+      "loss": 0.5844,
+      "step": 20642
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.157898882233421e-06,
+      "loss": 0.576,
+      "step": 20643
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.157043428692864e-06,
+      "loss": 0.5389,
+      "step": 20644
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.156188069317404e-06,
+      "loss": 0.577,
+      "step": 20645
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1553328041188113e-06,
+      "loss": 0.573,
+      "step": 20646
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.154477633108849e-06,
+      "loss": 0.6054,
+      "step": 20647
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.153622556299294e-06,
+      "loss": 0.6144,
+      "step": 20648
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.152767573701907e-06,
+      "loss": 0.5337,
+      "step": 20649
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1519126853284552e-06,
+      "loss": 0.5776,
+      "step": 20650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1510578911906998e-06,
+      "loss": 0.5681,
+      "step": 20651
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.150203191300407e-06,
+      "loss": 0.5108,
+      "step": 20652
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1493485856693386e-06,
+      "loss": 0.5964,
+      "step": 20653
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.148494074309253e-06,
+      "loss": 0.5824,
+      "step": 20654
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1476396572319057e-06,
+      "loss": 0.6068,
+      "step": 20655
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1467853344490618e-06,
+      "loss": 0.5775,
+      "step": 20656
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1459311059724716e-06,
+      "loss": 0.621,
+      "step": 20657
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.145076971813891e-06,
+      "loss": 0.6085,
+      "step": 20658
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1442229319850758e-06,
+      "loss": 0.5766,
+      "step": 20659
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1433689864977777e-06,
+      "loss": 0.5174,
+      "step": 20660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1425151353637472e-06,
+      "loss": 0.5659,
+      "step": 20661
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1416613785947313e-06,
+      "loss": 0.6,
+      "step": 20662
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.140807716202483e-06,
+      "loss": 0.576,
+      "step": 20663
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1399541481987484e-06,
+      "loss": 0.5619,
+      "step": 20664
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.139100674595272e-06,
+      "loss": 0.5608,
+      "step": 20665
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1382472954037967e-06,
+      "loss": 0.5733,
+      "step": 20666
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.13739401063607e-06,
+      "loss": 0.4996,
+      "step": 20667
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1365408203038326e-06,
+      "loss": 0.587,
+      "step": 20668
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.135687724418821e-06,
+      "loss": 0.5792,
+      "step": 20669
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.13483472299278e-06,
+      "loss": 0.6127,
+      "step": 20670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.133981816037447e-06,
+      "loss": 0.5492,
+      "step": 20671
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1331290035645555e-06,
+      "loss": 0.5205,
+      "step": 20672
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.13227628558584e-06,
+      "loss": 0.5472,
+      "step": 20673
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.131423662113041e-06,
+      "loss": 0.6338,
+      "step": 20674
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.130571133157888e-06,
+      "loss": 0.5617,
+      "step": 20675
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.129718698732108e-06,
+      "loss": 0.5754,
+      "step": 20676
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.12886635884744e-06,
+      "loss": 0.548,
+      "step": 20677
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1280141135156074e-06,
+      "loss": 0.5164,
+      "step": 20678
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1271619627483385e-06,
+      "loss": 0.608,
+      "step": 20679
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1263099065573576e-06,
+      "loss": 0.5766,
+      "step": 20680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1254579449543955e-06,
+      "loss": 0.6005,
+      "step": 20681
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.124606077951171e-06,
+      "loss": 0.5383,
+      "step": 20682
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.12375430555941e-06,
+      "loss": 0.5802,
+      "step": 20683
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.122902627790828e-06,
+      "loss": 0.5924,
+      "step": 20684
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.122051044657152e-06,
+      "loss": 0.5882,
+      "step": 20685
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.121199556170096e-06,
+      "loss": 0.5218,
+      "step": 20686
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1203481623413755e-06,
+      "loss": 0.549,
+      "step": 20687
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1194968631827127e-06,
+      "loss": 0.5871,
+      "step": 20688
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.118645658705819e-06,
+      "loss": 0.5881,
+      "step": 20689
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.117794548922408e-06,
+      "loss": 0.5049,
+      "step": 20690
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1169435338441878e-06,
+      "loss": 0.5793,
+      "step": 20691
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.116092613482875e-06,
+      "loss": 0.604,
+      "step": 20692
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1152417878501775e-06,
+      "loss": 0.5748,
+      "step": 20693
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.114391056957802e-06,
+      "loss": 0.5853,
+      "step": 20694
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1135404208174524e-06,
+      "loss": 0.5124,
+      "step": 20695
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.112689879440841e-06,
+      "loss": 0.6099,
+      "step": 20696
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1118394328396683e-06,
+      "loss": 0.5691,
+      "step": 20697
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.110989081025635e-06,
+      "loss": 0.676,
+      "step": 20698
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.110138824010448e-06,
+      "loss": 0.5978,
+      "step": 20699
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.109288661805804e-06,
+      "loss": 0.6052,
+      "step": 20700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.108438594423404e-06,
+      "loss": 0.5988,
+      "step": 20701
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.107588621874941e-06,
+      "loss": 0.5933,
+      "step": 20702
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1067387441721164e-06,
+      "loss": 0.5912,
+      "step": 20703
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.105888961326625e-06,
+      "loss": 0.593,
+      "step": 20704
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1050392733501587e-06,
+      "loss": 0.5015,
+      "step": 20705
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.104189680254408e-06,
+      "loss": 0.5683,
+      "step": 20706
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1033401820510687e-06,
+      "loss": 0.6126,
+      "step": 20707
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1024907787518287e-06,
+      "loss": 0.6124,
+      "step": 20708
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1016414703683737e-06,
+      "loss": 0.6482,
+      "step": 20709
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1007922569123962e-06,
+      "loss": 0.5603,
+      "step": 20710
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0999431383955802e-06,
+      "loss": 0.6144,
+      "step": 20711
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0990941148296084e-06,
+      "loss": 0.5166,
+      "step": 20712
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.098245186226162e-06,
+      "loss": 0.5793,
+      "step": 20713
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.097396352596931e-06,
+      "loss": 0.5923,
+      "step": 20714
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.09654761395359e-06,
+      "loss": 0.5556,
+      "step": 20715
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.09569897030782e-06,
+      "loss": 0.6168,
+      "step": 20716
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.094850421671295e-06,
+      "loss": 0.6146,
+      "step": 20717
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0940019680557e-06,
+      "loss": 0.5537,
+      "step": 20718
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0931536094727054e-06,
+      "loss": 0.5099,
+      "step": 20719
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0923053459339824e-06,
+      "loss": 0.5627,
+      "step": 20720
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0914571774512114e-06,
+      "loss": 0.616,
+      "step": 20721
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0906091040360596e-06,
+      "loss": 0.6158,
+      "step": 20722
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.089761125700198e-06,
+      "loss": 0.5926,
+      "step": 20723
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0889132424552914e-06,
+      "loss": 0.6093,
+      "step": 20724
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.088065454313015e-06,
+      "loss": 0.5622,
+      "step": 20725
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.087217761285032e-06,
+      "loss": 0.54,
+      "step": 20726
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0863701633830055e-06,
+      "loss": 0.5979,
+      "step": 20727
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0855226606185975e-06,
+      "loss": 0.5961,
+      "step": 20728
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0846752530034764e-06,
+      "loss": 0.5565,
+      "step": 20729
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0838279405492997e-06,
+      "loss": 0.5802,
+      "step": 20730
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0829807232677255e-06,
+      "loss": 0.5208,
+      "step": 20731
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0821336011704174e-06,
+      "loss": 0.59,
+      "step": 20732
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.081286574269029e-06,
+      "loss": 0.5885,
+      "step": 20733
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.080439642575217e-06,
+      "loss": 0.5863,
+      "step": 20734
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.079592806100632e-06,
+      "loss": 0.5907,
+      "step": 20735
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0787460648569344e-06,
+      "loss": 0.6196,
+      "step": 20736
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0778994188557722e-06,
+      "loss": 0.5852,
+      "step": 20737
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.077052868108793e-06,
+      "loss": 0.5858,
+      "step": 20738
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0762064126276525e-06,
+      "loss": 0.591,
+      "step": 20739
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0753600524239956e-06,
+      "loss": 0.589,
+      "step": 20740
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.07451378750947e-06,
+      "loss": 0.6157,
+      "step": 20741
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.073667617895716e-06,
+      "loss": 0.6254,
+      "step": 20742
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.072821543594384e-06,
+      "loss": 0.6268,
+      "step": 20743
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0719755646171146e-06,
+      "loss": 0.5918,
+      "step": 20744
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0711296809755497e-06,
+      "loss": 0.6358,
+      "step": 20745
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.070283892681325e-06,
+      "loss": 0.5276,
+      "step": 20746
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0694381997460865e-06,
+      "loss": 0.5852,
+      "step": 20747
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0685926021814673e-06,
+      "loss": 0.663,
+      "step": 20748
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.067747099999101e-06,
+      "loss": 0.5401,
+      "step": 20749
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.06690169321063e-06,
+      "loss": 0.5436,
+      "step": 20750
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0660563818276836e-06,
+      "loss": 0.5765,
+      "step": 20751
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0652111658618933e-06,
+      "loss": 0.5723,
+      "step": 20752
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.064366045324888e-06,
+      "loss": 0.6384,
+      "step": 20753
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0635210202283037e-06,
+      "loss": 0.5791,
+      "step": 20754
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0626760905837647e-06,
+      "loss": 0.6153,
+      "step": 20755
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0618312564028984e-06,
+      "loss": 0.5658,
+      "step": 20756
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0609865176973297e-06,
+      "loss": 0.576,
+      "step": 20757
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0601418744786848e-06,
+      "loss": 0.5179,
+      "step": 20758
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0592973267585844e-06,
+      "loss": 0.5224,
+      "step": 20759
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.05845287454865e-06,
+      "loss": 0.5917,
+      "step": 20760
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0576085178605055e-06,
+      "loss": 0.5309,
+      "step": 20761
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.056764256705768e-06,
+      "loss": 0.501,
+      "step": 20762
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.055920091096052e-06,
+      "loss": 0.5284,
+      "step": 20763
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0550760210429796e-06,
+      "loss": 0.5845,
+      "step": 20764
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.054232046558164e-06,
+      "loss": 0.5879,
+      "step": 20765
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0533881676532193e-06,
+      "loss": 0.5645,
+      "step": 20766
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.052544384339754e-06,
+      "loss": 0.5488,
+      "step": 20767
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0517006966293848e-06,
+      "loss": 0.6082,
+      "step": 20768
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.05085710453372e-06,
+      "loss": 0.5531,
+      "step": 20769
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0500136080643637e-06,
+      "loss": 0.5918,
+      "step": 20770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.049170207232931e-06,
+      "loss": 0.6279,
+      "step": 20771
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0483269020510232e-06,
+      "loss": 0.5941,
+      "step": 20772
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0474836925302453e-06,
+      "loss": 0.5315,
+      "step": 20773
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0466405786821984e-06,
+      "loss": 0.5938,
+      "step": 20774
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.04579756051849e-06,
+      "loss": 0.5597,
+      "step": 20775
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0449546380507166e-06,
+      "loss": 0.614,
+      "step": 20776
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.044111811290479e-06,
+      "loss": 0.5512,
+      "step": 20777
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0432690802493727e-06,
+      "loss": 0.6118,
+      "step": 20778
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0424264449389984e-06,
+      "loss": 0.5053,
+      "step": 20779
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0415839053709506e-06,
+      "loss": 0.6455,
+      "step": 20780
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0407414615568186e-06,
+      "loss": 0.6669,
+      "step": 20781
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0398991135082024e-06,
+      "loss": 0.59,
+      "step": 20782
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.039056861236691e-06,
+      "loss": 0.5455,
+      "step": 20783
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0382147047538724e-06,
+      "loss": 0.5843,
+      "step": 20784
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0373726440713335e-06,
+      "loss": 0.5752,
+      "step": 20785
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.036530679200669e-06,
+      "loss": 0.5142,
+      "step": 20786
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.035688810153461e-06,
+      "loss": 0.5398,
+      "step": 20787
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.034847036941293e-06,
+      "loss": 0.5989,
+      "step": 20788
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.034005359575748e-06,
+      "loss": 0.5973,
+      "step": 20789
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0331637780684133e-06,
+      "loss": 0.5714,
+      "step": 20790
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.032322292430866e-06,
+      "loss": 0.5362,
+      "step": 20791
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0314809026746826e-06,
+      "loss": 0.5467,
+      "step": 20792
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0306396088114486e-06,
+      "loss": 0.5969,
+      "step": 20793
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.029798410852738e-06,
+      "loss": 0.5293,
+      "step": 20794
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0289573088101255e-06,
+      "loss": 0.5956,
+      "step": 20795
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.028116302695183e-06,
+      "loss": 0.6043,
+      "step": 20796
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0272753925194886e-06,
+      "loss": 0.5601,
+      "step": 20797
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0264345782946127e-06,
+      "loss": 0.5604,
+      "step": 20798
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0255938600321234e-06,
+      "loss": 0.5923,
+      "step": 20799
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0247532377435885e-06,
+      "loss": 0.5792,
+      "step": 20800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0239127114405795e-06,
+      "loss": 0.6289,
+      "step": 20801
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0230722811346635e-06,
+      "loss": 0.5652,
+      "step": 20802
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0222319468373985e-06,
+      "loss": 0.5698,
+      "step": 20803
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.021391708560357e-06,
+      "loss": 0.6623,
+      "step": 20804
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0205515663150973e-06,
+      "loss": 0.5172,
+      "step": 20805
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.019711520113182e-06,
+      "loss": 0.6297,
+      "step": 20806
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0188715699661653e-06,
+      "loss": 0.5162,
+      "step": 20807
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0180317158856133e-06,
+      "loss": 0.586,
+      "step": 20808
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0171919578830797e-06,
+      "loss": 0.5878,
+      "step": 20809
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0163522959701207e-06,
+      "loss": 0.6171,
+      "step": 20810
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0155127301582863e-06,
+      "loss": 0.5488,
+      "step": 20811
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0146732604591377e-06,
+      "loss": 0.5125,
+      "step": 20812
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0138338868842234e-06,
+      "loss": 0.6858,
+      "step": 20813
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.01299460944509e-06,
+      "loss": 0.5331,
+      "step": 20814
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.012155428153293e-06,
+      "loss": 0.5894,
+      "step": 20815
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0113163430203775e-06,
+      "loss": 0.6452,
+      "step": 20816
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0104773540578892e-06,
+      "loss": 0.5656,
+      "step": 20817
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.009638461277372e-06,
+      "loss": 0.5537,
+      "step": 20818
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0087996646903737e-06,
+      "loss": 0.5964,
+      "step": 20819
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.007960964308436e-06,
+      "loss": 0.5319,
+      "step": 20820
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0071223601430988e-06,
+      "loss": 0.562,
+      "step": 20821
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0062838522059e-06,
+      "loss": 0.5911,
+      "step": 20822
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0054454405083834e-06,
+      "loss": 0.5483,
+      "step": 20823
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0046071250620836e-06,
+      "loss": 0.6229,
+      "step": 20824
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0037689058785325e-06,
+      "loss": 0.6135,
+      "step": 20825
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0029307829692722e-06,
+      "loss": 0.6011,
+      "step": 20826
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0020927563458325e-06,
+      "loss": 0.6226,
+      "step": 20827
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0012548260197462e-06,
+      "loss": 0.5692,
+      "step": 20828
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0004169920025385e-06,
+      "loss": 0.7008,
+      "step": 20829
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.999579254305748e-06,
+      "loss": 0.5634,
+      "step": 20830
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9987416129408964e-06,
+      "loss": 0.5771,
+      "step": 20831
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9979040679195103e-06,
+      "loss": 0.5996,
+      "step": 20832
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9970666192531195e-06,
+      "loss": 0.5657,
+      "step": 20833
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9962292669532457e-06,
+      "loss": 0.6548,
+      "step": 20834
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9953920110314106e-06,
+      "loss": 0.6008,
+      "step": 20835
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.994554851499133e-06,
+      "loss": 0.5691,
+      "step": 20836
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.99371778836794e-06,
+      "loss": 0.5563,
+      "step": 20837
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9928808216493456e-06,
+      "loss": 0.5865,
+      "step": 20838
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9920439513548693e-06,
+      "loss": 0.5381,
+      "step": 20839
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9912071774960216e-06,
+      "loss": 0.665,
+      "step": 20840
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9903705000843254e-06,
+      "loss": 0.591,
+      "step": 20841
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.98953391913129e-06,
+      "loss": 0.6168,
+      "step": 20842
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9886974346484254e-06,
+      "loss": 0.5815,
+      "step": 20843
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.987861046647247e-06,
+      "loss": 0.586,
+      "step": 20844
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.987024755139263e-06,
+      "loss": 0.6048,
+      "step": 20845
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9861885601359796e-06,
+      "loss": 0.5023,
+      "step": 20846
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9853524616489014e-06,
+      "loss": 0.572,
+      "step": 20847
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9845164596895404e-06,
+      "loss": 0.6149,
+      "step": 20848
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.983680554269397e-06,
+      "loss": 0.6726,
+      "step": 20849
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.982844745399974e-06,
+      "loss": 0.5686,
+      "step": 20850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.982009033092771e-06,
+      "loss": 0.5546,
+      "step": 20851
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.981173417359293e-06,
+      "loss": 0.622,
+      "step": 20852
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9803378982110354e-06,
+      "loss": 0.5538,
+      "step": 20853
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9795024756594937e-06,
+      "loss": 0.583,
+      "step": 20854
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9786671497161703e-06,
+      "loss": 0.5698,
+      "step": 20855
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9778319203925556e-06,
+      "loss": 0.5486,
+      "step": 20856
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.976996787700144e-06,
+      "loss": 0.6116,
+      "step": 20857
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.976161751650425e-06,
+      "loss": 0.6128,
+      "step": 20858
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.975326812254895e-06,
+      "loss": 0.6587,
+      "step": 20859
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9744919695250398e-06,
+      "loss": 0.5403,
+      "step": 20860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9736572234723483e-06,
+      "loss": 0.6348,
+      "step": 20861
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.972822574108305e-06,
+      "loss": 0.6112,
+      "step": 20862
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9719880214444e-06,
+      "loss": 0.6233,
+      "step": 20863
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.971153565492115e-06,
+      "loss": 0.5894,
+      "step": 20864
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9703192062629315e-06,
+      "loss": 0.6371,
+      "step": 20865
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9694849437683337e-06,
+      "loss": 0.5865,
+      "step": 20866
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9686507780198015e-06,
+      "loss": 0.5523,
+      "step": 20867
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9678167090288137e-06,
+      "loss": 0.5447,
+      "step": 20868
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9669827368068427e-06,
+      "loss": 0.571,
+      "step": 20869
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9661488613653723e-06,
+      "loss": 0.5744,
+      "step": 20870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.965315082715874e-06,
+      "loss": 0.6147,
+      "step": 20871
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.964481400869822e-06,
+      "loss": 0.6245,
+      "step": 20872
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.963647815838684e-06,
+      "loss": 0.6271,
+      "step": 20873
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9628143276339383e-06,
+      "loss": 0.533,
+      "step": 20874
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9619809362670513e-06,
+      "loss": 0.5862,
+      "step": 20875
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9611476417494868e-06,
+      "loss": 0.5533,
+      "step": 20876
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.960314444092719e-06,
+      "loss": 0.6223,
+      "step": 20877
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9594813433082103e-06,
+      "loss": 0.5722,
+      "step": 20878
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9586483394074237e-06,
+      "loss": 0.5854,
+      "step": 20879
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9578154324018214e-06,
+      "loss": 0.5434,
+      "step": 20880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9569826223028686e-06,
+      "loss": 0.5685,
+      "step": 20881
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9561499091220236e-06,
+      "loss": 0.5973,
+      "step": 20882
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9553172928707464e-06,
+      "loss": 0.6072,
+      "step": 20883
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.954484773560489e-06,
+      "loss": 0.5681,
+      "step": 20884
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9536523512027147e-06,
+      "loss": 0.5616,
+      "step": 20885
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.952820025808876e-06,
+      "loss": 0.6121,
+      "step": 20886
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.951987797390422e-06,
+      "loss": 0.5961,
+      "step": 20887
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9511556659588127e-06,
+      "loss": 0.561,
+      "step": 20888
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9503236315254956e-06,
+      "loss": 0.6427,
+      "step": 20889
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9494916941019193e-06,
+      "loss": 0.597,
+      "step": 20890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9486598536995283e-06,
+      "loss": 0.5778,
+      "step": 20891
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.947828110329778e-06,
+      "loss": 0.6075,
+      "step": 20892
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9469964640041083e-06,
+      "loss": 0.5345,
+      "step": 20893
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9461649147339624e-06,
+      "loss": 0.608,
+      "step": 20894
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.945333462530788e-06,
+      "loss": 0.5331,
+      "step": 20895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9445021074060245e-06,
+      "loss": 0.5924,
+      "step": 20896
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.943670849371111e-06,
+      "loss": 0.599,
+      "step": 20897
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.942839688437483e-06,
+      "loss": 0.581,
+      "step": 20898
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.942008624616587e-06,
+      "loss": 0.5817,
+      "step": 20899
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9411776579198524e-06,
+      "loss": 0.541,
+      "step": 20900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9403467883587154e-06,
+      "loss": 0.5854,
+      "step": 20901
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.939516015944608e-06,
+      "loss": 0.5878,
+      "step": 20902
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.938685340688966e-06,
+      "loss": 0.5731,
+      "step": 20903
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9378547626032195e-06,
+      "loss": 0.5771,
+      "step": 20904
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9370242816987927e-06,
+      "loss": 0.593,
+      "step": 20905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9361938979871218e-06,
+      "loss": 0.6282,
+      "step": 20906
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.935363611479629e-06,
+      "loss": 0.5578,
+      "step": 20907
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.934533422187741e-06,
+      "loss": 0.5556,
+      "step": 20908
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.933703330122877e-06,
+      "loss": 0.5534,
+      "step": 20909
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.932873335296469e-06,
+      "loss": 0.5441,
+      "step": 20910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9320434377199327e-06,
+      "loss": 0.5876,
+      "step": 20911
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9312136374046884e-06,
+      "loss": 0.556,
+      "step": 20912
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.930383934362153e-06,
+      "loss": 0.5731,
+      "step": 20913
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.929554328603751e-06,
+      "loss": 0.5399,
+      "step": 20914
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9287248201408923e-06,
+      "loss": 0.5736,
+      "step": 20915
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.927895408984992e-06,
+      "loss": 0.6519,
+      "step": 20916
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.927066095147467e-06,
+      "loss": 0.5451,
+      "step": 20917
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9262368786397278e-06,
+      "loss": 0.5509,
+      "step": 20918
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9254077594731846e-06,
+      "loss": 0.5547,
+      "step": 20919
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9245787376592425e-06,
+      "loss": 0.5946,
+      "step": 20920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.923749813209319e-06,
+      "loss": 0.6003,
+      "step": 20921
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.922920986134815e-06,
+      "loss": 0.5488,
+      "step": 20922
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9220922564471366e-06,
+      "loss": 0.5251,
+      "step": 20923
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9212636241576854e-06,
+      "loss": 0.5678,
+      "step": 20924
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.920435089277869e-06,
+      "loss": 0.553,
+      "step": 20925
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.919606651819087e-06,
+      "loss": 0.5913,
+      "step": 20926
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.918778311792736e-06,
+      "loss": 0.6792,
+      "step": 20927
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.917950069210219e-06,
+      "loss": 0.5296,
+      "step": 20928
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.917121924082934e-06,
+      "loss": 0.6108,
+      "step": 20929
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9162938764222736e-06,
+      "loss": 0.5655,
+      "step": 20930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9154659262396313e-06,
+      "loss": 0.593,
+      "step": 20931
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9146380735464052e-06,
+      "loss": 0.5055,
+      "step": 20932
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9138103183539856e-06,
+      "loss": 0.5893,
+      "step": 20933
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9129826606737623e-06,
+      "loss": 0.6176,
+      "step": 20934
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.912155100517122e-06,
+      "loss": 0.6253,
+      "step": 20935
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.911327637895458e-06,
+      "loss": 0.5804,
+      "step": 20936
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.910500272820155e-06,
+      "loss": 0.5897,
+      "step": 20937
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9096730053025933e-06,
+      "loss": 0.5606,
+      "step": 20938
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9088458353541647e-06,
+      "loss": 0.5382,
+      "step": 20939
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9080187629862488e-06,
+      "loss": 0.5665,
+      "step": 20940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.907191788210226e-06,
+      "loss": 0.5918,
+      "step": 20941
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9063649110374737e-06,
+      "loss": 0.5688,
+      "step": 20942
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.905538131479376e-06,
+      "loss": 0.6188,
+      "step": 20943
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9047114495473085e-06,
+      "loss": 0.5668,
+      "step": 20944
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.903884865252644e-06,
+      "loss": 0.5962,
+      "step": 20945
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9030583786067566e-06,
+      "loss": 0.6157,
+      "step": 20946
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.902231989621026e-06,
+      "loss": 0.5448,
+      "step": 20947
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.901405698306818e-06,
+      "loss": 0.5731,
+      "step": 20948
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9005795046755037e-06,
+      "loss": 0.543,
+      "step": 20949
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.899753408738455e-06,
+      "loss": 0.5222,
+      "step": 20950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.89892741050704e-06,
+      "loss": 0.5968,
+      "step": 20951
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.898101509992621e-06,
+      "loss": 0.6277,
+      "step": 20952
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8972757072065637e-06,
+      "loss": 0.5869,
+      "step": 20953
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8964500021602373e-06,
+      "loss": 0.6253,
+      "step": 20954
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.895624394865e-06,
+      "loss": 0.5695,
+      "step": 20955
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.894798885332211e-06,
+      "loss": 0.5855,
+      "step": 20956
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.893973473573235e-06,
+      "loss": 0.5424,
+      "step": 20957
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.893148159599427e-06,
+      "loss": 0.5873,
+      "step": 20958
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8923229434221465e-06,
+      "loss": 0.5724,
+      "step": 20959
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8914978250527437e-06,
+      "loss": 0.5952,
+      "step": 20960
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.89067280450258e-06,
+      "loss": 0.6598,
+      "step": 20961
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8898478817830055e-06,
+      "loss": 0.563,
+      "step": 20962
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8890230569053723e-06,
+      "loss": 0.5811,
+      "step": 20963
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.888198329881027e-06,
+      "loss": 0.6783,
+      "step": 20964
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.887373700721323e-06,
+      "loss": 0.571,
+      "step": 20965
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.886549169437608e-06,
+      "loss": 0.5949,
+      "step": 20966
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8857247360412244e-06,
+      "loss": 0.6016,
+      "step": 20967
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8849004005435244e-06,
+      "loss": 0.5517,
+      "step": 20968
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8840761629558447e-06,
+      "loss": 0.5879,
+      "step": 20969
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8832520232895257e-06,
+      "loss": 0.6196,
+      "step": 20970
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8824279815559152e-06,
+      "loss": 0.5419,
+      "step": 20971
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8816040377663503e-06,
+      "loss": 0.5518,
+      "step": 20972
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.880780191932169e-06,
+      "loss": 0.602,
+      "step": 20973
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.879956444064703e-06,
+      "loss": 0.5684,
+      "step": 20974
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8791327941752967e-06,
+      "loss": 0.5945,
+      "step": 20975
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.87830924227528e-06,
+      "loss": 0.589,
+      "step": 20976
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8774857883759866e-06,
+      "loss": 0.5975,
+      "step": 20977
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8766624324887427e-06,
+      "loss": 0.5713,
+      "step": 20978
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.875839174624887e-06,
+      "loss": 0.5937,
+      "step": 20979
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.875016014795744e-06,
+      "loss": 0.561,
+      "step": 20980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.874192953012638e-06,
+      "loss": 0.6438,
+      "step": 20981
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8733699892869004e-06,
+      "loss": 0.5993,
+      "step": 20982
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.872547123629854e-06,
+      "loss": 0.6081,
+      "step": 20983
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.871724356052823e-06,
+      "loss": 0.6599,
+      "step": 20984
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8709016865671244e-06,
+      "loss": 0.5458,
+      "step": 20985
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8700791151840857e-06,
+      "loss": 0.6084,
+      "step": 20986
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.869256641915024e-06,
+      "loss": 0.5599,
+      "step": 20987
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.868434266771256e-06,
+      "loss": 0.5948,
+      "step": 20988
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.867611989764096e-06,
+      "loss": 0.6105,
+      "step": 20989
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8667898109048653e-06,
+      "loss": 0.5952,
+      "step": 20990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8659677302048738e-06,
+      "loss": 0.5354,
+      "step": 20991
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.865145747675433e-06,
+      "loss": 0.5392,
+      "step": 20992
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.864323863327858e-06,
+      "loss": 0.6613,
+      "step": 20993
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8635020771734567e-06,
+      "loss": 0.6833,
+      "step": 20994
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8626803892235375e-06,
+      "loss": 0.5664,
+      "step": 20995
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8618587994894053e-06,
+      "loss": 0.5737,
+      "step": 20996
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8610373079823707e-06,
+      "loss": 0.5126,
+      "step": 20997
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8602159147137355e-06,
+      "loss": 0.6459,
+      "step": 20998
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8593946196947997e-06,
+      "loss": 0.5843,
+      "step": 20999
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.858573422936871e-06,
+      "loss": 0.582,
+      "step": 21000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8577523244512473e-06,
+      "loss": 0.5717,
+      "step": 21001
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.856931324249228e-06,
+      "loss": 0.6006,
+      "step": 21002
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8561104223421056e-06,
+      "loss": 0.5517,
+      "step": 21003
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.855289618741185e-06,
+      "loss": 0.5571,
+      "step": 21004
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8544689134577575e-06,
+      "loss": 0.5907,
+      "step": 21005
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.853648306503116e-06,
+      "loss": 0.5678,
+      "step": 21006
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.85282779788855e-06,
+      "loss": 0.5387,
+      "step": 21007
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.852007387625356e-06,
+      "loss": 0.5547,
+      "step": 21008
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8511870757248215e-06,
+      "loss": 0.594,
+      "step": 21009
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8503668621982317e-06,
+      "loss": 0.5502,
+      "step": 21010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8495467470568794e-06,
+      "loss": 0.6834,
+      "step": 21011
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.848726730312046e-06,
+      "loss": 0.5832,
+      "step": 21012
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8479068119750176e-06,
+      "loss": 0.6126,
+      "step": 21013
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.847086992057071e-06,
+      "loss": 0.5894,
+      "step": 21014
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8462672705694973e-06,
+      "loss": 0.6055,
+      "step": 21015
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8454476475235706e-06,
+      "loss": 0.566,
+      "step": 21016
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8446281229305716e-06,
+      "loss": 0.5547,
+      "step": 21017
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.843808696801773e-06,
+      "loss": 0.5438,
+      "step": 21018
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.842989369148459e-06,
+      "loss": 0.6215,
+      "step": 21019
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.842170139981899e-06,
+      "loss": 0.5788,
+      "step": 21020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8413510093133646e-06,
+      "loss": 0.5854,
+      "step": 21021
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.840531977154133e-06,
+      "loss": 0.6007,
+      "step": 21022
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.839713043515473e-06,
+      "loss": 0.5696,
+      "step": 21023
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8388942084086525e-06,
+      "loss": 0.5288,
+      "step": 21024
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8380754718449366e-06,
+      "loss": 0.6123,
+      "step": 21025
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8372568338355988e-06,
+      "loss": 0.5366,
+      "step": 21026
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8364382943919012e-06,
+      "loss": 0.5405,
+      "step": 21027
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8356198535251066e-06,
+      "loss": 0.5216,
+      "step": 21028
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8348015112464743e-06,
+      "loss": 0.562,
+      "step": 21029
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.833983267567273e-06,
+      "loss": 0.6456,
+      "step": 21030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.833165122498759e-06,
+      "loss": 0.5502,
+      "step": 21031
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8323470760521866e-06,
+      "loss": 0.6054,
+      "step": 21032
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.831529128238819e-06,
+      "loss": 0.6455,
+      "step": 21033
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8307112790699097e-06,
+      "loss": 0.5994,
+      "step": 21034
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.829893528556713e-06,
+      "loss": 0.5861,
+      "step": 21035
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.829075876710479e-06,
+      "loss": 0.6098,
+      "step": 21036
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8282583235424643e-06,
+      "loss": 0.5232,
+      "step": 21037
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8274408690639164e-06,
+      "loss": 0.5676,
+      "step": 21038
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.826623513286085e-06,
+      "loss": 0.533,
+      "step": 21039
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8258062562202137e-06,
+      "loss": 0.5281,
+      "step": 21040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.824989097877555e-06,
+      "loss": 0.5484,
+      "step": 21041
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8241720382693514e-06,
+      "loss": 0.6582,
+      "step": 21042
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.823355077406843e-06,
+      "loss": 0.6172,
+      "step": 21043
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8225382153012772e-06,
+      "loss": 0.5553,
+      "step": 21044
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8217214519638926e-06,
+      "loss": 0.5763,
+      "step": 21045
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.820904787405927e-06,
+      "loss": 0.5971,
+      "step": 21046
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8200882216386182e-06,
+      "loss": 0.6111,
+      "step": 21047
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.819271754673206e-06,
+      "loss": 0.5544,
+      "step": 21048
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.818455386520925e-06,
+      "loss": 0.5626,
+      "step": 21049
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8176391171930073e-06,
+      "loss": 0.5525,
+      "step": 21050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8168229467006846e-06,
+      "loss": 0.5449,
+      "step": 21051
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8160068750551917e-06,
+      "loss": 0.5794,
+      "step": 21052
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.815190902267757e-06,
+      "loss": 0.6039,
+      "step": 21053
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8143750283496054e-06,
+      "loss": 0.5671,
+      "step": 21054
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8135592533119706e-06,
+      "loss": 0.5489,
+      "step": 21055
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.812743577166075e-06,
+      "loss": 0.5798,
+      "step": 21056
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.811927999923143e-06,
+      "loss": 0.5261,
+      "step": 21057
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8111125215943944e-06,
+      "loss": 0.6415,
+      "step": 21058
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8102971421910576e-06,
+      "loss": 0.561,
+      "step": 21059
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8094818617243492e-06,
+      "loss": 0.5987,
+      "step": 21060
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8086666802054853e-06,
+      "loss": 0.5815,
+      "step": 21061
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8078515976456897e-06,
+      "loss": 0.553,
+      "step": 21062
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8070366140561765e-06,
+      "loss": 0.5707,
+      "step": 21063
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.80622172944816e-06,
+      "loss": 0.521,
+      "step": 21064
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.805406943832849e-06,
+      "loss": 0.6602,
+      "step": 21065
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8045922572214644e-06,
+      "loss": 0.5574,
+      "step": 21066
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8037776696252127e-06,
+      "loss": 0.5452,
+      "step": 21067
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8029631810553026e-06,
+      "loss": 0.5695,
+      "step": 21068
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8021487915229407e-06,
+      "loss": 0.6127,
+      "step": 21069
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.801334501039339e-06,
+      "loss": 0.5658,
+      "step": 21070
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8005203096156998e-06,
+      "loss": 0.6146,
+      "step": 21071
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.799706217263224e-06,
+      "loss": 0.5765,
+      "step": 21072
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7988922239931206e-06,
+      "loss": 0.5701,
+      "step": 21073
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.798078329816588e-06,
+      "loss": 0.5713,
+      "step": 21074
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7972645347448246e-06,
+      "loss": 0.5474,
+      "step": 21075
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7964508387890277e-06,
+      "loss": 0.5352,
+      "step": 21076
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7956372419604005e-06,
+      "loss": 0.5907,
+      "step": 21077
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.794823744270134e-06,
+      "loss": 0.6041,
+      "step": 21078
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7940103457294254e-06,
+      "loss": 0.6483,
+      "step": 21079
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.793197046349462e-06,
+      "loss": 0.542,
+      "step": 21080
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.792383846141442e-06,
+      "loss": 0.5614,
+      "step": 21081
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7915707451165553e-06,
+      "loss": 0.5872,
+      "step": 21082
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7907577432859843e-06,
+      "loss": 0.531,
+      "step": 21083
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.789944840660925e-06,
+      "loss": 0.5755,
+      "step": 21084
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.789132037252561e-06,
+      "loss": 0.5827,
+      "step": 21085
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7883193330720747e-06,
+      "loss": 0.5951,
+      "step": 21086
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7875067281306487e-06,
+      "loss": 0.4988,
+      "step": 21087
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.786694222439471e-06,
+      "loss": 0.5831,
+      "step": 21088
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.785881816009718e-06,
+      "loss": 0.5538,
+      "step": 21089
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.78506950885257e-06,
+      "loss": 0.5779,
+      "step": 21090
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.784257300979204e-06,
+      "loss": 0.5695,
+      "step": 21091
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.783445192400799e-06,
+      "loss": 0.627,
+      "step": 21092
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.78263318312853e-06,
+      "loss": 0.5172,
+      "step": 21093
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7818212731735674e-06,
+      "loss": 0.5456,
+      "step": 21094
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7810094625470884e-06,
+      "loss": 0.5487,
+      "step": 21095
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.780197751260263e-06,
+      "loss": 0.5572,
+      "step": 21096
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7793861393242605e-06,
+      "loss": 0.5704,
+      "step": 21097
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7785746267502468e-06,
+      "loss": 0.6444,
+      "step": 21098
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.777763213549394e-06,
+      "loss": 0.5463,
+      "step": 21099
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7769518997328647e-06,
+      "loss": 0.5789,
+      "step": 21100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7761406853118242e-06,
+      "loss": 0.5144,
+      "step": 21101
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7753295702974338e-06,
+      "loss": 0.5392,
+      "step": 21102
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.774518554700858e-06,
+      "loss": 0.5397,
+      "step": 21103
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.773707638533256e-06,
+      "loss": 0.5739,
+      "step": 21104
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.772896821805784e-06,
+      "loss": 0.5757,
+      "step": 21105
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7720861045296053e-06,
+      "loss": 0.5719,
+      "step": 21106
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.771275486715872e-06,
+      "loss": 0.5094,
+      "step": 21107
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.77046496837574e-06,
+      "loss": 0.5179,
+      "step": 21108
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7696545495203596e-06,
+      "loss": 0.6514,
+      "step": 21109
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7688442301608886e-06,
+      "loss": 0.5823,
+      "step": 21110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7680340103084758e-06,
+      "loss": 0.5662,
+      "step": 21111
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7672238899742676e-06,
+      "loss": 0.5799,
+      "step": 21112
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7664138691694122e-06,
+      "loss": 0.5627,
+      "step": 21113
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7656039479050613e-06,
+      "loss": 0.6087,
+      "step": 21114
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.764794126192356e-06,
+      "loss": 0.5776,
+      "step": 21115
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.763984404042438e-06,
+      "loss": 0.5639,
+      "step": 21116
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7631747814664568e-06,
+      "loss": 0.5148,
+      "step": 21117
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.762365258475548e-06,
+      "loss": 0.6145,
+      "step": 21118
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7615558350808535e-06,
+      "loss": 0.5203,
+      "step": 21119
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7607465112935074e-06,
+      "loss": 0.6216,
+      "step": 21120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7599372871246543e-06,
+      "loss": 0.6114,
+      "step": 21121
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.759128162585425e-06,
+      "loss": 0.5745,
+      "step": 21122
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.758319137686951e-06,
+      "loss": 0.5652,
+      "step": 21123
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.757510212440373e-06,
+      "loss": 0.6572,
+      "step": 21124
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7567013868568182e-06,
+      "loss": 0.6163,
+      "step": 21125
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.755892660947416e-06,
+      "loss": 0.5707,
+      "step": 21126
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7550840347232933e-06,
+      "loss": 0.6362,
+      "step": 21127
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7542755081955843e-06,
+      "loss": 0.5654,
+      "step": 21128
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7534670813754096e-06,
+      "loss": 0.6262,
+      "step": 21129
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7526587542738957e-06,
+      "loss": 0.5872,
+      "step": 21130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7518505269021623e-06,
+      "loss": 0.6063,
+      "step": 21131
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7510423992713374e-06,
+      "loss": 0.5552,
+      "step": 21132
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.750234371392538e-06,
+      "loss": 0.6054,
+      "step": 21133
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7494264432768826e-06,
+      "loss": 0.5297,
+      "step": 21134
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.748618614935492e-06,
+      "loss": 0.5486,
+      "step": 21135
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7478108863794806e-06,
+      "loss": 0.5559,
+      "step": 21136
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7470032576199646e-06,
+      "loss": 0.5887,
+      "step": 21137
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.746195728668053e-06,
+      "loss": 0.5978,
+      "step": 21138
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7453882995348657e-06,
+      "loss": 0.5141,
+      "step": 21139
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7445809702315095e-06,
+      "loss": 0.5473,
+      "step": 21140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7437737407690944e-06,
+      "loss": 0.5688,
+      "step": 21141
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.742966611158725e-06,
+      "loss": 0.5805,
+      "step": 21142
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7421595814115153e-06,
+      "loss": 0.5072,
+      "step": 21143
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.741352651538567e-06,
+      "loss": 0.5864,
+      "step": 21144
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7405458215509805e-06,
+      "loss": 0.6247,
+      "step": 21145
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7397390914598663e-06,
+      "loss": 0.5925,
+      "step": 21146
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7389324612763213e-06,
+      "loss": 0.6285,
+      "step": 21147
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.738125931011445e-06,
+      "loss": 0.522,
+      "step": 21148
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7373195006763343e-06,
+      "loss": 0.5318,
+      "step": 21149
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7365131702820913e-06,
+      "loss": 0.5572,
+      "step": 21150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7357069398398105e-06,
+      "loss": 0.6289,
+      "step": 21151
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7349008093605846e-06,
+      "loss": 0.5314,
+      "step": 21152
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.734094778855504e-06,
+      "loss": 0.5812,
+      "step": 21153
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.733288848335667e-06,
+      "loss": 0.5506,
+      "step": 21154
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7324830178121595e-06,
+      "loss": 0.6428,
+      "step": 21155
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7316772872960697e-06,
+      "loss": 0.5497,
+      "step": 21156
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7308716567984896e-06,
+      "loss": 0.5966,
+      "step": 21157
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.730066126330503e-06,
+      "loss": 0.5858,
+      "step": 21158
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.729260695903193e-06,
+      "loss": 0.5956,
+      "step": 21159
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7284553655276426e-06,
+      "loss": 0.6022,
+      "step": 21160
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7276501352149386e-06,
+      "loss": 0.614,
+      "step": 21161
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7268450049761585e-06,
+      "loss": 0.6715,
+      "step": 21162
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7260399748223832e-06,
+      "loss": 0.5875,
+      "step": 21163
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7252350447646847e-06,
+      "loss": 0.5547,
+      "step": 21164
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7244302148141476e-06,
+      "loss": 0.5035,
+      "step": 21165
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7236254849818443e-06,
+      "loss": 0.6558,
+      "step": 21166
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7228208552788436e-06,
+      "loss": 0.5633,
+      "step": 21167
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7220163257162267e-06,
+      "loss": 0.539,
+      "step": 21168
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.721211896305059e-06,
+      "loss": 0.601,
+      "step": 21169
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7204075670564122e-06,
+      "loss": 0.5854,
+      "step": 21170
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7196033379813503e-06,
+      "loss": 0.6374,
+      "step": 21171
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7187992090909465e-06,
+      "loss": 0.5778,
+      "step": 21172
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7179951803962636e-06,
+      "loss": 0.5478,
+      "step": 21173
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.717191251908362e-06,
+      "loss": 0.5896,
+      "step": 21174
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7163874236383115e-06,
+      "loss": 0.5787,
+      "step": 21175
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7155836955971694e-06,
+      "loss": 0.6534,
+      "step": 21176
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7147800677959967e-06,
+      "loss": 0.5686,
+      "step": 21177
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7139765402458494e-06,
+      "loss": 0.5744,
+      "step": 21178
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7131731129577922e-06,
+      "loss": 0.5902,
+      "step": 21179
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7123697859428722e-06,
+      "loss": 0.5784,
+      "step": 21180
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7115665592121453e-06,
+      "loss": 0.5768,
+      "step": 21181
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7107634327766698e-06,
+      "loss": 0.5637,
+      "step": 21182
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7099604066474936e-06,
+      "loss": 0.6325,
+      "step": 21183
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.709157480835668e-06,
+      "loss": 0.6187,
+      "step": 21184
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7083546553522388e-06,
+      "loss": 0.5804,
+      "step": 21185
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7075519302082596e-06,
+      "loss": 0.6404,
+      "step": 21186
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.706749305414774e-06,
+      "loss": 0.5459,
+      "step": 21187
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7059467809828234e-06,
+      "loss": 0.5796,
+      "step": 21188
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.705144356923457e-06,
+      "loss": 0.5824,
+      "step": 21189
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.704342033247713e-06,
+      "loss": 0.5824,
+      "step": 21190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7035398099666343e-06,
+      "loss": 0.5627,
+      "step": 21191
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.702737687091256e-06,
+      "loss": 0.5101,
+      "step": 21192
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7019356646326223e-06,
+      "loss": 0.6375,
+      "step": 21193
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7011337426017657e-06,
+      "loss": 0.5679,
+      "step": 21194
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.700331921009722e-06,
+      "loss": 0.5794,
+      "step": 21195
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.699530199867523e-06,
+      "loss": 0.6192,
+      "step": 21196
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6987285791862063e-06,
+      "loss": 0.5769,
+      "step": 21197
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6979270589767994e-06,
+      "loss": 0.5887,
+      "step": 21198
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6971256392503286e-06,
+      "loss": 0.5846,
+      "step": 21199
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6963243200178292e-06,
+      "loss": 0.5845,
+      "step": 21200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.695523101290325e-06,
+      "loss": 0.6082,
+      "step": 21201
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.694721983078842e-06,
+      "loss": 0.6461,
+      "step": 21202
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6939209653944e-06,
+      "loss": 0.5795,
+      "step": 21203
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.693120048248027e-06,
+      "loss": 0.5596,
+      "step": 21204
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.692319231650744e-06,
+      "loss": 0.5108,
+      "step": 21205
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6915185156135694e-06,
+      "loss": 0.5856,
+      "step": 21206
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.690717900147518e-06,
+      "loss": 0.5865,
+      "step": 21207
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.689917385263614e-06,
+      "loss": 0.5872,
+      "step": 21208
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6891169709728704e-06,
+      "loss": 0.5869,
+      "step": 21209
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6883166572862975e-06,
+      "loss": 0.6006,
+      "step": 21210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6875164442149147e-06,
+      "loss": 0.5147,
+      "step": 21211
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.686716331769732e-06,
+      "loss": 0.5139,
+      "step": 21212
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.685916319961759e-06,
+      "loss": 0.633,
+      "step": 21213
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6851164088019997e-06,
+      "loss": 0.5405,
+      "step": 21214
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6843165983014698e-06,
+      "loss": 0.558,
+      "step": 21215
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6835168884711715e-06,
+      "loss": 0.6274,
+      "step": 21216
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.682717279322107e-06,
+      "loss": 0.5611,
+      "step": 21217
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6819177708652844e-06,
+      "loss": 0.5447,
+      "step": 21218
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6811183631117044e-06,
+      "loss": 0.5789,
+      "step": 21219
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6803190560723668e-06,
+      "loss": 0.6508,
+      "step": 21220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6795198497582663e-06,
+      "loss": 0.595,
+      "step": 21221
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6787207441804097e-06,
+      "loss": 0.5939,
+      "step": 21222
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.677921739349787e-06,
+      "loss": 0.5108,
+      "step": 21223
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6771228352773958e-06,
+      "loss": 0.5828,
+      "step": 21224
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6763240319742257e-06,
+      "loss": 0.5482,
+      "step": 21225
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.675525329451275e-06,
+      "loss": 0.5954,
+      "step": 21226
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.674726727719532e-06,
+      "loss": 0.6192,
+      "step": 21227
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6739282267899824e-06,
+      "loss": 0.5589,
+      "step": 21228
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6731298266736214e-06,
+      "loss": 0.6036,
+      "step": 21229
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.672331527381431e-06,
+      "loss": 0.5528,
+      "step": 21230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6715333289243983e-06,
+      "loss": 0.5515,
+      "step": 21231
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6707352313135027e-06,
+      "loss": 0.6216,
+      "step": 21232
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.669937234559733e-06,
+      "loss": 0.5502,
+      "step": 21233
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.669139338674068e-06,
+      "loss": 0.5835,
+      "step": 21234
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6683415436674873e-06,
+      "loss": 0.5666,
+      "step": 21235
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6675438495509666e-06,
+      "loss": 0.5902,
+      "step": 21236
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6667462563354885e-06,
+      "loss": 0.5359,
+      "step": 21237
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6659487640320247e-06,
+      "loss": 0.5582,
+      "step": 21238
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6651513726515468e-06,
+      "loss": 0.6173,
+      "step": 21239
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6643540822050353e-06,
+      "loss": 0.5522,
+      "step": 21240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.663556892703456e-06,
+      "loss": 0.6117,
+      "step": 21241
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6627598041577807e-06,
+      "loss": 0.5906,
+      "step": 21242
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6619628165789744e-06,
+      "loss": 0.5777,
+      "step": 21243
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6611659299780113e-06,
+      "loss": 0.5584,
+      "step": 21244
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.660369144365853e-06,
+      "loss": 0.5428,
+      "step": 21245
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6595724597534642e-06,
+      "loss": 0.5984,
+      "step": 21246
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6587758761518057e-06,
+      "loss": 0.5773,
+      "step": 21247
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6579793935718445e-06,
+      "loss": 0.5728,
+      "step": 21248
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.657183012024538e-06,
+      "loss": 0.5343,
+      "step": 21249
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6563867315208413e-06,
+      "loss": 0.5772,
+      "step": 21250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6555905520717205e-06,
+      "loss": 0.5745,
+      "step": 21251
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.654794473688126e-06,
+      "loss": 0.5551,
+      "step": 21252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.653998496381014e-06,
+      "loss": 0.5291,
+      "step": 21253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.653202620161335e-06,
+      "loss": 0.5646,
+      "step": 21254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.652406845040045e-06,
+      "loss": 0.6396,
+      "step": 21255
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6516111710280932e-06,
+      "loss": 0.5706,
+      "step": 21256
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6508155981364305e-06,
+      "loss": 0.5492,
+      "step": 21257
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.650020126375997e-06,
+      "loss": 0.6138,
+      "step": 21258
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.64922475575775e-06,
+      "loss": 0.5542,
+      "step": 21259
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6484294862926286e-06,
+      "loss": 0.5844,
+      "step": 21260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.647634317991573e-06,
+      "loss": 0.5679,
+      "step": 21261
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.646839250865534e-06,
+      "loss": 0.5872,
+      "step": 21262
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6460442849254473e-06,
+      "loss": 0.5772,
+      "step": 21263
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6452494201822533e-06,
+      "loss": 0.5364,
+      "step": 21264
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6444546566468864e-06,
+      "loss": 0.6224,
+      "step": 21265
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.64365999433029e-06,
+      "loss": 0.5775,
+      "step": 21266
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6428654332433958e-06,
+      "loss": 0.542,
+      "step": 21267
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6420709733971375e-06,
+      "loss": 0.6347,
+      "step": 21268
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.641276614802445e-06,
+      "loss": 0.632,
+      "step": 21269
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6404823574702556e-06,
+      "loss": 0.5437,
+      "step": 21270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6396882014114956e-06,
+      "loss": 0.5527,
+      "step": 21271
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.63889414663709e-06,
+      "loss": 0.5404,
+      "step": 21272
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.638100193157972e-06,
+      "loss": 0.5764,
+      "step": 21273
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.637306340985063e-06,
+      "loss": 0.6241,
+      "step": 21274
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6365125901292896e-06,
+      "loss": 0.63,
+      "step": 21275
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6357189406015693e-06,
+      "loss": 0.5738,
+      "step": 21276
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6349253924128313e-06,
+      "loss": 0.5976,
+      "step": 21277
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.63413194557399e-06,
+      "loss": 0.6385,
+      "step": 21278
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.633338600095964e-06,
+      "loss": 0.608,
+      "step": 21279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.632545355989673e-06,
+      "loss": 0.5774,
+      "step": 21280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6317522132660324e-06,
+      "loss": 0.5438,
+      "step": 21281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6309591719359563e-06,
+      "loss": 0.5829,
+      "step": 21282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6301662320103526e-06,
+      "loss": 0.515,
+      "step": 21283
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.629373393500141e-06,
+      "loss": 0.5235,
+      "step": 21284
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6285806564162276e-06,
+      "loss": 0.5668,
+      "step": 21285
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6277880207695216e-06,
+      "loss": 0.5844,
+      "step": 21286
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.626995486570928e-06,
+      "loss": 0.586,
+      "step": 21287
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6262030538313564e-06,
+      "loss": 0.6289,
+      "step": 21288
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.625410722561712e-06,
+      "loss": 0.597,
+      "step": 21289
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6246184927728913e-06,
+      "loss": 0.6099,
+      "step": 21290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6238263644758043e-06,
+      "loss": 0.6028,
+      "step": 21291
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.623034337681347e-06,
+      "loss": 0.5406,
+      "step": 21292
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.622242412400421e-06,
+      "loss": 0.6049,
+      "step": 21293
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.621450588643918e-06,
+      "loss": 0.6109,
+      "step": 21294
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6206588664227404e-06,
+      "loss": 0.5132,
+      "step": 21295
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6198672457477825e-06,
+      "loss": 0.5222,
+      "step": 21296
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.619075726629935e-06,
+      "loss": 0.6188,
+      "step": 21297
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.618284309080088e-06,
+      "loss": 0.6111,
+      "step": 21298
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6174929931091387e-06,
+      "loss": 0.5642,
+      "step": 21299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6167017787279725e-06,
+      "loss": 0.5708,
+      "step": 21300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.615910665947473e-06,
+      "loss": 0.6317,
+      "step": 21301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.615119654778535e-06,
+      "loss": 0.5694,
+      "step": 21302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6143287452320397e-06,
+      "loss": 0.617,
+      "step": 21303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6135379373188706e-06,
+      "loss": 0.5831,
+      "step": 21304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.612747231049907e-06,
+      "loss": 0.5843,
+      "step": 21305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.611956626436034e-06,
+      "loss": 0.6494,
+      "step": 21306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6111661234881312e-06,
+      "loss": 0.6249,
+      "step": 21307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6103757222170745e-06,
+      "loss": 0.5565,
+      "step": 21308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6095854226337367e-06,
+      "loss": 0.5788,
+      "step": 21309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6087952247490024e-06,
+      "loss": 0.5932,
+      "step": 21310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6080051285737384e-06,
+      "loss": 0.5521,
+      "step": 21311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.607215134118817e-06,
+      "loss": 0.5943,
+      "step": 21312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.606425241395114e-06,
+      "loss": 0.648,
+      "step": 21313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6056354504134962e-06,
+      "loss": 0.5985,
+      "step": 21314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6048457611848322e-06,
+      "loss": 0.6128,
+      "step": 21315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6040561737199855e-06,
+      "loss": 0.5348,
+      "step": 21316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.603266688029826e-06,
+      "loss": 0.5894,
+      "step": 21317
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.602477304125218e-06,
+      "loss": 0.567,
+      "step": 21318
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6016880220170215e-06,
+      "loss": 0.5799,
+      "step": 21319
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6008988417160964e-06,
+      "loss": 0.5969,
+      "step": 21320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.600109763233307e-06,
+      "loss": 0.5188,
+      "step": 21321
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.599320786579509e-06,
+      "loss": 0.5845,
+      "step": 21322
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5985319117655573e-06,
+      "loss": 0.5418,
+      "step": 21323
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.597743138802313e-06,
+      "loss": 0.6001,
+      "step": 21324
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5969544677006275e-06,
+      "loss": 0.5714,
+      "step": 21325
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5961658984713522e-06,
+      "loss": 0.6119,
+      "step": 21326
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5953774311253377e-06,
+      "loss": 0.6213,
+      "step": 21327
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5945890656734396e-06,
+      "loss": 0.5934,
+      "step": 21328
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5938008021265016e-06,
+      "loss": 0.5318,
+      "step": 21329
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5930126404953738e-06,
+      "loss": 0.5764,
+      "step": 21330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5922245807908964e-06,
+      "loss": 0.6211,
+      "step": 21331
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.591436623023921e-06,
+      "loss": 0.5722,
+      "step": 21332
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.590648767205287e-06,
+      "loss": 0.5858,
+      "step": 21333
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5898610133458337e-06,
+      "loss": 0.541,
+      "step": 21334
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.589073361456407e-06,
+      "loss": 0.6284,
+      "step": 21335
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5882858115478424e-06,
+      "loss": 0.5756,
+      "step": 21336
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.587498363630978e-06,
+      "loss": 0.6199,
+      "step": 21337
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5867110177166456e-06,
+      "loss": 0.6085,
+      "step": 21338
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5859237738156875e-06,
+      "loss": 0.5761,
+      "step": 21339
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5851366319389314e-06,
+      "loss": 0.5439,
+      "step": 21340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5843495920972083e-06,
+      "loss": 0.5785,
+      "step": 21341
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.583562654301355e-06,
+      "loss": 0.5689,
+      "step": 21342
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.582775818562194e-06,
+      "loss": 0.5371,
+      "step": 21343
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.581989084890557e-06,
+      "loss": 0.557,
+      "step": 21344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.581202453297265e-06,
+      "loss": 0.5232,
+      "step": 21345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.580415923793147e-06,
+      "loss": 0.5571,
+      "step": 21346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5796294963890267e-06,
+      "loss": 0.6006,
+      "step": 21347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.578843171095725e-06,
+      "loss": 0.5529,
+      "step": 21348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5780569479240593e-06,
+      "loss": 0.537,
+      "step": 21349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5772708268848535e-06,
+      "loss": 0.644,
+      "step": 21350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.576484807988924e-06,
+      "loss": 0.5602,
+      "step": 21351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5756988912470825e-06,
+      "loss": 0.5772,
+      "step": 21352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5749130766701514e-06,
+      "loss": 0.59,
+      "step": 21353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5741273642689412e-06,
+      "loss": 0.5614,
+      "step": 21354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5733417540542627e-06,
+      "loss": 0.5818,
+      "step": 21355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5725562460369247e-06,
+      "loss": 0.5169,
+      "step": 21356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5717708402277407e-06,
+      "loss": 0.5369,
+      "step": 21357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.570985536637519e-06,
+      "loss": 0.6147,
+      "step": 21358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5702003352770632e-06,
+      "loss": 0.5918,
+      "step": 21359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5694152361571756e-06,
+      "loss": 0.4969,
+      "step": 21360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5686302392886675e-06,
+      "loss": 0.5956,
+      "step": 21361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.567845344682336e-06,
+      "loss": 0.6434,
+      "step": 21362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.56706055234898e-06,
+      "loss": 0.5696,
+      "step": 21363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5662758622994056e-06,
+      "loss": 0.6028,
+      "step": 21364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5654912745444073e-06,
+      "loss": 0.5666,
+      "step": 21365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.564706789094782e-06,
+      "loss": 0.551,
+      "step": 21366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.56392240596132e-06,
+      "loss": 0.5895,
+      "step": 21367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.563138125154824e-06,
+      "loss": 0.5978,
+      "step": 21368
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5623539466860813e-06,
+      "loss": 0.6593,
+      "step": 21369
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.561569870565883e-06,
+      "loss": 0.5829,
+      "step": 21370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.560785896805018e-06,
+      "loss": 0.5865,
+      "step": 21371
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5600020254142776e-06,
+      "loss": 0.5746,
+      "step": 21372
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5592182564044465e-06,
+      "loss": 0.5241,
+      "step": 21373
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5584345897863074e-06,
+      "loss": 0.5486,
+      "step": 21374
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.55765102557065e-06,
+      "loss": 0.6291,
+      "step": 21375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.556867563768254e-06,
+      "loss": 0.5061,
+      "step": 21376
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5560842043899005e-06,
+      "loss": 0.6185,
+      "step": 21377
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5553009474463665e-06,
+      "loss": 0.5824,
+      "step": 21378
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5545177929484355e-06,
+      "loss": 0.5373,
+      "step": 21379
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5537347409068826e-06,
+      "loss": 0.6183,
+      "step": 21380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.552951791332482e-06,
+      "loss": 0.596,
+      "step": 21381
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5521689442360044e-06,
+      "loss": 0.6165,
+      "step": 21382
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5513861996282306e-06,
+      "loss": 0.594,
+      "step": 21383
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5506035575199273e-06,
+      "loss": 0.6018,
+      "step": 21384
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.549821017921862e-06,
+      "loss": 0.5533,
+      "step": 21385
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.549038580844808e-06,
+      "loss": 0.5907,
+      "step": 21386
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5482562462995296e-06,
+      "loss": 0.5993,
+      "step": 21387
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5474740142967946e-06,
+      "loss": 0.5773,
+      "step": 21388
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.546691884847361e-06,
+      "loss": 0.5637,
+      "step": 21389
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5459098579620034e-06,
+      "loss": 0.5622,
+      "step": 21390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.545127933651472e-06,
+      "loss": 0.5903,
+      "step": 21391
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.544346111926528e-06,
+      "loss": 0.6017,
+      "step": 21392
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.543564392797936e-06,
+      "loss": 0.5586,
+      "step": 21393
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.542782776276449e-06,
+      "loss": 0.5601,
+      "step": 21394
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.542001262372821e-06,
+      "loss": 0.599,
+      "step": 21395
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5412198510978116e-06,
+      "loss": 0.5849,
+      "step": 21396
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.54043854246217e-06,
+      "loss": 0.5994,
+      "step": 21397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5396573364766507e-06,
+      "loss": 0.5658,
+      "step": 21398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5388762331519967e-06,
+      "loss": 0.5954,
+      "step": 21399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.538095232498966e-06,
+      "loss": 0.5611,
+      "step": 21400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.537314334528301e-06,
+      "loss": 0.5604,
+      "step": 21401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.536533539250747e-06,
+      "loss": 0.5742,
+      "step": 21402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.535752846677049e-06,
+      "loss": 0.5449,
+      "step": 21403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5349722568179514e-06,
+      "loss": 0.6086,
+      "step": 21404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5341917696841954e-06,
+      "loss": 0.5584,
+      "step": 21405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5334113852865185e-06,
+      "loss": 0.6019,
+      "step": 21406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.532631103635663e-06,
+      "loss": 0.5965,
+      "step": 21407
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.531850924742366e-06,
+      "loss": 0.6045,
+      "step": 21408
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.531070848617362e-06,
+      "loss": 0.5762,
+      "step": 21409
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5302908752713828e-06,
+      "loss": 0.5791,
+      "step": 21410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5295110047151683e-06,
+      "loss": 0.5425,
+      "step": 21411
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5287312369594463e-06,
+      "loss": 0.6823,
+      "step": 21412
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.527951572014947e-06,
+      "loss": 0.6237,
+      "step": 21413
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5271720098923968e-06,
+      "loss": 0.5973,
+      "step": 21414
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5263925506025288e-06,
+      "loss": 0.4601,
+      "step": 21415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.525613194156067e-06,
+      "loss": 0.594,
+      "step": 21416
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.524833940563731e-06,
+      "loss": 0.6275,
+      "step": 21417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5240547898362522e-06,
+      "loss": 0.5983,
+      "step": 21418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.523275741984348e-06,
+      "loss": 0.5281,
+      "step": 21419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5224967970187386e-06,
+      "loss": 0.6543,
+      "step": 21420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5217179549501415e-06,
+      "loss": 0.6086,
+      "step": 21421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5209392157892788e-06,
+      "loss": 0.4828,
+      "step": 21422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5201605795468642e-06,
+      "loss": 0.6041,
+      "step": 21423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.519382046233613e-06,
+      "loss": 0.6252,
+      "step": 21424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5186036158602344e-06,
+      "loss": 0.5991,
+      "step": 21425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5178252884374465e-06,
+      "loss": 0.5503,
+      "step": 21426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.517047063975957e-06,
+      "loss": 0.5642,
+      "step": 21427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5162689424864728e-06,
+      "loss": 0.6093,
+      "step": 21428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5154909239797064e-06,
+      "loss": 0.5608,
+      "step": 21429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5147130084663607e-06,
+      "loss": 0.5224,
+      "step": 21430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.513935195957142e-06,
+      "loss": 0.5663,
+      "step": 21431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5131574864627494e-06,
+      "loss": 0.5206,
+      "step": 21432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5123798799938913e-06,
+      "loss": 0.5233,
+      "step": 21433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.511602376561265e-06,
+      "loss": 0.5731,
+      "step": 21434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5108249761755676e-06,
+      "loss": 0.6382,
+      "step": 21435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5100476788475027e-06,
+      "loss": 0.6013,
+      "step": 21436
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.509270484587761e-06,
+      "loss": 0.5737,
+      "step": 21437
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.508493393407041e-06,
+      "loss": 0.5799,
+      "step": 21438
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.507716405316031e-06,
+      "loss": 0.6068,
+      "step": 21439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5069395203254287e-06,
+      "loss": 0.5945,
+      "step": 21440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5061627384459243e-06,
+      "loss": 0.6179,
+      "step": 21441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5053860596882042e-06,
+      "loss": 0.6213,
+      "step": 21442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5046094840629543e-06,
+      "loss": 0.5804,
+      "step": 21443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5038330115808673e-06,
+      "loss": 0.6137,
+      "step": 21444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5030566422526236e-06,
+      "loss": 0.5257,
+      "step": 21445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.502280376088906e-06,
+      "loss": 0.5775,
+      "step": 21446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.501504213100402e-06,
+      "loss": 0.522,
+      "step": 21447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.500728153297788e-06,
+      "loss": 0.5386,
+      "step": 21448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.499952196691744e-06,
+      "loss": 0.5089,
+      "step": 21449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.499176343292945e-06,
+      "loss": 0.5217,
+      "step": 21450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4984005931120735e-06,
+      "loss": 0.6028,
+      "step": 21451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4976249461598013e-06,
+      "loss": 0.573,
+      "step": 21452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4968494024468016e-06,
+      "loss": 0.6354,
+      "step": 21453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4960739619837436e-06,
+      "loss": 0.5557,
+      "step": 21454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.495298624781306e-06,
+      "loss": 0.5769,
+      "step": 21455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.494523390850152e-06,
+      "loss": 0.5814,
+      "step": 21456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.493748260200949e-06,
+      "loss": 0.5211,
+      "step": 21457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4929732328443677e-06,
+      "loss": 0.558,
+      "step": 21458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.492198308791072e-06,
+      "loss": 0.6058,
+      "step": 21459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.491423488051723e-06,
+      "loss": 0.5824,
+      "step": 21460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.490648770636983e-06,
+      "loss": 0.5418,
+      "step": 21461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4898741565575157e-06,
+      "loss": 0.5694,
+      "step": 21462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4890996458239804e-06,
+      "loss": 0.5883,
+      "step": 21463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4883252384470325e-06,
+      "loss": 0.5566,
+      "step": 21464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.487550934437327e-06,
+      "loss": 0.529,
+      "step": 21465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.486776733805526e-06,
+      "loss": 0.6048,
+      "step": 21466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.486002636562277e-06,
+      "loss": 0.6297,
+      "step": 21467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4852286427182313e-06,
+      "loss": 0.6337,
+      "step": 21468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.484454752284047e-06,
+      "loss": 0.5761,
+      "step": 21469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4836809652703673e-06,
+      "loss": 0.6366,
+      "step": 21470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.482907281687843e-06,
+      "loss": 0.6057,
+      "step": 21471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.482133701547117e-06,
+      "loss": 0.6037,
+      "step": 21472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4813602248588387e-06,
+      "loss": 0.5373,
+      "step": 21473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.480586851633651e-06,
+      "loss": 0.5822,
+      "step": 21474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4798135818821945e-06,
+      "loss": 0.686,
+      "step": 21475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4790404156151095e-06,
+      "loss": 0.5591,
+      "step": 21476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.478267352843039e-06,
+      "loss": 0.5776,
+      "step": 21477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4774943935766184e-06,
+      "loss": 0.543,
+      "step": 21478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.476721537826482e-06,
+      "loss": 0.5487,
+      "step": 21479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4759487856032714e-06,
+      "loss": 0.5808,
+      "step": 21480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4751761369176165e-06,
+      "loss": 0.6001,
+      "step": 21481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.474403591780149e-06,
+      "loss": 0.5549,
+      "step": 21482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4736311502014986e-06,
+      "loss": 0.5558,
+      "step": 21483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.472858812192299e-06,
+      "loss": 0.5834,
+      "step": 21484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4720865777631775e-06,
+      "loss": 0.6281,
+      "step": 21485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4713144469247584e-06,
+      "loss": 0.5842,
+      "step": 21486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4705424196876647e-06,
+      "loss": 0.6128,
+      "step": 21487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4697704960625256e-06,
+      "loss": 0.6053,
+      "step": 21488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4689986760599626e-06,
+      "loss": 0.6098,
+      "step": 21489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4682269596905915e-06,
+      "loss": 0.569,
+      "step": 21490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4674553469650387e-06,
+      "loss": 0.5749,
+      "step": 21491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4666838378939185e-06,
+      "loss": 0.5317,
+      "step": 21492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.465912432487848e-06,
+      "loss": 0.5343,
+      "step": 21493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.465141130757439e-06,
+      "loss": 0.5859,
+      "step": 21494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.464369932713313e-06,
+      "loss": 0.6288,
+      "step": 21495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.463598838366077e-06,
+      "loss": 0.537,
+      "step": 21496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.462827847726341e-06,
+      "loss": 0.5159,
+      "step": 21497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4620569608047197e-06,
+      "loss": 0.616,
+      "step": 21498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4612861776118168e-06,
+      "loss": 0.6081,
+      "step": 21499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4605154981582415e-06,
+      "loss": 0.6176,
+      "step": 21500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4597449224545933e-06,
+      "loss": 0.5614,
+      "step": 21501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.458974450511484e-06,
+      "loss": 0.5372,
+      "step": 21502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4582040823395126e-06,
+      "loss": 0.5827,
+      "step": 21503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4574338179492797e-06,
+      "loss": 0.5348,
+      "step": 21504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4566636573513826e-06,
+      "loss": 0.5732,
+      "step": 21505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4558936005564237e-06,
+      "loss": 0.5408,
+      "step": 21506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4551236475749985e-06,
+      "loss": 0.5993,
+      "step": 21507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.454353798417698e-06,
+      "loss": 0.6026,
+      "step": 21508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.453584053095123e-06,
+      "loss": 0.6251,
+      "step": 21509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.452814411617861e-06,
+      "loss": 0.5671,
+      "step": 21510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.452044873996505e-06,
+      "loss": 0.4886,
+      "step": 21511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4512754402416407e-06,
+      "loss": 0.5878,
+      "step": 21512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.450506110363863e-06,
+      "loss": 0.6362,
+      "step": 21513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4497368843737533e-06,
+      "loss": 0.6014,
+      "step": 21514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4489677622818997e-06,
+      "loss": 0.6259,
+      "step": 21515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4481987440988807e-06,
+      "loss": 0.5698,
+      "step": 21516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.447429829835286e-06,
+      "loss": 0.5554,
+      "step": 21517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4466610195016925e-06,
+      "loss": 0.6003,
+      "step": 21518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4458923131086787e-06,
+      "loss": 0.6197,
+      "step": 21519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4451237106668258e-06,
+      "loss": 0.614,
+      "step": 21520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4443552121867098e-06,
+      "loss": 0.6118,
+      "step": 21521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4435868176789047e-06,
+      "loss": 0.599,
+      "step": 21522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.442818527153982e-06,
+      "loss": 0.6169,
+      "step": 21523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4420503406225194e-06,
+      "loss": 0.5898,
+      "step": 21524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4412822580950846e-06,
+      "loss": 0.5556,
+      "step": 21525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.440514279582248e-06,
+      "loss": 0.5166,
+      "step": 21526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4397464050945753e-06,
+      "loss": 0.6027,
+      "step": 21527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4389786346426368e-06,
+      "loss": 0.5794,
+      "step": 21528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4382109682369958e-06,
+      "loss": 0.5416,
+      "step": 21529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.437443405888215e-06,
+      "loss": 0.5551,
+      "step": 21530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4366759476068593e-06,
+      "loss": 0.6112,
+      "step": 21531
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4359085934034888e-06,
+      "loss": 0.5417,
+      "step": 21532
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4351413432886616e-06,
+      "loss": 0.6376,
+      "step": 21533
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.434374197272935e-06,
+      "loss": 0.5749,
+      "step": 21534
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4336071553668696e-06,
+      "loss": 0.5993,
+      "step": 21535
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4328402175810175e-06,
+      "loss": 0.6094,
+      "step": 21536
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4320733839259337e-06,
+      "loss": 0.5335,
+      "step": 21537
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4313066544121676e-06,
+      "loss": 0.5656,
+      "step": 21538
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4305400290502745e-06,
+      "loss": 0.5079,
+      "step": 21539
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.429773507850801e-06,
+      "loss": 0.5595,
+      "step": 21540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.429007090824295e-06,
+      "loss": 0.6482,
+      "step": 21541
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4282407779813054e-06,
+      "loss": 0.5605,
+      "step": 21542
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4274745693323753e-06,
+      "loss": 0.6021,
+      "step": 21543
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4267084648880503e-06,
+      "loss": 0.5421,
+      "step": 21544
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.425942464658867e-06,
+      "loss": 0.575,
+      "step": 21545
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.425176568655374e-06,
+      "loss": 0.5764,
+      "step": 21546
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4244107768881074e-06,
+      "loss": 0.5765,
+      "step": 21547
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4236450893676055e-06,
+      "loss": 0.5027,
+      "step": 21548
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4228795061044007e-06,
+      "loss": 0.5507,
+      "step": 21549
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.422114027109035e-06,
+      "loss": 0.5872,
+      "step": 21550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4213486523920393e-06,
+      "loss": 0.5344,
+      "step": 21551
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.420583381963941e-06,
+      "loss": 0.6776,
+      "step": 21552
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.419818215835279e-06,
+      "loss": 0.5909,
+      "step": 21553
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4190531540165798e-06,
+      "loss": 0.6527,
+      "step": 21554
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.418288196518369e-06,
+      "loss": 0.5403,
+      "step": 21555
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.417523343351171e-06,
+      "loss": 0.5837,
+      "step": 21556
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.416758594525519e-06,
+      "loss": 0.6228,
+      "step": 21557
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4159939500519313e-06,
+      "loss": 0.5202,
+      "step": 21558
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.415229409940928e-06,
+      "loss": 0.5223,
+      "step": 21559
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4144649742030348e-06,
+      "loss": 0.5538,
+      "step": 21560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4137006428487696e-06,
+      "loss": 0.5746,
+      "step": 21561
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4129364158886492e-06,
+      "loss": 0.6075,
+      "step": 21562
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.412172293333187e-06,
+      "loss": 0.588,
+      "step": 21563
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4114082751929045e-06,
+      "loss": 0.5419,
+      "step": 21564
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4106443614783116e-06,
+      "loss": 0.5798,
+      "step": 21565
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.409880552199921e-06,
+      "loss": 0.6211,
+      "step": 21566
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4091168473682414e-06,
+      "loss": 0.5555,
+      "step": 21567
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4083532469937866e-06,
+      "loss": 0.5813,
+      "step": 21568
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4075897510870604e-06,
+      "loss": 0.6239,
+      "step": 21569
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4068263596585685e-06,
+      "loss": 0.5389,
+      "step": 21570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4060630727188205e-06,
+      "loss": 0.5504,
+      "step": 21571
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4052998902783176e-06,
+      "loss": 0.5222,
+      "step": 21572
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.404536812347561e-06,
+      "loss": 0.5461,
+      "step": 21573
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.40377383893705e-06,
+      "loss": 0.5623,
+      "step": 21574
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4030109700572877e-06,
+      "loss": 0.5411,
+      "step": 21575
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.402248205718769e-06,
+      "loss": 0.6099,
+      "step": 21576
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.401485545931992e-06,
+      "loss": 0.6027,
+      "step": 21577
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.400722990707447e-06,
+      "loss": 0.5598,
+      "step": 21578
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3999605400556347e-06,
+      "loss": 0.4802,
+      "step": 21579
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3991981939870424e-06,
+      "loss": 0.547,
+      "step": 21580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.398435952512159e-06,
+      "loss": 0.5151,
+      "step": 21581
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3976738156414804e-06,
+      "loss": 0.5763,
+      "step": 21582
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3969117833854884e-06,
+      "loss": 0.5478,
+      "step": 21583
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3961498557546716e-06,
+      "loss": 0.5049,
+      "step": 21584
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3953880327595113e-06,
+      "loss": 0.6102,
+      "step": 21585
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3946263144104966e-06,
+      "loss": 0.5297,
+      "step": 21586
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.393864700718106e-06,
+      "loss": 0.6196,
+      "step": 21587
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3931031916928203e-06,
+      "loss": 0.64,
+      "step": 21588
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.392341787345116e-06,
+      "loss": 0.6396,
+      "step": 21589
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3915804876854764e-06,
+      "loss": 0.5373,
+      "step": 21590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3908192927243747e-06,
+      "loss": 0.5656,
+      "step": 21591
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.390058202472283e-06,
+      "loss": 0.5134,
+      "step": 21592
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.38929721693968e-06,
+      "loss": 0.5595,
+      "step": 21593
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.388536336137034e-06,
+      "loss": 0.6063,
+      "step": 21594
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3877755600748164e-06,
+      "loss": 0.596,
+      "step": 21595
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3870148887634926e-06,
+      "loss": 0.5868,
+      "step": 21596
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3862543222135358e-06,
+      "loss": 0.5459,
+      "step": 21597
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3854938604354105e-06,
+      "loss": 0.5884,
+      "step": 21598
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3847335034395803e-06,
+      "loss": 0.558,
+      "step": 21599
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3839732512365056e-06,
+      "loss": 0.6377,
+      "step": 21600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3832131038366556e-06,
+      "loss": 0.618,
+      "step": 21601
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3824530612504816e-06,
+      "loss": 0.6532,
+      "step": 21602
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.381693123488449e-06,
+      "loss": 0.5298,
+      "step": 21603
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.380933290561014e-06,
+      "loss": 0.6295,
+      "step": 21604
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3801735624786305e-06,
+      "loss": 0.5396,
+      "step": 21605
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.379413939251751e-06,
+      "loss": 0.5091,
+      "step": 21606
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3786544208908356e-06,
+      "loss": 0.5512,
+      "step": 21607
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3778950074063314e-06,
+      "loss": 0.5903,
+      "step": 21608
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.37713569880869e-06,
+      "loss": 0.6045,
+      "step": 21609
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3763764951083556e-06,
+      "loss": 0.6007,
+      "step": 21610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3756173963157825e-06,
+      "loss": 0.6186,
+      "step": 21611
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3748584024414147e-06,
+      "loss": 0.5988,
+      "step": 21612
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3740995134956914e-06,
+      "loss": 0.5578,
+      "step": 21613
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.373340729489062e-06,
+      "loss": 0.596,
+      "step": 21614
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.372582050431965e-06,
+      "loss": 0.6632,
+      "step": 21615
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3718234763348426e-06,
+      "loss": 0.5571,
+      "step": 21616
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3710650072081277e-06,
+      "loss": 0.575,
+      "step": 21617
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3703066430622654e-06,
+      "loss": 0.5183,
+      "step": 21618
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.369548383907687e-06,
+      "loss": 0.5953,
+      "step": 21619
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.368790229754828e-06,
+      "loss": 0.5949,
+      "step": 21620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3680321806141182e-06,
+      "loss": 0.6054,
+      "step": 21621
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3672742364959944e-06,
+      "loss": 0.5554,
+      "step": 21622
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.366516397410884e-06,
+      "loss": 0.5967,
+      "step": 21623
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.365758663369212e-06,
+      "loss": 0.5842,
+      "step": 21624
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3650010343814113e-06,
+      "loss": 0.6384,
+      "step": 21625
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3642435104579054e-06,
+      "loss": 0.5427,
+      "step": 21626
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3634860916091183e-06,
+      "loss": 0.6502,
+      "step": 21627
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.36272877784547e-06,
+      "loss": 0.5822,
+      "step": 21628
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.361971569177387e-06,
+      "loss": 0.6341,
+      "step": 21629
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.361214465615287e-06,
+      "loss": 0.5562,
+      "step": 21630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3604574671695867e-06,
+      "loss": 0.5758,
+      "step": 21631
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.359700573850703e-06,
+      "loss": 0.5509,
+      "step": 21632
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3589437856690543e-06,
+      "loss": 0.5289,
+      "step": 21633
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3581871026350546e-06,
+      "loss": 0.5592,
+      "step": 21634
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.357430524759111e-06,
+      "loss": 0.5883,
+      "step": 21635
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3566740520516416e-06,
+      "loss": 0.5992,
+      "step": 21636
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3559176845230534e-06,
+      "loss": 0.6413,
+      "step": 21637
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.355161422183756e-06,
+      "loss": 0.5472,
+      "step": 21638
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3544052650441505e-06,
+      "loss": 0.5613,
+      "step": 21639
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.353649213114649e-06,
+      "loss": 0.5708,
+      "step": 21640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3528932664056537e-06,
+      "loss": 0.5953,
+      "step": 21641
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3521374249275664e-06,
+      "loss": 0.5704,
+      "step": 21642
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3513816886907847e-06,
+      "loss": 0.5802,
+      "step": 21643
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3506260577057148e-06,
+      "loss": 0.5519,
+      "step": 21644
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.349870531982753e-06,
+      "loss": 0.4953,
+      "step": 21645
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.34911511153229e-06,
+      "loss": 0.6201,
+      "step": 21646
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.34835979636473e-06,
+      "loss": 0.6349,
+      "step": 21647
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3476045864904605e-06,
+      "loss": 0.5425,
+      "step": 21648
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3468494819198773e-06,
+      "loss": 0.5967,
+      "step": 21649
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.346094482663367e-06,
+      "loss": 0.7011,
+      "step": 21650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3453395887313247e-06,
+      "loss": 0.6015,
+      "step": 21651
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.344584800134135e-06,
+      "loss": 0.554,
+      "step": 21652
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3438301168821864e-06,
+      "loss": 0.6045,
+      "step": 21653
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3430755389858596e-06,
+      "loss": 0.5164,
+      "step": 21654
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3423210664555428e-06,
+      "loss": 0.5919,
+      "step": 21655
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3415666993016184e-06,
+      "loss": 0.5991,
+      "step": 21656
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.340812437534462e-06,
+      "loss": 0.5586,
+      "step": 21657
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.340058281164459e-06,
+      "loss": 0.5132,
+      "step": 21658
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.339304230201984e-06,
+      "loss": 0.5763,
+      "step": 21659
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.338550284657416e-06,
+      "loss": 0.5767,
+      "step": 21660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3377964445411237e-06,
+      "loss": 0.5913,
+      "step": 21661
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3370427098634883e-06,
+      "loss": 0.5831,
+      "step": 21662
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.336289080634878e-06,
+      "loss": 0.5819,
+      "step": 21663
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.335535556865661e-06,
+      "loss": 0.5773,
+      "step": 21664
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.334782138566213e-06,
+      "loss": 0.5778,
+      "step": 21665
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.334028825746897e-06,
+      "loss": 0.535,
+      "step": 21666
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3332756184180815e-06,
+      "loss": 0.5647,
+      "step": 21667
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3325225165901266e-06,
+      "loss": 0.6005,
+      "step": 21668
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3317695202734013e-06,
+      "loss": 0.6083,
+      "step": 21669
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3310166294782664e-06,
+      "loss": 0.5553,
+      "step": 21670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3302638442150804e-06,
+      "loss": 0.5656,
+      "step": 21671
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3295111644942003e-06,
+      "loss": 0.5015,
+      "step": 21672
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3287585903259903e-06,
+      "loss": 0.6365,
+      "step": 21673
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.328006121720802e-06,
+      "loss": 0.5418,
+      "step": 21674
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3272537586889875e-06,
+      "loss": 0.5364,
+      "step": 21675
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3265015012409063e-06,
+      "loss": 0.6277,
+      "step": 21676
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3257493493869066e-06,
+      "loss": 0.6019,
+      "step": 21677
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3249973031373394e-06,
+      "loss": 0.5616,
+      "step": 21678
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3242453625025507e-06,
+      "loss": 0.5399,
+      "step": 21679
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3234935274928925e-06,
+      "loss": 0.5056,
+      "step": 21680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3227417981187083e-06,
+      "loss": 0.6356,
+      "step": 21681
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.321990174390344e-06,
+      "loss": 0.6076,
+      "step": 21682
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.321238656318138e-06,
+      "loss": 0.5505,
+      "step": 21683
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.320487243912437e-06,
+      "loss": 0.6054,
+      "step": 21684
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3197359371835802e-06,
+      "loss": 0.5628,
+      "step": 21685
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.318984736141903e-06,
+      "loss": 0.6322,
+      "step": 21686
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.318233640797747e-06,
+      "loss": 0.5015,
+      "step": 21687
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3174826511614467e-06,
+      "loss": 0.5874,
+      "step": 21688
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3167317672433355e-06,
+      "loss": 0.5114,
+      "step": 21689
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.315980989053742e-06,
+      "loss": 0.5685,
+      "step": 21690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3152303166030067e-06,
+      "loss": 0.5532,
+      "step": 21691
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3144797499014537e-06,
+      "loss": 0.5474,
+      "step": 21692
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.313729288959412e-06,
+      "loss": 0.5212,
+      "step": 21693
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3129789337872064e-06,
+      "loss": 0.5589,
+      "step": 21694
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.312228684395168e-06,
+      "loss": 0.5696,
+      "step": 21695
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.311478540793619e-06,
+      "loss": 0.5925,
+      "step": 21696
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.310728502992877e-06,
+      "loss": 0.5322,
+      "step": 21697
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3099785710032706e-06,
+      "loss": 0.5321,
+      "step": 21698
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.309228744835116e-06,
+      "loss": 0.5597,
+      "step": 21699
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.308479024498731e-06,
+      "loss": 0.5736,
+      "step": 21700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3077294100044314e-06,
+      "loss": 0.599,
+      "step": 21701
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.306979901362535e-06,
+      "loss": 0.5956,
+      "step": 21702
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.306230498583356e-06,
+      "loss": 0.5878,
+      "step": 21703
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3054812016772067e-06,
+      "loss": 0.556,
+      "step": 21704
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3047320106543922e-06,
+      "loss": 0.5398,
+      "step": 21705
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3039829255252312e-06,
+      "loss": 0.4949,
+      "step": 21706
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.303233946300026e-06,
+      "loss": 0.582,
+      "step": 21707
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3024850729890834e-06,
+      "loss": 0.5792,
+      "step": 21708
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.301736305602712e-06,
+      "loss": 0.5878,
+      "step": 21709
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3009876441512137e-06,
+      "loss": 0.5959,
+      "step": 21710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.30023908864489e-06,
+      "loss": 0.5878,
+      "step": 21711
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2994906390940395e-06,
+      "loss": 0.6316,
+      "step": 21712
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2987422955089668e-06,
+      "loss": 0.6153,
+      "step": 21713
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.297994057899967e-06,
+      "loss": 0.5682,
+      "step": 21714
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.297245926277336e-06,
+      "loss": 0.6108,
+      "step": 21715
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2964979006513666e-06,
+      "loss": 0.5229,
+      "step": 21716
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.295749981032358e-06,
+      "loss": 0.5274,
+      "step": 21717
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2950021674306e-06,
+      "loss": 0.5946,
+      "step": 21718
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2942544598563785e-06,
+      "loss": 0.619,
+      "step": 21719
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2935068583199902e-06,
+      "loss": 0.5351,
+      "step": 21720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.292759362831718e-06,
+      "loss": 0.6198,
+      "step": 21721
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2920119734018497e-06,
+      "loss": 0.5716,
+      "step": 21722
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2912646900406665e-06,
+      "loss": 0.6047,
+      "step": 21723
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2905175127584578e-06,
+      "loss": 0.5153,
+      "step": 21724
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.289770441565502e-06,
+      "loss": 0.5747,
+      "step": 21725
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2890234764720778e-06,
+      "loss": 0.5956,
+      "step": 21726
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2882766174884673e-06,
+      "loss": 0.5482,
+      "step": 21727
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2875298646249477e-06,
+      "loss": 0.5684,
+      "step": 21728
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.286783217891794e-06,
+      "loss": 0.5488,
+      "step": 21729
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.286036677299277e-06,
+      "loss": 0.5769,
+      "step": 21730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2852902428576763e-06,
+      "loss": 0.5571,
+      "step": 21731
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2845439145772607e-06,
+      "loss": 0.5528,
+      "step": 21732
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2837976924683014e-06,
+      "loss": 0.5673,
+      "step": 21733
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.283051576541062e-06,
+      "loss": 0.6311,
+      "step": 21734
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2823055668058157e-06,
+      "loss": 0.5804,
+      "step": 21735
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.281559663272828e-06,
+      "loss": 0.5873,
+      "step": 21736
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2808138659523573e-06,
+      "loss": 0.6156,
+      "step": 21737
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.280068174854674e-06,
+      "loss": 0.5965,
+      "step": 21738
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.279322589990035e-06,
+      "loss": 0.4772,
+      "step": 21739
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2785771113687018e-06,
+      "loss": 0.5783,
+      "step": 21740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2778317390009284e-06,
+      "loss": 0.6479,
+      "step": 21741
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2770864728969798e-06,
+      "loss": 0.5877,
+      "step": 21742
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2763413130671066e-06,
+      "loss": 0.5844,
+      "step": 21743
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2755962595215643e-06,
+      "loss": 0.5635,
+      "step": 21744
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2748513122706008e-06,
+      "loss": 0.5822,
+      "step": 21745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2741064713244744e-06,
+      "loss": 0.5506,
+      "step": 21746
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2733617366934314e-06,
+      "loss": 0.5632,
+      "step": 21747
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.272617108387718e-06,
+      "loss": 0.6375,
+      "step": 21748
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2718725864175852e-06,
+      "loss": 0.5963,
+      "step": 21749
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2711281707932752e-06,
+      "loss": 0.5499,
+      "step": 21750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.270383861525034e-06,
+      "loss": 0.6088,
+      "step": 21751
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2696396586230994e-06,
+      "loss": 0.5163,
+      "step": 21752
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2688955620977194e-06,
+      "loss": 0.5306,
+      "step": 21753
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2681515719591284e-06,
+      "loss": 0.5324,
+      "step": 21754
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.267407688217567e-06,
+      "loss": 0.5868,
+      "step": 21755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2666639108832656e-06,
+      "loss": 0.643,
+      "step": 21756
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.265920239966468e-06,
+      "loss": 0.5865,
+      "step": 21757
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.265176675477403e-06,
+      "loss": 0.5342,
+      "step": 21758
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2644332174263006e-06,
+      "loss": 0.5852,
+      "step": 21759
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2636898658233973e-06,
+      "loss": 0.5615,
+      "step": 21760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.26294662067892e-06,
+      "loss": 0.6328,
+      "step": 21761
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2622034820030936e-06,
+      "loss": 0.5371,
+      "step": 21762
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2614604498061453e-06,
+      "loss": 0.5515,
+      "step": 21763
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2607175240983027e-06,
+      "loss": 0.6056,
+      "step": 21764
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.259974704889788e-06,
+      "loss": 0.5303,
+      "step": 21765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.259231992190821e-06,
+      "loss": 0.6464,
+      "step": 21766
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.258489386011622e-06,
+      "loss": 0.5874,
+      "step": 21767
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.257746886362413e-06,
+      "loss": 0.5747,
+      "step": 21768
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.257004493253411e-06,
+      "loss": 0.6372,
+      "step": 21769
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.256262206694827e-06,
+      "loss": 0.6264,
+      "step": 21770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2555200266968822e-06,
+      "loss": 0.5488,
+      "step": 21771
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2547779532697876e-06,
+      "loss": 0.5609,
+      "step": 21772
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.254035986423754e-06,
+      "loss": 0.6305,
+      "step": 21773
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2532941261689866e-06,
+      "loss": 0.531,
+      "step": 21774
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.252552372515704e-06,
+      "loss": 0.6055,
+      "step": 21775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2518107254741072e-06,
+      "loss": 0.5775,
+      "step": 21776
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.251069185054403e-06,
+      "loss": 0.5662,
+      "step": 21777
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.250327751266792e-06,
+      "loss": 0.5408,
+      "step": 21778
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2495864241214848e-06,
+      "loss": 0.596,
+      "step": 21779
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2488452036286777e-06,
+      "loss": 0.5664,
+      "step": 21780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2481040897985696e-06,
+      "loss": 0.5745,
+      "step": 21781
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.247363082641362e-06,
+      "loss": 0.5427,
+      "step": 21782
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2466221821672517e-06,
+      "loss": 0.6043,
+      "step": 21783
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2458813883864315e-06,
+      "loss": 0.5604,
+      "step": 21784
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2451407013090955e-06,
+      "loss": 0.5205,
+      "step": 21785
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.244400120945439e-06,
+      "loss": 0.5324,
+      "step": 21786
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2436596473056527e-06,
+      "loss": 0.5682,
+      "step": 21787
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2429192803999213e-06,
+      "loss": 0.5664,
+      "step": 21788
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.24217902023844e-06,
+      "loss": 0.6613,
+      "step": 21789
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2414388668313912e-06,
+      "loss": 0.5717,
+      "step": 21790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.240698820188961e-06,
+      "loss": 0.616,
+      "step": 21791
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.23995888032133e-06,
+      "loss": 0.5551,
+      "step": 21792
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2392190472386865e-06,
+      "loss": 0.5334,
+      "step": 21793
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.238479320951209e-06,
+      "loss": 0.6736,
+      "step": 21794
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.237739701469075e-06,
+      "loss": 0.5783,
+      "step": 21795
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2370001888024606e-06,
+      "loss": 0.6753,
+      "step": 21796
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2362607829615466e-06,
+      "loss": 0.5657,
+      "step": 21797
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2355214839565074e-06,
+      "loss": 0.5801,
+      "step": 21798
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2347822917975116e-06,
+      "loss": 0.5814,
+      "step": 21799
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2340432064947383e-06,
+      "loss": 0.5505,
+      "step": 21800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.233304228058353e-06,
+      "loss": 0.6187,
+      "step": 21801
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.232565356498526e-06,
+      "loss": 0.5352,
+      "step": 21802
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2318265918254234e-06,
+      "loss": 0.5827,
+      "step": 21803
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.231087934049214e-06,
+      "loss": 0.6412,
+      "step": 21804
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2303493831800616e-06,
+      "loss": 0.5443,
+      "step": 21805
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2296109392281286e-06,
+      "loss": 0.5397,
+      "step": 21806
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2288726022035732e-06,
+      "loss": 0.5562,
+      "step": 21807
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.228134372116563e-06,
+      "loss": 0.5163,
+      "step": 21808
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.227396248977253e-06,
+      "loss": 0.5644,
+      "step": 21809
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2266582327957962e-06,
+      "loss": 0.52,
+      "step": 21810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.225920323582359e-06,
+      "loss": 0.6029,
+      "step": 21811
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2251825213470856e-06,
+      "loss": 0.6076,
+      "step": 21812
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2244448261001295e-06,
+      "loss": 0.569,
+      "step": 21813
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.223707237851648e-06,
+      "loss": 0.5644,
+      "step": 21814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.222969756611787e-06,
+      "loss": 0.5194,
+      "step": 21815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2222323823906965e-06,
+      "loss": 0.526,
+      "step": 21816
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.22149511519852e-06,
+      "loss": 0.5901,
+      "step": 21817
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2207579550454074e-06,
+      "loss": 0.5521,
+      "step": 21818
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.220020901941501e-06,
+      "loss": 0.5374,
+      "step": 21819
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2192839558969406e-06,
+      "loss": 0.6028,
+      "step": 21820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.218547116921872e-06,
+      "loss": 0.5575,
+      "step": 21821
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2178103850264334e-06,
+      "loss": 0.5498,
+      "step": 21822
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2170737602207605e-06,
+      "loss": 0.5781,
+      "step": 21823
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2163372425149897e-06,
+      "loss": 0.6041,
+      "step": 21824
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.21560083191926e-06,
+      "loss": 0.5364,
+      "step": 21825
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.214864528443703e-06,
+      "loss": 0.598,
+      "step": 21826
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2141283320984506e-06,
+      "loss": 0.5534,
+      "step": 21827
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2133922428936306e-06,
+      "loss": 0.5869,
+      "step": 21828
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2126562608393785e-06,
+      "loss": 0.622,
+      "step": 21829
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2119203859458195e-06,
+      "loss": 0.569,
+      "step": 21830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2111846182230754e-06,
+      "loss": 0.5959,
+      "step": 21831
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.210448957681277e-06,
+      "loss": 0.5962,
+      "step": 21832
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2097134043305467e-06,
+      "loss": 0.5778,
+      "step": 21833
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2089779581810044e-06,
+      "loss": 0.5623,
+      "step": 21834
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2082426192427685e-06,
+      "loss": 0.557,
+      "step": 21835
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2075073875259645e-06,
+      "loss": 0.5337,
+      "step": 21836
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2067722630407042e-06,
+      "loss": 0.5527,
+      "step": 21837
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2060372457971067e-06,
+      "loss": 0.6019,
+      "step": 21838
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.205302335805282e-06,
+      "loss": 0.6847,
+      "step": 21839
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.204567533075348e-06,
+      "loss": 0.6069,
+      "step": 21840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2038328376174146e-06,
+      "loss": 0.5733,
+      "step": 21841
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2030982494415896e-06,
+      "loss": 0.5471,
+      "step": 21842
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2023637685579856e-06,
+      "loss": 0.5425,
+      "step": 21843
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2016293949767085e-06,
+      "loss": 0.5601,
+      "step": 21844
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.200895128707863e-06,
+      "loss": 0.5701,
+      "step": 21845
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.200160969761551e-06,
+      "loss": 0.5346,
+      "step": 21846
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.19942691814788e-06,
+      "loss": 0.5819,
+      "step": 21847
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.198692973876949e-06,
+      "loss": 0.5292,
+      "step": 21848
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.197959136958857e-06,
+      "loss": 0.5535,
+      "step": 21849
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1972254074037004e-06,
+      "loss": 0.5784,
+      "step": 21850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1964917852215807e-06,
+      "loss": 0.5087,
+      "step": 21851
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.195758270422591e-06,
+      "loss": 0.5841,
+      "step": 21852
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1950248630168223e-06,
+      "loss": 0.6339,
+      "step": 21853
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1942915630143723e-06,
+      "loss": 0.5701,
+      "step": 21854
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1935583704253283e-06,
+      "loss": 0.5829,
+      "step": 21855
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.192825285259782e-06,
+      "loss": 0.5632,
+      "step": 21856
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1920923075278153e-06,
+      "loss": 0.5935,
+      "step": 21857
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.191359437239523e-06,
+      "loss": 0.5608,
+      "step": 21858
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1906266744049853e-06,
+      "loss": 0.5504,
+      "step": 21859
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1898940190342875e-06,
+      "loss": 0.5916,
+      "step": 21860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1891614711375076e-06,
+      "loss": 0.6087,
+      "step": 21861
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1884290307247313e-06,
+      "loss": 0.5559,
+      "step": 21862
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1876966978060353e-06,
+      "loss": 0.5401,
+      "step": 21863
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.186964472391495e-06,
+      "loss": 0.5759,
+      "step": 21864
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.186232354491191e-06,
+      "loss": 0.5865,
+      "step": 21865
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1855003441151957e-06,
+      "loss": 0.6221,
+      "step": 21866
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1847684412735828e-06,
+      "loss": 0.6085,
+      "step": 21867
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1840366459764194e-06,
+      "loss": 0.5828,
+      "step": 21868
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.183304958233782e-06,
+      "loss": 0.5904,
+      "step": 21869
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.182573378055738e-06,
+      "loss": 0.5267,
+      "step": 21870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.181841905452352e-06,
+      "loss": 0.5608,
+      "step": 21871
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1811105404336896e-06,
+      "loss": 0.5342,
+      "step": 21872
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.180379283009818e-06,
+      "loss": 0.5673,
+      "step": 21873
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.179648133190799e-06,
+      "loss": 0.5507,
+      "step": 21874
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.178917090986691e-06,
+      "loss": 0.601,
+      "step": 21875
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.178186156407558e-06,
+      "loss": 0.5899,
+      "step": 21876
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.177455329463456e-06,
+      "loss": 0.5859,
+      "step": 21877
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1767246101644425e-06,
+      "loss": 0.5834,
+      "step": 21878
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1759939985205704e-06,
+      "loss": 0.5958,
+      "step": 21879
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.175263494541897e-06,
+      "loss": 0.6348,
+      "step": 21880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.174533098238475e-06,
+      "loss": 0.5862,
+      "step": 21881
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.17380280962035e-06,
+      "loss": 0.5919,
+      "step": 21882
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.173072628697578e-06,
+      "loss": 0.6093,
+      "step": 21883
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1723425554802036e-06,
+      "loss": 0.5895,
+      "step": 21884
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1716125899782734e-06,
+      "loss": 0.5293,
+      "step": 21885
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1708827322018313e-06,
+      "loss": 0.6312,
+      "step": 21886
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1701529821609235e-06,
+      "loss": 0.5954,
+      "step": 21887
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1694233398655906e-06,
+      "loss": 0.5911,
+      "step": 21888
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1686938053258734e-06,
+      "loss": 0.6153,
+      "step": 21889
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1679643785518077e-06,
+      "loss": 0.5698,
+      "step": 21890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1672350595534373e-06,
+      "loss": 0.6285,
+      "step": 21891
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.166505848340794e-06,
+      "loss": 0.6228,
+      "step": 21892
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1657767449239118e-06,
+      "loss": 0.6436,
+      "step": 21893
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1650477493128274e-06,
+      "loss": 0.6035,
+      "step": 21894
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1643188615175705e-06,
+      "loss": 0.6237,
+      "step": 21895
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1635900815481716e-06,
+      "loss": 0.596,
+      "step": 21896
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.162861409414655e-06,
+      "loss": 0.555,
+      "step": 21897
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.162132845127055e-06,
+      "loss": 0.57,
+      "step": 21898
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1614043886953937e-06,
+      "loss": 0.6059,
+      "step": 21899
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1606760401296954e-06,
+      "loss": 0.5575,
+      "step": 21900
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.159947799439981e-06,
+      "loss": 0.5918,
+      "step": 21901
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1592196666362763e-06,
+      "loss": 0.5971,
+      "step": 21902
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1584916417285985e-06,
+      "loss": 0.5376,
+      "step": 21903
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1577637247269635e-06,
+      "loss": 0.5626,
+      "step": 21904
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.157035915641392e-06,
+      "loss": 0.6854,
+      "step": 21905
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.156308214481899e-06,
+      "loss": 0.5691,
+      "step": 21906
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1555806212584963e-06,
+      "loss": 0.5741,
+      "step": 21907
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.154853135981195e-06,
+      "loss": 0.5895,
+      "step": 21908
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1541257586600097e-06,
+      "loss": 0.6216,
+      "step": 21909
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1533984893049487e-06,
+      "loss": 0.56,
+      "step": 21910
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.152671327926019e-06,
+      "loss": 0.584,
+      "step": 21911
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.151944274533224e-06,
+      "loss": 0.5319,
+      "step": 21912
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1512173291365747e-06,
+      "loss": 0.5732,
+      "step": 21913
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.150490491746071e-06,
+      "loss": 0.5909,
+      "step": 21914
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1497637623717126e-06,
+      "loss": 0.6094,
+      "step": 21915
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1490371410235056e-06,
+      "loss": 0.5064,
+      "step": 21916
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.148310627711446e-06,
+      "loss": 0.5844,
+      "step": 21917
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1475842224455323e-06,
+      "loss": 0.5729,
+      "step": 21918
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.146857925235756e-06,
+      "loss": 0.6034,
+      "step": 21919
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1461317360921174e-06,
+      "loss": 0.5885,
+      "step": 21920
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1454056550246073e-06,
+      "loss": 0.588,
+      "step": 21921
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.144679682043217e-06,
+      "loss": 0.5981,
+      "step": 21922
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1439538171579333e-06,
+      "loss": 0.577,
+      "step": 21923
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1432280603787514e-06,
+      "loss": 0.5653,
+      "step": 21924
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1425024117156545e-06,
+      "loss": 0.6512,
+      "step": 21925
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.141776871178627e-06,
+      "loss": 0.5317,
+      "step": 21926
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.141051438777656e-06,
+      "loss": 0.6082,
+      "step": 21927
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.140326114522723e-06,
+      "loss": 0.5713,
+      "step": 21928
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1396008984238094e-06,
+      "loss": 0.6208,
+      "step": 21929
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1388757904908908e-06,
+      "loss": 0.5917,
+      "step": 21930
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1381507907339516e-06,
+      "loss": 0.5755,
+      "step": 21931
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.137425899162966e-06,
+      "loss": 0.5305,
+      "step": 21932
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.136701115787908e-06,
+      "loss": 0.6184,
+      "step": 21933
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1359764406187487e-06,
+      "loss": 0.6127,
+      "step": 21934
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.135251873665468e-06,
+      "loss": 0.6194,
+      "step": 21935
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.134527414938031e-06,
+      "loss": 0.5433,
+      "step": 21936
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1338030644464057e-06,
+      "loss": 0.5577,
+      "step": 21937
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.133078822200564e-06,
+      "loss": 0.5564,
+      "step": 21938
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.132354688210471e-06,
+      "loss": 0.535,
+      "step": 21939
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1316306624860907e-06,
+      "loss": 0.5339,
+      "step": 21940
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.130906745037382e-06,
+      "loss": 0.5937,
+      "step": 21941
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1301829358743154e-06,
+      "loss": 0.639,
+      "step": 21942
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1294592350068454e-06,
+      "loss": 0.5782,
+      "step": 21943
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1287356424449303e-06,
+      "loss": 0.6213,
+      "step": 21944
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1280121581985314e-06,
+      "loss": 0.5907,
+      "step": 21945
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1272887822776025e-06,
+      "loss": 0.5624,
+      "step": 21946
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1265655146920972e-06,
+      "loss": 0.5318,
+      "step": 21947
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1258423554519646e-06,
+      "loss": 0.5988,
+      "step": 21948
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1251193045671647e-06,
+      "loss": 0.5228,
+      "step": 21949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1243963620476415e-06,
+      "loss": 0.5886,
+      "step": 21950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.123673527903345e-06,
+      "loss": 0.5377,
+      "step": 21951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.122950802144219e-06,
+      "loss": 0.542,
+      "step": 21952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.122228184780213e-06,
+      "loss": 0.6441,
+      "step": 21953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1215056758212703e-06,
+      "loss": 0.5546,
+      "step": 21954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1207832752773293e-06,
+      "loss": 0.6326,
+      "step": 21955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.120060983158336e-06,
+      "loss": 0.5921,
+      "step": 21956
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1193387994742275e-06,
+      "loss": 0.5963,
+      "step": 21957
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.118616724234942e-06,
+      "loss": 0.5795,
+      "step": 21958
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1178947574504128e-06,
+      "loss": 0.5932,
+      "step": 21959
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1171728991305797e-06,
+      "loss": 0.4937,
+      "step": 21960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1164511492853744e-06,
+      "loss": 0.6221,
+      "step": 21961
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1157295079247287e-06,
+      "loss": 0.6006,
+      "step": 21962
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1150079750585697e-06,
+      "loss": 0.5862,
+      "step": 21963
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1142865506968315e-06,
+      "loss": 0.4984,
+      "step": 21964
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1135652348494405e-06,
+      "loss": 0.5945,
+      "step": 21965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1128440275263175e-06,
+      "loss": 0.5376,
+      "step": 21966
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1121229287373945e-06,
+      "loss": 0.5808,
+      "step": 21967
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.111401938492591e-06,
+      "loss": 0.6735,
+      "step": 21968
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1106810568018275e-06,
+      "loss": 0.5977,
+      "step": 21969
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.109960283675023e-06,
+      "loss": 0.5498,
+      "step": 21970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1092396191220997e-06,
+      "loss": 0.5703,
+      "step": 21971
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1085190631529738e-06,
+      "loss": 0.5342,
+      "step": 21972
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1077986157775586e-06,
+      "loss": 0.591,
+      "step": 21973
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1070782770057674e-06,
+      "loss": 0.6017,
+      "step": 21974
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1063580468475165e-06,
+      "loss": 0.5839,
+      "step": 21975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1056379253127144e-06,
+      "loss": 0.5618,
+      "step": 21976
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1049179124112694e-06,
+      "loss": 0.5551,
+      "step": 21977
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1041980081530932e-06,
+      "loss": 0.6666,
+      "step": 21978
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.10347821254809e-06,
+      "loss": 0.5641,
+      "step": 21979
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1027585256061654e-06,
+      "loss": 0.625,
+      "step": 21980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1020389473372206e-06,
+      "loss": 0.5424,
+      "step": 21981
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1013194777511615e-06,
+      "loss": 0.5544,
+      "step": 21982
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.100600116857887e-06,
+      "loss": 0.5472,
+      "step": 21983
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0998808646672964e-06,
+      "loss": 0.5336,
+      "step": 21984
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0991617211892836e-06,
+      "loss": 0.6105,
+      "step": 21985
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0984426864337505e-06,
+      "loss": 0.5296,
+      "step": 21986
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0977237604105883e-06,
+      "loss": 0.5334,
+      "step": 21987
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.097004943129689e-06,
+      "loss": 0.5207,
+      "step": 21988
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0962862346009483e-06,
+      "loss": 0.5695,
+      "step": 21989
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0955676348342533e-06,
+      "loss": 0.6133,
+      "step": 21990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0948491438394934e-06,
+      "loss": 0.4854,
+      "step": 21991
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.094130761626553e-06,
+      "loss": 0.6204,
+      "step": 21992
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0934124882053218e-06,
+      "loss": 0.6282,
+      "step": 21993
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.092694323585683e-06,
+      "loss": 0.5506,
+      "step": 21994
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.091976267777518e-06,
+      "loss": 0.6204,
+      "step": 21995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0912583207907055e-06,
+      "loss": 0.5454,
+      "step": 21996
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0905404826351305e-06,
+      "loss": 0.6365,
+      "step": 21997
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.089822753320668e-06,
+      "loss": 0.6068,
+      "step": 21998
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.089105132857193e-06,
+      "loss": 0.5505,
+      "step": 21999
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.088387621254584e-06,
+      "loss": 0.518,
+      "step": 22000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.087670218522714e-06,
+      "loss": 0.5563,
+      "step": 22001
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0869529246714536e-06,
+      "loss": 0.5798,
+      "step": 22002
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0862357397106727e-06,
+      "loss": 0.5373,
+      "step": 22003
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0855186636502424e-06,
+      "loss": 0.5005,
+      "step": 22004
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.084801696500031e-06,
+      "loss": 0.5848,
+      "step": 22005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0840848382698998e-06,
+      "loss": 0.6126,
+      "step": 22006
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0833680889697194e-06,
+      "loss": 0.6261,
+      "step": 22007
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0826514486093496e-06,
+      "loss": 0.5939,
+      "step": 22008
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0819349171986536e-06,
+      "loss": 0.5032,
+      "step": 22009
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.081218494747488e-06,
+      "loss": 0.5749,
+      "step": 22010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0805021812657168e-06,
+      "loss": 0.5609,
+      "step": 22011
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0797859767631936e-06,
+      "loss": 0.5914,
+      "step": 22012
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0790698812497744e-06,
+      "loss": 0.5815,
+      "step": 22013
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.078353894735312e-06,
+      "loss": 0.5495,
+      "step": 22014
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0776380172296628e-06,
+      "loss": 0.5813,
+      "step": 22015
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0769222487426753e-06,
+      "loss": 0.5957,
+      "step": 22016
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0762065892841967e-06,
+      "loss": 0.5775,
+      "step": 22017
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0754910388640814e-06,
+      "loss": 0.6059,
+      "step": 22018
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0747755974921714e-06,
+      "loss": 0.5724,
+      "step": 22019
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.074060265178314e-06,
+      "loss": 0.5211,
+      "step": 22020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0733450419323477e-06,
+      "loss": 0.6156,
+      "step": 22021
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.072629927764125e-06,
+      "loss": 0.6175,
+      "step": 22022
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0719149226834765e-06,
+      "loss": 0.5556,
+      "step": 22023
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.071200026700243e-06,
+      "loss": 0.5331,
+      "step": 22024
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.070485239824267e-06,
+      "loss": 0.5364,
+      "step": 22025
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0697705620653806e-06,
+      "loss": 0.6335,
+      "step": 22026
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.069055993433421e-06,
+      "loss": 0.5307,
+      "step": 22027
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.068341533938215e-06,
+      "loss": 0.5797,
+      "step": 22028
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.067627183589602e-06,
+      "loss": 0.5506,
+      "step": 22029
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.06691294239741e-06,
+      "loss": 0.5573,
+      "step": 22030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0661988103714637e-06,
+      "loss": 0.5667,
+      "step": 22031
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0654847875215944e-06,
+      "loss": 0.4896,
+      "step": 22032
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.064770873857628e-06,
+      "loss": 0.6048,
+      "step": 22033
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0640570693893856e-06,
+      "loss": 0.599,
+      "step": 22034
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0633433741266884e-06,
+      "loss": 0.5966,
+      "step": 22035
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.062629788079363e-06,
+      "loss": 0.5743,
+      "step": 22036
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0619163112572262e-06,
+      "loss": 0.5255,
+      "step": 22037
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0612029436700963e-06,
+      "loss": 0.6283,
+      "step": 22038
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0604896853277867e-06,
+      "loss": 0.6105,
+      "step": 22039
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.059776536240118e-06,
+      "loss": 0.5811,
+      "step": 22040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0590634964169008e-06,
+      "loss": 0.5701,
+      "step": 22041
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0583505658679447e-06,
+      "loss": 0.5622,
+      "step": 22042
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.057637744603066e-06,
+      "loss": 0.5153,
+      "step": 22043
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.05692503263207e-06,
+      "loss": 0.5215,
+      "step": 22044
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.056212429964766e-06,
+      "loss": 0.5414,
+      "step": 22045
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.055499936610955e-06,
+      "loss": 0.5348,
+      "step": 22046
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0547875525804496e-06,
+      "loss": 0.5811,
+      "step": 22047
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0540752778830475e-06,
+      "loss": 0.6417,
+      "step": 22048
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0533631125285504e-06,
+      "loss": 0.5103,
+      "step": 22049
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.05265105652676e-06,
+      "loss": 0.5871,
+      "step": 22050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.051939109887475e-06,
+      "loss": 0.573,
+      "step": 22051
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.051227272620493e-06,
+      "loss": 0.613,
+      "step": 22052
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.050515544735604e-06,
+      "loss": 0.5707,
+      "step": 22053
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0498039262426096e-06,
+      "loss": 0.5311,
+      "step": 22054
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.049092417151298e-06,
+      "loss": 0.5913,
+      "step": 22055
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0483810174714613e-06,
+      "loss": 0.5048,
+      "step": 22056
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0476697272128864e-06,
+      "loss": 0.5243,
+      "step": 22057
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.046958546385366e-06,
+      "loss": 0.6039,
+      "step": 22058
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0462474749986837e-06,
+      "loss": 0.5736,
+      "step": 22059
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0455365130626228e-06,
+      "loss": 0.5581,
+      "step": 22060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.044825660586972e-06,
+      "loss": 0.5388,
+      "step": 22061
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.044114917581509e-06,
+      "loss": 0.6043,
+      "step": 22062
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0434042840560165e-06,
+      "loss": 0.5893,
+      "step": 22063
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.042693760020269e-06,
+      "loss": 0.5782,
+      "step": 22064
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.04198334548405e-06,
+      "loss": 0.5165,
+      "step": 22065
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0412730404571325e-06,
+      "loss": 0.5647,
+      "step": 22066
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.040562844949292e-06,
+      "loss": 0.6181,
+      "step": 22067
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0398527589702967e-06,
+      "loss": 0.5692,
+      "step": 22068
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0391427825299246e-06,
+      "loss": 0.5315,
+      "step": 22069
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0384329156379436e-06,
+      "loss": 0.5601,
+      "step": 22070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.037723158304119e-06,
+      "loss": 0.5817,
+      "step": 22071
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0370135105382215e-06,
+      "loss": 0.5567,
+      "step": 22072
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0363039723500155e-06,
+      "loss": 0.5961,
+      "step": 22073
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.035594543749264e-06,
+      "loss": 0.5202,
+      "step": 22074
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0348852247457273e-06,
+      "loss": 0.5262,
+      "step": 22075
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.034176015349172e-06,
+      "loss": 0.5743,
+      "step": 22076
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.033466915569353e-06,
+      "loss": 0.5972,
+      "step": 22077
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.03275792541603e-06,
+      "loss": 0.6138,
+      "step": 22078
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.032049044898956e-06,
+      "loss": 0.5225,
+      "step": 22079
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.031340274027891e-06,
+      "loss": 0.5116,
+      "step": 22080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.030631612812586e-06,
+      "loss": 0.6369,
+      "step": 22081
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0299230612627885e-06,
+      "loss": 0.5702,
+      "step": 22082
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.029214619388257e-06,
+      "loss": 0.6288,
+      "step": 22083
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.028506287198735e-06,
+      "loss": 0.5742,
+      "step": 22084
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.027798064703971e-06,
+      "loss": 0.6094,
+      "step": 22085
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.027089951913709e-06,
+      "loss": 0.575,
+      "step": 22086
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0263819488376956e-06,
+      "loss": 0.5825,
+      "step": 22087
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0256740554856747e-06,
+      "loss": 0.5889,
+      "step": 22088
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0249662718673844e-06,
+      "loss": 0.589,
+      "step": 22089
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0242585979925633e-06,
+      "loss": 0.5563,
+      "step": 22090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.023551033870955e-06,
+      "loss": 0.5807,
+      "step": 22091
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0228435795122937e-06,
+      "loss": 0.5512,
+      "step": 22092
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0221362349263106e-06,
+      "loss": 0.5848,
+      "step": 22093
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0214290001227457e-06,
+      "loss": 0.5775,
+      "step": 22094
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0207218751113287e-06,
+      "loss": 0.5452,
+      "step": 22095
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.020014859901789e-06,
+      "loss": 0.6176,
+      "step": 22096
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.019307954503855e-06,
+      "loss": 0.6354,
+      "step": 22097
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.018601158927258e-06,
+      "loss": 0.5691,
+      "step": 22098
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.017894473181723e-06,
+      "loss": 0.5787,
+      "step": 22099
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.017187897276972e-06,
+      "loss": 0.5899,
+      "step": 22100
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0164814312227288e-06,
+      "loss": 0.6084,
+      "step": 22101
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.015775075028719e-06,
+      "loss": 0.6102,
+      "step": 22102
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.015068828704658e-06,
+      "loss": 0.6287,
+      "step": 22103
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0143626922602643e-06,
+      "loss": 0.5671,
+      "step": 22104
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0136566657052592e-06,
+      "loss": 0.547,
+      "step": 22105
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.012950749049356e-06,
+      "loss": 0.5964,
+      "step": 22106
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0122449423022694e-06,
+      "loss": 0.5346,
+      "step": 22107
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.011539245473707e-06,
+      "loss": 0.6118,
+      "step": 22108
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0108336585733877e-06,
+      "loss": 0.5542,
+      "step": 22109
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.010128181611016e-06,
+      "loss": 0.5766,
+      "step": 22110
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.009422814596299e-06,
+      "loss": 0.5481,
+      "step": 22111
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0087175575389484e-06,
+      "loss": 0.5794,
+      "step": 22112
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0080124104486653e-06,
+      "loss": 0.6126,
+      "step": 22113
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0073073733351534e-06,
+      "loss": 0.5667,
+      "step": 22114
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0066024462081124e-06,
+      "loss": 0.6062,
+      "step": 22115
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0058976290772493e-06,
+      "loss": 0.5716,
+      "step": 22116
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.005192921952258e-06,
+      "loss": 0.5678,
+      "step": 22117
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.004488324842837e-06,
+      "loss": 0.5778,
+      "step": 22118
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0037838377586793e-06,
+      "loss": 0.605,
+      "step": 22119
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.003079460709485e-06,
+      "loss": 0.5449,
+      "step": 22120
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0023751937049442e-06,
+      "loss": 0.6267,
+      "step": 22121
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0016710367547455e-06,
+      "loss": 0.5489,
+      "step": 22122
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0009669898685836e-06,
+      "loss": 0.5114,
+      "step": 22123
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.000263053056145e-06,
+      "loss": 0.567,
+      "step": 22124
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9995592263271156e-06,
+      "loss": 0.6386,
+      "step": 22125
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9988555096911775e-06,
+      "loss": 0.6453,
+      "step": 22126
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.998151903158022e-06,
+      "loss": 0.5844,
+      "step": 22127
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9974484067373267e-06,
+      "loss": 0.6027,
+      "step": 22128
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9967450204387738e-06,
+      "loss": 0.5933,
+      "step": 22129
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.996041744272038e-06,
+      "loss": 0.5446,
+      "step": 22130
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.995338578246804e-06,
+      "loss": 0.5606,
+      "step": 22131
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9946355223727443e-06,
+      "loss": 0.5737,
+      "step": 22132
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9939325766595317e-06,
+      "loss": 0.6024,
+      "step": 22133
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.993229741116843e-06,
+      "loss": 0.5687,
+      "step": 22134
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.992527015754351e-06,
+      "loss": 0.5522,
+      "step": 22135
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.991824400581721e-06,
+      "loss": 0.6397,
+      "step": 22136
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.991121895608622e-06,
+      "loss": 0.4732,
+      "step": 22137
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.990419500844725e-06,
+      "loss": 0.5821,
+      "step": 22138
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9897172162996938e-06,
+      "loss": 0.5704,
+      "step": 22139
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.989015041983192e-06,
+      "loss": 0.5517,
+      "step": 22140
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9883129779048803e-06,
+      "loss": 0.5339,
+      "step": 22141
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.987611024074424e-06,
+      "loss": 0.5958,
+      "step": 22142
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9869091805014806e-06,
+      "loss": 0.5478,
+      "step": 22143
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.986207447195705e-06,
+      "loss": 0.5966,
+      "step": 22144
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.985505824166759e-06,
+      "loss": 0.552,
+      "step": 22145
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.984804311424294e-06,
+      "loss": 0.6343,
+      "step": 22146
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.984102908977966e-06,
+      "loss": 0.6156,
+      "step": 22147
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.983401616837423e-06,
+      "loss": 0.5345,
+      "step": 22148
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9827004350123193e-06,
+      "loss": 0.5358,
+      "step": 22149
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9819993635123014e-06,
+      "loss": 0.5979,
+      "step": 22150
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9812984023470196e-06,
+      "loss": 0.5708,
+      "step": 22151
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.980597551526113e-06,
+      "loss": 0.6121,
+      "step": 22152
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.979896811059234e-06,
+      "loss": 0.6088,
+      "step": 22153
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.979196180956022e-06,
+      "loss": 0.5864,
+      "step": 22154
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9784956612261163e-06,
+      "loss": 0.568,
+      "step": 22155
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.977795251879161e-06,
+      "loss": 0.5332,
+      "step": 22156
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9770949529247906e-06,
+      "loss": 0.5961,
+      "step": 22157
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9763947643726444e-06,
+      "loss": 0.5651,
+      "step": 22158
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9756946862323534e-06,
+      "loss": 0.5545,
+      "step": 22159
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9749947185135576e-06,
+      "loss": 0.5931,
+      "step": 22160
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.974294861225885e-06,
+      "loss": 0.5607,
+      "step": 22161
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.973595114378968e-06,
+      "loss": 0.5128,
+      "step": 22162
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.972895477982432e-06,
+      "loss": 0.6331,
+      "step": 22163
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.972195952045909e-06,
+      "loss": 0.5755,
+      "step": 22164
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.971496536579025e-06,
+      "loss": 0.5738,
+      "step": 22165
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9707972315913993e-06,
+      "loss": 0.6249,
+      "step": 22166
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9700980370926616e-06,
+      "loss": 0.5862,
+      "step": 22167
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.969398953092431e-06,
+      "loss": 0.5811,
+      "step": 22168
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.968699979600327e-06,
+      "loss": 0.5729,
+      "step": 22169
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.968001116625966e-06,
+      "loss": 0.5969,
+      "step": 22170
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.967302364178969e-06,
+      "loss": 0.562,
+      "step": 22171
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9666037222689504e-06,
+      "loss": 0.5658,
+      "step": 22172
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9659051909055194e-06,
+      "loss": 0.6113,
+      "step": 22173
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.965206770098297e-06,
+      "loss": 0.6348,
+      "step": 22174
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9645084598568876e-06,
+      "loss": 0.5948,
+      "step": 22175
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9638102601909027e-06,
+      "loss": 0.4884,
+      "step": 22176
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9631121711099475e-06,
+      "loss": 0.5188,
+      "step": 22177
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.962414192623634e-06,
+      "loss": 0.5841,
+      "step": 22178
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9617163247415627e-06,
+      "loss": 0.5563,
+      "step": 22179
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9610185674733386e-06,
+      "loss": 0.626,
+      "step": 22180
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.960320920828559e-06,
+      "loss": 0.5492,
+      "step": 22181
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.959623384816831e-06,
+      "loss": 0.6004,
+      "step": 22182
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9589259594477503e-06,
+      "loss": 0.5803,
+      "step": 22183
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.958228644730912e-06,
+      "loss": 0.5663,
+      "step": 22184
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.957531440675915e-06,
+      "loss": 0.6449,
+      "step": 22185
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9568343472923524e-06,
+      "loss": 0.4954,
+      "step": 22186
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9561373645898164e-06,
+      "loss": 0.519,
+      "step": 22187
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9554404925778957e-06,
+      "loss": 0.6392,
+      "step": 22188
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9547437312661856e-06,
+      "loss": 0.5535,
+      "step": 22189
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9540470806642707e-06,
+      "loss": 0.5872,
+      "step": 22190
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9533505407817377e-06,
+      "loss": 0.521,
+      "step": 22191
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.952654111628168e-06,
+      "loss": 0.5617,
+      "step": 22192
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9519577932131527e-06,
+      "loss": 0.6202,
+      "step": 22193
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.951261585546269e-06,
+      "loss": 0.5817,
+      "step": 22194
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.950565488637096e-06,
+      "loss": 0.554,
+      "step": 22195
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9498695024952173e-06,
+      "loss": 0.5429,
+      "step": 22196
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9491736271302077e-06,
+      "loss": 0.5708,
+      "step": 22197
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9484778625516434e-06,
+      "loss": 0.5879,
+      "step": 22198
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.947782208769096e-06,
+      "loss": 0.5756,
+      "step": 22199
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9470866657921416e-06,
+      "loss": 0.6348,
+      "step": 22200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9463912336303526e-06,
+      "loss": 0.6175,
+      "step": 22201
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.945695912293295e-06,
+      "loss": 0.5603,
+      "step": 22202
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.945000701790537e-06,
+      "loss": 0.5496,
+      "step": 22203
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9443056021316496e-06,
+      "loss": 0.5783,
+      "step": 22204
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9436106133261955e-06,
+      "loss": 0.5477,
+      "step": 22205
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9429157353837356e-06,
+      "loss": 0.6251,
+      "step": 22206
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.942220968313838e-06,
+      "loss": 0.5915,
+      "step": 22207
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.94152631212606e-06,
+      "loss": 0.6263,
+      "step": 22208
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9408317668299605e-06,
+      "loss": 0.5516,
+      "step": 22209
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9401373324350947e-06,
+      "loss": 0.5932,
+      "step": 22210
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.939443008951023e-06,
+      "loss": 0.5991,
+      "step": 22211
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9387487963872988e-06,
+      "loss": 0.5313,
+      "step": 22212
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.938054694753475e-06,
+      "loss": 0.5484,
+      "step": 22213
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9373607040590992e-06,
+      "loss": 0.6003,
+      "step": 22214
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9366668243137266e-06,
+      "loss": 0.602,
+      "step": 22215
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.935973055526904e-06,
+      "loss": 0.5543,
+      "step": 22216
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.935279397708175e-06,
+      "loss": 0.516,
+      "step": 22217
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9345858508670913e-06,
+      "loss": 0.5474,
+      "step": 22218
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9338924150131933e-06,
+      "loss": 0.6298,
+      "step": 22219
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9331990901560216e-06,
+      "loss": 0.5944,
+      "step": 22220
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9325058763051176e-06,
+      "loss": 0.5722,
+      "step": 22221
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.931812773470023e-06,
+      "loss": 0.6192,
+      "step": 22222
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9311197816602746e-06,
+      "loss": 0.538,
+      "step": 22223
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9304269008854073e-06,
+      "loss": 0.5464,
+      "step": 22224
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9297341311549543e-06,
+      "loss": 0.5679,
+      "step": 22225
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9290414724784535e-06,
+      "loss": 0.5086,
+      "step": 22226
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.928348924865433e-06,
+      "loss": 0.5753,
+      "step": 22227
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.927656488325421e-06,
+      "loss": 0.6108,
+      "step": 22228
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9269641628679527e-06,
+      "loss": 0.5391,
+      "step": 22229
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9262719485025506e-06,
+      "loss": 0.6046,
+      "step": 22230
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.92557984523874e-06,
+      "loss": 0.551,
+      "step": 22231
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9248878530860436e-06,
+      "loss": 0.6233,
+      "step": 22232
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9241959720539902e-06,
+      "loss": 0.6042,
+      "step": 22233
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9235042021520946e-06,
+      "loss": 0.5778,
+      "step": 22234
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9228125433898748e-06,
+      "loss": 0.5747,
+      "step": 22235
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.922120995776854e-06,
+      "loss": 0.6234,
+      "step": 22236
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9214295593225464e-06,
+      "loss": 0.6185,
+      "step": 22237
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.920738234036463e-06,
+      "loss": 0.5596,
+      "step": 22238
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.920047019928125e-06,
+      "loss": 0.6126,
+      "step": 22239
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9193559170070387e-06,
+      "loss": 0.5526,
+      "step": 22240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.918664925282715e-06,
+      "loss": 0.556,
+      "step": 22241
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.91797404476466e-06,
+      "loss": 0.6115,
+      "step": 22242
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.917283275462387e-06,
+      "loss": 0.5704,
+      "step": 22243
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9165926173853977e-06,
+      "loss": 0.611,
+      "step": 22244
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9159020705431974e-06,
+      "loss": 0.5777,
+      "step": 22245
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9152116349452845e-06,
+      "loss": 0.5298,
+      "step": 22246
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.914521310601166e-06,
+      "loss": 0.5752,
+      "step": 22247
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9138310975203402e-06,
+      "loss": 0.562,
+      "step": 22248
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9131409957122993e-06,
+      "loss": 0.6172,
+      "step": 22249
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9124510051865475e-06,
+      "loss": 0.5552,
+      "step": 22250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9117611259525758e-06,
+      "loss": 0.6053,
+      "step": 22251
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.911071358019878e-06,
+      "loss": 0.5989,
+      "step": 22252
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.910381701397944e-06,
+      "loss": 0.591,
+      "step": 22253
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.909692156096268e-06,
+      "loss": 0.6173,
+      "step": 22254
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9090027221243366e-06,
+      "loss": 0.6052,
+      "step": 22255
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.908313399491637e-06,
+      "loss": 0.5464,
+      "step": 22256
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.907624188207653e-06,
+      "loss": 0.6219,
+      "step": 22257
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9069350882818726e-06,
+      "loss": 0.6332,
+      "step": 22258
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9062460997237775e-06,
+      "loss": 0.5531,
+      "step": 22259
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9055572225428442e-06,
+      "loss": 0.5927,
+      "step": 22260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9048684567485587e-06,
+      "loss": 0.5439,
+      "step": 22261
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9041798023503965e-06,
+      "loss": 0.5519,
+      "step": 22262
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9034912593578348e-06,
+      "loss": 0.5894,
+      "step": 22263
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9028028277803435e-06,
+      "loss": 0.5919,
+      "step": 22264
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9021145076274028e-06,
+      "loss": 0.5591,
+      "step": 22265
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.901426298908483e-06,
+      "loss": 0.6098,
+      "step": 22266
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.90073820163305e-06,
+      "loss": 0.5609,
+      "step": 22267
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9000502158105783e-06,
+      "loss": 0.5252,
+      "step": 22268
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8993623414505335e-06,
+      "loss": 0.6153,
+      "step": 22269
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8986745785623806e-06,
+      "loss": 0.5603,
+      "step": 22270
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8979869271555817e-06,
+      "loss": 0.5425,
+      "step": 22271
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8972993872396038e-06,
+      "loss": 0.5434,
+      "step": 22272
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8966119588239064e-06,
+      "loss": 0.5632,
+      "step": 22273
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8959246419179479e-06,
+      "loss": 0.5792,
+      "step": 22274
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8952374365311855e-06,
+      "loss": 0.6168,
+      "step": 22275
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8945503426730794e-06,
+      "loss": 0.5967,
+      "step": 22276
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8938633603530832e-06,
+      "loss": 0.5325,
+      "step": 22277
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8931764895806458e-06,
+      "loss": 0.5261,
+      "step": 22278
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8924897303652278e-06,
+      "loss": 0.5018,
+      "step": 22279
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8918030827162737e-06,
+      "loss": 0.5573,
+      "step": 22280
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8911165466432346e-06,
+      "loss": 0.5416,
+      "step": 22281
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8904301221555533e-06,
+      "loss": 0.5787,
+      "step": 22282
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.889743809262682e-06,
+      "loss": 0.5621,
+      "step": 22283
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.889057607974063e-06,
+      "loss": 0.6195,
+      "step": 22284
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.888371518299137e-06,
+      "loss": 0.6116,
+      "step": 22285
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8876855402473448e-06,
+      "loss": 0.5671,
+      "step": 22286
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8869996738281304e-06,
+      "loss": 0.5811,
+      "step": 22287
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8863139190509283e-06,
+      "loss": 0.548,
+      "step": 22288
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8856282759251742e-06,
+      "loss": 0.6153,
+      "step": 22289
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8849427444603063e-06,
+      "loss": 0.5089,
+      "step": 22290
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.884257324665758e-06,
+      "loss": 0.5828,
+      "step": 22291
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8835720165509598e-06,
+      "loss": 0.5698,
+      "step": 22292
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8828868201253391e-06,
+      "loss": 0.5531,
+      "step": 22293
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8822017353983312e-06,
+      "loss": 0.5704,
+      "step": 22294
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8815167623793606e-06,
+      "loss": 0.6542,
+      "step": 22295
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.880831901077853e-06,
+      "loss": 0.5923,
+      "step": 22296
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.88014715150323e-06,
+      "loss": 0.6078,
+      "step": 22297
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8794625136649181e-06,
+      "loss": 0.6127,
+      "step": 22298
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8787779875723389e-06,
+      "loss": 0.5601,
+      "step": 22299
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.878093573234907e-06,
+      "loss": 0.5336,
+      "step": 22300
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8774092706620473e-06,
+      "loss": 0.5258,
+      "step": 22301
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8767250798631731e-06,
+      "loss": 0.5328,
+      "step": 22302
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8760410008476993e-06,
+      "loss": 0.5356,
+      "step": 22303
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.875357033625037e-06,
+      "loss": 0.5266,
+      "step": 22304
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8746731782046035e-06,
+      "loss": 0.524,
+      "step": 22305
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8739894345958065e-06,
+      "loss": 0.571,
+      "step": 22306
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8733058028080552e-06,
+      "loss": 0.5462,
+      "step": 22307
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8726222828507546e-06,
+      "loss": 0.5526,
+      "step": 22308
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8719388747333145e-06,
+      "loss": 0.5363,
+      "step": 22309
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8712555784651375e-06,
+      "loss": 0.5971,
+      "step": 22310
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8705723940556242e-06,
+      "loss": 0.5376,
+      "step": 22311
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.86988932151418e-06,
+      "loss": 0.6009,
+      "step": 22312
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8692063608502032e-06,
+      "loss": 0.5621,
+      "step": 22313
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8685235120730904e-06,
+      "loss": 0.593,
+      "step": 22314
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8678407751922368e-06,
+      "loss": 0.5113,
+      "step": 22315
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8671581502170421e-06,
+      "loss": 0.5623,
+      "step": 22316
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8664756371568981e-06,
+      "loss": 0.5415,
+      "step": 22317
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.865793236021195e-06,
+      "loss": 0.5565,
+      "step": 22318
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.865110946819323e-06,
+      "loss": 0.531,
+      "step": 22319
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8644287695606744e-06,
+      "loss": 0.5101,
+      "step": 22320
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8637467042546343e-06,
+      "loss": 0.5538,
+      "step": 22321
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8630647509105859e-06,
+      "loss": 0.6406,
+      "step": 22322
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8623829095379188e-06,
+      "loss": 0.5849,
+      "step": 22323
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8617011801460128e-06,
+      "loss": 0.5826,
+      "step": 22324
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8610195627442507e-06,
+      "loss": 0.5984,
+      "step": 22325
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8603380573420072e-06,
+      "loss": 0.584,
+      "step": 22326
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.859656663948668e-06,
+      "loss": 0.548,
+      "step": 22327
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8589753825736045e-06,
+      "loss": 0.595,
+      "step": 22328
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8582942132261916e-06,
+      "loss": 0.561,
+      "step": 22329
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.857613155915806e-06,
+      "loss": 0.5577,
+      "step": 22330
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8569322106518184e-06,
+      "loss": 0.6223,
+      "step": 22331
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8562513774435986e-06,
+      "loss": 0.5512,
+      "step": 22332
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8555706563005126e-06,
+      "loss": 0.5294,
+      "step": 22333
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8548900472319342e-06,
+      "loss": 0.61,
+      "step": 22334
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8542095502472247e-06,
+      "loss": 0.5518,
+      "step": 22335
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8535291653557498e-06,
+      "loss": 0.555,
+      "step": 22336
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8528488925668685e-06,
+      "loss": 0.6116,
+      "step": 22337
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8521687318899483e-06,
+      "loss": 0.5621,
+      "step": 22338
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.851488683334346e-06,
+      "loss": 0.6142,
+      "step": 22339
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.850808746909416e-06,
+      "loss": 0.6476,
+      "step": 22340
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8501289226245212e-06,
+      "loss": 0.5592,
+      "step": 22341
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8494492104890137e-06,
+      "loss": 0.5924,
+      "step": 22342
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8487696105122466e-06,
+      "loss": 0.6603,
+      "step": 22343
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8480901227035696e-06,
+      "loss": 0.5964,
+      "step": 22344
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8474107470723368e-06,
+      "loss": 0.5471,
+      "step": 22345
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.846731483627897e-06,
+      "loss": 0.5832,
+      "step": 22346
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8460523323795954e-06,
+      "loss": 0.6439,
+      "step": 22347
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8453732933367762e-06,
+      "loss": 0.5456,
+      "step": 22348
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.844694366508788e-06,
+      "loss": 0.595,
+      "step": 22349
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8440155519049707e-06,
+      "loss": 0.6187,
+      "step": 22350
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.843336849534665e-06,
+      "loss": 0.582,
+      "step": 22351
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8426582594072118e-06,
+      "loss": 0.5547,
+      "step": 22352
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8419797815319495e-06,
+      "loss": 0.6393,
+      "step": 22353
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.841301415918214e-06,
+      "loss": 0.5945,
+      "step": 22354
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.840623162575338e-06,
+      "loss": 0.5499,
+      "step": 22355
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.839945021512658e-06,
+      "loss": 0.6337,
+      "step": 22356
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8392669927395047e-06,
+      "loss": 0.525,
+      "step": 22357
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8385890762652082e-06,
+      "loss": 0.624,
+      "step": 22358
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8379112720990955e-06,
+      "loss": 0.5984,
+      "step": 22359
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.837233580250497e-06,
+      "loss": 0.5459,
+      "step": 22360
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8365560007287376e-06,
+      "loss": 0.5406,
+      "step": 22361
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8358785335431383e-06,
+      "loss": 0.5748,
+      "step": 22362
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8352011787030266e-06,
+      "loss": 0.5589,
+      "step": 22363
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8345239362177203e-06,
+      "loss": 0.5366,
+      "step": 22364
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8338468060965398e-06,
+      "loss": 0.5886,
+      "step": 22365
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8331697883487999e-06,
+      "loss": 0.5349,
+      "step": 22366
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8324928829838218e-06,
+      "loss": 0.5316,
+      "step": 22367
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8318160900109194e-06,
+      "loss": 0.6354,
+      "step": 22368
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8311394094394041e-06,
+      "loss": 0.5993,
+      "step": 22369
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8304628412785852e-06,
+      "loss": 0.5751,
+      "step": 22370
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8297863855377796e-06,
+      "loss": 0.5132,
+      "step": 22371
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.829110042226292e-06,
+      "loss": 0.5556,
+      "step": 22372
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8284338113534273e-06,
+      "loss": 0.6119,
+      "step": 22373
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8277576929284956e-06,
+      "loss": 0.536,
+      "step": 22374
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.827081686960799e-06,
+      "loss": 0.5938,
+      "step": 22375
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8264057934596403e-06,
+      "loss": 0.6204,
+      "step": 22376
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.825730012434316e-06,
+      "loss": 0.5038,
+      "step": 22377
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8250543438941326e-06,
+      "loss": 0.5421,
+      "step": 22378
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8243787878483853e-06,
+      "loss": 0.6573,
+      "step": 22379
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.823703344306369e-06,
+      "loss": 0.5831,
+      "step": 22380
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8230280132773758e-06,
+      "loss": 0.6068,
+      "step": 22381
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8223527947707053e-06,
+      "loss": 0.5384,
+      "step": 22382
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8216776887956456e-06,
+      "loss": 0.5048,
+      "step": 22383
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8210026953614846e-06,
+      "loss": 0.613,
+      "step": 22384
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8203278144775162e-06,
+      "loss": 0.5984,
+      "step": 22385
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8196530461530238e-06,
+      "loss": 0.6472,
+      "step": 22386
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8189783903972934e-06,
+      "loss": 0.5754,
+      "step": 22387
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8183038472196057e-06,
+      "loss": 0.5717,
+      "step": 22388
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8176294166292497e-06,
+      "loss": 0.6044,
+      "step": 22389
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8169550986355022e-06,
+      "loss": 0.5851,
+      "step": 22390
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8162808932476406e-06,
+      "loss": 0.5399,
+      "step": 22391
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8156068004749472e-06,
+      "loss": 0.6051,
+      "step": 22392
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8149328203266947e-06,
+      "loss": 0.58,
+      "step": 22393
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.81425895281216e-06,
+      "loss": 0.6224,
+      "step": 22394
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8135851979406115e-06,
+      "loss": 0.6024,
+      "step": 22395
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8129115557213262e-06,
+      "loss": 0.6003,
+      "step": 22396
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8122380261635719e-06,
+      "loss": 0.5885,
+      "step": 22397
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.811564609276617e-06,
+      "loss": 0.5641,
+      "step": 22398
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8108913050697252e-06,
+      "loss": 0.5271,
+      "step": 22399
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8102181135521668e-06,
+      "loss": 0.5867,
+      "step": 22400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8095450347332045e-06,
+      "loss": 0.5107,
+      "step": 22401
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8088720686220962e-06,
+      "loss": 0.5461,
+      "step": 22402
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.808199215228108e-06,
+      "loss": 0.593,
+      "step": 22403
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8075264745604958e-06,
+      "loss": 0.5456,
+      "step": 22404
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.806853846628519e-06,
+      "loss": 0.4994,
+      "step": 22405
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.80618133144143e-06,
+      "loss": 0.5729,
+      "step": 22406
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.805508929008487e-06,
+      "loss": 0.5744,
+      "step": 22407
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8048366393389427e-06,
+      "loss": 0.549,
+      "step": 22408
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8041644624420462e-06,
+      "loss": 0.5931,
+      "step": 22409
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.803492398327046e-06,
+      "loss": 0.6215,
+      "step": 22410
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.802820447003195e-06,
+      "loss": 0.6259,
+      "step": 22411
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8021486084797368e-06,
+      "loss": 0.6436,
+      "step": 22412
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8014768827659158e-06,
+      "loss": 0.5283,
+      "step": 22413
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8008052698709776e-06,
+      "loss": 0.5561,
+      "step": 22414
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8001337698041643e-06,
+      "loss": 0.6102,
+      "step": 22415
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7994623825747159e-06,
+      "loss": 0.6277,
+      "step": 22416
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7987911081918673e-06,
+      "loss": 0.6211,
+      "step": 22417
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7981199466648613e-06,
+      "loss": 0.5496,
+      "step": 22418
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7974488980029325e-06,
+      "loss": 0.5715,
+      "step": 22419
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7967779622153147e-06,
+      "loss": 0.5332,
+      "step": 22420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7961071393112372e-06,
+      "loss": 0.4968,
+      "step": 22421
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.795436429299936e-06,
+      "loss": 0.5148,
+      "step": 22422
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7947658321906393e-06,
+      "loss": 0.5328,
+      "step": 22423
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7940953479925715e-06,
+      "loss": 0.5467,
+      "step": 22424
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7934249767149647e-06,
+      "loss": 0.5983,
+      "step": 22425
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7927547183670413e-06,
+      "loss": 0.5949,
+      "step": 22426
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7920845729580249e-06,
+      "loss": 0.6159,
+      "step": 22427
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.791414540497134e-06,
+      "loss": 0.5515,
+      "step": 22428
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7907446209935941e-06,
+      "loss": 0.5987,
+      "step": 22429
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7900748144566216e-06,
+      "loss": 0.612,
+      "step": 22430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7894051208954334e-06,
+      "loss": 0.5953,
+      "step": 22431
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.788735540319243e-06,
+      "loss": 0.6597,
+      "step": 22432
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.788066072737269e-06,
+      "loss": 0.5944,
+      "step": 22433
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7873967181587215e-06,
+      "loss": 0.5442,
+      "step": 22434
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7867274765928088e-06,
+      "loss": 0.5412,
+      "step": 22435
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7860583480487449e-06,
+      "loss": 0.5481,
+      "step": 22436
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7853893325357364e-06,
+      "loss": 0.6456,
+      "step": 22437
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7847204300629873e-06,
+      "loss": 0.5588,
+      "step": 22438
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7840516406397025e-06,
+      "loss": 0.5745,
+      "step": 22439
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7833829642750888e-06,
+      "loss": 0.5705,
+      "step": 22440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.782714400978345e-06,
+      "loss": 0.6061,
+      "step": 22441
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7820459507586708e-06,
+      "loss": 0.5748,
+      "step": 22442
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7813776136252626e-06,
+      "loss": 0.613,
+      "step": 22443
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.780709389587326e-06,
+      "loss": 0.6015,
+      "step": 22444
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7800412786540445e-06,
+      "loss": 0.613,
+      "step": 22445
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7793732808346209e-06,
+      "loss": 0.5705,
+      "step": 22446
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.778705396138244e-06,
+      "loss": 0.5598,
+      "step": 22447
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7780376245741038e-06,
+      "loss": 0.6227,
+      "step": 22448
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7773699661513887e-06,
+      "loss": 0.5505,
+      "step": 22449
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7767024208792905e-06,
+      "loss": 0.6011,
+      "step": 22450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7760349887669925e-06,
+      "loss": 0.6016,
+      "step": 22451
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7753676698236789e-06,
+      "loss": 0.6238,
+      "step": 22452
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7747004640585307e-06,
+      "loss": 0.5425,
+      "step": 22453
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7740333714807345e-06,
+      "loss": 0.5707,
+      "step": 22454
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.773366392099466e-06,
+      "loss": 0.6346,
+      "step": 22455
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7726995259239034e-06,
+      "loss": 0.6061,
+      "step": 22456
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7720327729632258e-06,
+      "loss": 0.5844,
+      "step": 22457
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7713661332266075e-06,
+      "loss": 0.5392,
+      "step": 22458
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7706996067232208e-06,
+      "loss": 0.5719,
+      "step": 22459
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.770033193462236e-06,
+      "loss": 0.5812,
+      "step": 22460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7693668934528296e-06,
+      "loss": 0.5786,
+      "step": 22461
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7687007067041661e-06,
+      "loss": 0.5691,
+      "step": 22462
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7680346332254128e-06,
+      "loss": 0.5361,
+      "step": 22463
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7673686730257345e-06,
+      "loss": 0.5317,
+      "step": 22464
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7667028261142993e-06,
+      "loss": 0.5735,
+      "step": 22465
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7660370925002678e-06,
+      "loss": 0.6133,
+      "step": 22466
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7653714721927984e-06,
+      "loss": 0.5362,
+      "step": 22467
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7647059652010557e-06,
+      "loss": 0.588,
+      "step": 22468
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7640405715341958e-06,
+      "loss": 0.57,
+      "step": 22469
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7633752912013736e-06,
+      "loss": 0.6111,
+      "step": 22470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7627101242117417e-06,
+      "loss": 0.5728,
+      "step": 22471
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7620450705744597e-06,
+      "loss": 0.5478,
+      "step": 22472
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7613801302986756e-06,
+      "loss": 0.5609,
+      "step": 22473
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7607153033935398e-06,
+      "loss": 0.594,
+      "step": 22474
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7600505898681996e-06,
+      "loss": 0.5767,
+      "step": 22475
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7593859897318055e-06,
+      "loss": 0.598,
+      "step": 22476
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7587215029935e-06,
+      "loss": 0.614,
+      "step": 22477
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7580571296624272e-06,
+      "loss": 0.6218,
+      "step": 22478
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7573928697477316e-06,
+      "loss": 0.5847,
+      "step": 22479
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7567287232585527e-06,
+      "loss": 0.5259,
+      "step": 22480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7560646902040312e-06,
+      "loss": 0.5618,
+      "step": 22481
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7554007705932996e-06,
+      "loss": 0.6028,
+      "step": 22482
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7547369644355007e-06,
+      "loss": 0.581,
+      "step": 22483
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7540732717397657e-06,
+      "loss": 0.5383,
+      "step": 22484
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7534096925152277e-06,
+      "loss": 0.5614,
+      "step": 22485
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7527462267710161e-06,
+      "loss": 0.6269,
+      "step": 22486
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7520828745162666e-06,
+      "loss": 0.6233,
+      "step": 22487
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7514196357601033e-06,
+      "loss": 0.5721,
+      "step": 22488
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.750756510511651e-06,
+      "loss": 0.589,
+      "step": 22489
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.75009349878004e-06,
+      "loss": 0.5832,
+      "step": 22490
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.749430600574391e-06,
+      "loss": 0.6139,
+      "step": 22491
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7487678159038268e-06,
+      "loss": 0.5673,
+      "step": 22492
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7481051447774656e-06,
+      "loss": 0.5824,
+      "step": 22493
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.747442587204431e-06,
+      "loss": 0.5546,
+      "step": 22494
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.746780143193837e-06,
+      "loss": 0.5346,
+      "step": 22495
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7461178127547984e-06,
+      "loss": 0.5411,
+      "step": 22496
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7454555958964336e-06,
+      "loss": 0.5504,
+      "step": 22497
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.744793492627852e-06,
+      "loss": 0.6327,
+      "step": 22498
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7441315029581663e-06,
+      "loss": 0.5457,
+      "step": 22499
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7434696268964834e-06,
+      "loss": 0.5617,
+      "step": 22500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7428078644519154e-06,
+      "loss": 0.5336,
+      "step": 22501
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7421462156335657e-06,
+      "loss": 0.6005,
+      "step": 22502
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7414846804505403e-06,
+      "loss": 0.5975,
+      "step": 22503
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.74082325891194e-06,
+      "loss": 0.6142,
+      "step": 22504
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.740161951026872e-06,
+      "loss": 0.5903,
+      "step": 22505
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7395007568044331e-06,
+      "loss": 0.6289,
+      "step": 22506
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7388396762537185e-06,
+      "loss": 0.6018,
+      "step": 22507
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.738178709383832e-06,
+      "loss": 0.5891,
+      "step": 22508
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7375178562038663e-06,
+      "loss": 0.5083,
+      "step": 22509
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7368571167229154e-06,
+      "loss": 0.5821,
+      "step": 22510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7361964909500673e-06,
+      "loss": 0.5993,
+      "step": 22511
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7355359788944204e-06,
+      "loss": 0.5294,
+      "step": 22512
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7348755805650597e-06,
+      "loss": 0.6073,
+      "step": 22513
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7342152959710735e-06,
+      "loss": 0.5781,
+      "step": 22514
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7335551251215443e-06,
+      "loss": 0.5726,
+      "step": 22515
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7328950680255642e-06,
+      "loss": 0.5563,
+      "step": 22516
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7322351246922108e-06,
+      "loss": 0.5687,
+      "step": 22517
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7315752951305654e-06,
+      "loss": 0.6089,
+      "step": 22518
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7309155793497112e-06,
+      "loss": 0.5737,
+      "step": 22519
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7302559773587236e-06,
+      "loss": 0.5684,
+      "step": 22520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7295964891666817e-06,
+      "loss": 0.566,
+      "step": 22521
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7289371147826561e-06,
+      "loss": 0.5797,
+      "step": 22522
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7282778542157253e-06,
+      "loss": 0.5185,
+      "step": 22523
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7276187074749595e-06,
+      "loss": 0.6046,
+      "step": 22524
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7269596745694295e-06,
+      "loss": 0.6047,
+      "step": 22525
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7263007555082011e-06,
+      "loss": 0.5393,
+      "step": 22526
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7256419503003464e-06,
+      "loss": 0.598,
+      "step": 22527
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7249832589549297e-06,
+      "loss": 0.6111,
+      "step": 22528
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.724324681481011e-06,
+      "loss": 0.5397,
+      "step": 22529
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7236662178876595e-06,
+      "loss": 0.6037,
+      "step": 22530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7230078681839334e-06,
+      "loss": 0.5935,
+      "step": 22531
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7223496323788912e-06,
+      "loss": 0.5669,
+      "step": 22532
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.721691510481589e-06,
+      "loss": 0.6028,
+      "step": 22533
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7210335025010883e-06,
+      "loss": 0.5761,
+      "step": 22534
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7203756084464407e-06,
+      "loss": 0.5551,
+      "step": 22535
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7197178283267003e-06,
+      "loss": 0.645,
+      "step": 22536
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7190601621509151e-06,
+      "loss": 0.5825,
+      "step": 22537
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7184026099281414e-06,
+      "loss": 0.6085,
+      "step": 22538
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7177451716674242e-06,
+      "loss": 0.5402,
+      "step": 22539
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7170878473778086e-06,
+      "loss": 0.5461,
+      "step": 22540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7164306370683448e-06,
+      "loss": 0.5769,
+      "step": 22541
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7157735407480746e-06,
+      "loss": 0.5858,
+      "step": 22542
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7151165584260388e-06,
+      "loss": 0.5469,
+      "step": 22543
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7144596901112764e-06,
+      "loss": 0.5448,
+      "step": 22544
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7138029358128316e-06,
+      "loss": 0.5566,
+      "step": 22545
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7131462955397405e-06,
+      "loss": 0.5953,
+      "step": 22546
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7124897693010334e-06,
+      "loss": 0.5735,
+      "step": 22547
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7118333571057532e-06,
+      "loss": 0.5543,
+      "step": 22548
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7111770589629272e-06,
+      "loss": 0.5642,
+      "step": 22549
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7105208748815894e-06,
+      "loss": 0.566,
+      "step": 22550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7098648048707644e-06,
+      "loss": 0.5292,
+      "step": 22551
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7092088489394876e-06,
+      "loss": 0.5959,
+      "step": 22552
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7085530070967815e-06,
+      "loss": 0.5654,
+      "step": 22553
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7078972793516713e-06,
+      "loss": 0.5527,
+      "step": 22554
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7072416657131784e-06,
+      "loss": 0.5865,
+      "step": 22555
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7065861661903293e-06,
+      "loss": 0.5973,
+      "step": 22556
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7059307807921421e-06,
+      "loss": 0.5489,
+      "step": 22557
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7052755095276319e-06,
+      "loss": 0.5876,
+      "step": 22558
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7046203524058225e-06,
+      "loss": 0.5394,
+      "step": 22559
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7039653094357256e-06,
+      "loss": 0.5204,
+      "step": 22560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.703310380626355e-06,
+      "loss": 0.5347,
+      "step": 22561
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7026555659867217e-06,
+      "loss": 0.5149,
+      "step": 22562
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7020008655258413e-06,
+      "loss": 0.5648,
+      "step": 22563
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7013462792527203e-06,
+      "loss": 0.5512,
+      "step": 22564
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7006918071763657e-06,
+      "loss": 0.6364,
+      "step": 22565
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7000374493057825e-06,
+      "loss": 0.5559,
+      "step": 22566
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.699383205649978e-06,
+      "loss": 0.5339,
+      "step": 22567
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6987290762179553e-06,
+      "loss": 0.5626,
+      "step": 22568
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6980750610187124e-06,
+      "loss": 0.6239,
+      "step": 22569
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6974211600612533e-06,
+      "loss": 0.55,
+      "step": 22570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6967673733545743e-06,
+      "loss": 0.5936,
+      "step": 22571
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6961137009076723e-06,
+      "loss": 0.5921,
+      "step": 22572
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6954601427295403e-06,
+      "loss": 0.6061,
+      "step": 22573
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6948066988291756e-06,
+      "loss": 0.5853,
+      "step": 22574
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6941533692155675e-06,
+      "loss": 0.5834,
+      "step": 22575
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6935001538977092e-06,
+      "loss": 0.6323,
+      "step": 22576
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.692847052884583e-06,
+      "loss": 0.5604,
+      "step": 22577
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6921940661851844e-06,
+      "loss": 0.5564,
+      "step": 22578
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6915411938084947e-06,
+      "loss": 0.6368,
+      "step": 22579
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.690888435763497e-06,
+      "loss": 0.5472,
+      "step": 22580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6902357920591773e-06,
+      "loss": 0.608,
+      "step": 22581
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6895832627045139e-06,
+      "loss": 0.5656,
+      "step": 22582
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6889308477084887e-06,
+      "loss": 0.5599,
+      "step": 22583
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6882785470800734e-06,
+      "loss": 0.5716,
+      "step": 22584
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6876263608282528e-06,
+      "loss": 0.5714,
+      "step": 22585
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6869742889619967e-06,
+      "loss": 0.6507,
+      "step": 22586
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6863223314902788e-06,
+      "loss": 0.5502,
+      "step": 22587
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6856704884220675e-06,
+      "loss": 0.5546,
+      "step": 22588
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6850187597663392e-06,
+      "loss": 0.5895,
+      "step": 22589
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6843671455320587e-06,
+      "loss": 0.4868,
+      "step": 22590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6837156457281912e-06,
+      "loss": 0.5462,
+      "step": 22591
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.683064260363705e-06,
+      "loss": 0.5371,
+      "step": 22592
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6824129894475637e-06,
+      "loss": 0.6348,
+      "step": 22593
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6817618329887276e-06,
+      "loss": 0.5286,
+      "step": 22594
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6811107909961555e-06,
+      "loss": 0.5625,
+      "step": 22595
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6804598634788106e-06,
+      "loss": 0.606,
+      "step": 22596
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6798090504456478e-06,
+      "loss": 0.5405,
+      "step": 22597
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6791583519056243e-06,
+      "loss": 0.5583,
+      "step": 22598
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6785077678676898e-06,
+      "loss": 0.5588,
+      "step": 22599
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6778572983408026e-06,
+      "loss": 0.5945,
+      "step": 22600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6772069433339121e-06,
+      "loss": 0.5571,
+      "step": 22601
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.676556702855965e-06,
+      "loss": 0.5725,
+      "step": 22602
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6759065769159133e-06,
+      "loss": 0.5639,
+      "step": 22603
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6752565655227015e-06,
+      "loss": 0.5455,
+      "step": 22604
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6746066686852734e-06,
+      "loss": 0.6038,
+      "step": 22605
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.673956886412571e-06,
+      "loss": 0.5202,
+      "step": 22606
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6733072187135402e-06,
+      "loss": 0.577,
+      "step": 22607
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6726576655971184e-06,
+      "loss": 0.6062,
+      "step": 22608
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6720082270722415e-06,
+      "loss": 0.5334,
+      "step": 22609
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6713589031478516e-06,
+      "loss": 0.606,
+      "step": 22610
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6707096938328814e-06,
+      "loss": 0.5184,
+      "step": 22611
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.670060599136265e-06,
+      "loss": 0.5569,
+      "step": 22612
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6694116190669319e-06,
+      "loss": 0.5515,
+      "step": 22613
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6687627536338158e-06,
+      "loss": 0.6088,
+      "step": 22614
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6681140028458464e-06,
+      "loss": 0.5133,
+      "step": 22615
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6674653667119478e-06,
+      "loss": 0.6191,
+      "step": 22616
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.666816845241046e-06,
+      "loss": 0.5676,
+      "step": 22617
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6661684384420685e-06,
+      "loss": 0.5297,
+      "step": 22618
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6655201463239367e-06,
+      "loss": 0.5717,
+      "step": 22619
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6648719688955672e-06,
+      "loss": 0.5862,
+      "step": 22620
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6642239061658872e-06,
+      "loss": 0.5795,
+      "step": 22621
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6635759581438093e-06,
+      "loss": 0.5904,
+      "step": 22622
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6629281248382523e-06,
+      "loss": 0.597,
+      "step": 22623
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6622804062581276e-06,
+      "loss": 0.5743,
+      "step": 22624
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6616328024123529e-06,
+      "loss": 0.5726,
+      "step": 22625
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6609853133098375e-06,
+      "loss": 0.5801,
+      "step": 22626
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.660337938959491e-06,
+      "loss": 0.5874,
+      "step": 22627
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6596906793702205e-06,
+      "loss": 0.5811,
+      "step": 22628
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.659043534550936e-06,
+      "loss": 0.5794,
+      "step": 22629
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6583965045105433e-06,
+      "loss": 0.55,
+      "step": 22630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6577495892579398e-06,
+      "loss": 0.5845,
+      "step": 22631
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6571027888020364e-06,
+      "loss": 0.4943,
+      "step": 22632
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6564561031517278e-06,
+      "loss": 0.5715,
+      "step": 22633
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6558095323159161e-06,
+      "loss": 0.5144,
+      "step": 22634
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6551630763034931e-06,
+      "loss": 0.5456,
+      "step": 22635
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6545167351233615e-06,
+      "loss": 0.5485,
+      "step": 22636
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6538705087844132e-06,
+      "loss": 0.6118,
+      "step": 22637
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6532243972955397e-06,
+      "loss": 0.5733,
+      "step": 22638
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.652578400665631e-06,
+      "loss": 0.5748,
+      "step": 22639
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6519325189035796e-06,
+      "loss": 0.5791,
+      "step": 22640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.651286752018273e-06,
+      "loss": 0.5426,
+      "step": 22641
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.650641100018593e-06,
+      "loss": 0.56,
+      "step": 22642
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.64999556291343e-06,
+      "loss": 0.5647,
+      "step": 22643
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6493501407116663e-06,
+      "loss": 0.5896,
+      "step": 22644
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6487048334221812e-06,
+      "loss": 0.539,
+      "step": 22645
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6480596410538541e-06,
+      "loss": 0.6033,
+      "step": 22646
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6474145636155669e-06,
+      "loss": 0.5584,
+      "step": 22647
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.646769601116195e-06,
+      "loss": 0.6047,
+      "step": 22648
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.646124753564613e-06,
+      "loss": 0.5738,
+      "step": 22649
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.645480020969692e-06,
+      "loss": 0.4731,
+      "step": 22650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6448354033403091e-06,
+      "loss": 0.6556,
+      "step": 22651
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.644190900685333e-06,
+      "loss": 0.6427,
+      "step": 22652
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6435465130136297e-06,
+      "loss": 0.5775,
+      "step": 22653
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6429022403340711e-06,
+      "loss": 0.5304,
+      "step": 22654
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6422580826555234e-06,
+      "loss": 0.6698,
+      "step": 22655
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.641614039986844e-06,
+      "loss": 0.5929,
+      "step": 22656
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6409701123369015e-06,
+      "loss": 0.5287,
+      "step": 22657
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6403262997145553e-06,
+      "loss": 0.6391,
+      "step": 22658
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6396826021286649e-06,
+      "loss": 0.6316,
+      "step": 22659
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6390390195880845e-06,
+      "loss": 0.5416,
+      "step": 22660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.638395552101677e-06,
+      "loss": 0.5598,
+      "step": 22661
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6377521996782942e-06,
+      "loss": 0.5782,
+      "step": 22662
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6371089623267856e-06,
+      "loss": 0.548,
+      "step": 22663
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6364658400560085e-06,
+      "loss": 0.6023,
+      "step": 22664
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6358228328748115e-06,
+      "loss": 0.5685,
+      "step": 22665
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6351799407920399e-06,
+      "loss": 0.5869,
+      "step": 22666
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6345371638165408e-06,
+      "loss": 0.6128,
+      "step": 22667
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.633894501957164e-06,
+      "loss": 0.6235,
+      "step": 22668
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.633251955222749e-06,
+      "loss": 0.5547,
+      "step": 22669
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6326095236221385e-06,
+      "loss": 0.5414,
+      "step": 22670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6319672071641712e-06,
+      "loss": 0.6388,
+      "step": 22671
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6313250058576913e-06,
+      "loss": 0.5675,
+      "step": 22672
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6306829197115315e-06,
+      "loss": 0.5806,
+      "step": 22673
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.630040948734526e-06,
+      "loss": 0.6164,
+      "step": 22674
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6293990929355152e-06,
+      "loss": 0.5854,
+      "step": 22675
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.628757352323327e-06,
+      "loss": 0.6011,
+      "step": 22676
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6281157269067937e-06,
+      "loss": 0.5573,
+      "step": 22677
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.627474216694741e-06,
+      "loss": 0.5912,
+      "step": 22678
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6268328216960027e-06,
+      "loss": 0.534,
+      "step": 22679
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6261915419194019e-06,
+      "loss": 0.5382,
+      "step": 22680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6255503773737635e-06,
+      "loss": 0.5916,
+      "step": 22681
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6249093280679084e-06,
+      "loss": 0.6096,
+      "step": 22682
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6242683940106618e-06,
+      "loss": 0.628,
+      "step": 22683
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6236275752108421e-06,
+      "loss": 0.5205,
+      "step": 22684
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6229868716772646e-06,
+      "loss": 0.4815,
+      "step": 22685
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6223462834187508e-06,
+      "loss": 0.5808,
+      "step": 22686
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6217058104441131e-06,
+      "loss": 0.572,
+      "step": 22687
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.621065452762165e-06,
+      "loss": 0.5877,
+      "step": 22688
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6204252103817175e-06,
+      "loss": 0.5535,
+      "step": 22689
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6197850833115835e-06,
+      "loss": 0.6017,
+      "step": 22690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6191450715605706e-06,
+      "loss": 0.5419,
+      "step": 22691
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6185051751374847e-06,
+      "loss": 0.5391,
+      "step": 22692
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6178653940511301e-06,
+      "loss": 0.5745,
+      "step": 22693
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6172257283103166e-06,
+      "loss": 0.6186,
+      "step": 22694
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.616586177923841e-06,
+      "loss": 0.5771,
+      "step": 22695
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6159467429005049e-06,
+      "loss": 0.5908,
+      "step": 22696
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6153074232491106e-06,
+      "loss": 0.5801,
+      "step": 22697
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6146682189784523e-06,
+      "loss": 0.6385,
+      "step": 22698
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6140291300973287e-06,
+      "loss": 0.6027,
+      "step": 22699
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6133901566145304e-06,
+      "loss": 0.5772,
+      "step": 22700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.612751298538855e-06,
+      "loss": 0.6117,
+      "step": 22701
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6121125558790907e-06,
+      "loss": 0.5837,
+      "step": 22702
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6114739286440284e-06,
+      "loss": 0.549,
+      "step": 22703
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6108354168424533e-06,
+      "loss": 0.6476,
+      "step": 22704
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6101970204831563e-06,
+      "loss": 0.5757,
+      "step": 22705
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6095587395749202e-06,
+      "loss": 0.54,
+      "step": 22706
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6089205741265269e-06,
+      "loss": 0.6049,
+      "step": 22707
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6082825241467614e-06,
+      "loss": 0.5227,
+      "step": 22708
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6076445896444014e-06,
+      "loss": 0.589,
+      "step": 22709
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6070067706282266e-06,
+      "loss": 0.5677,
+      "step": 22710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6063690671070109e-06,
+      "loss": 0.5585,
+      "step": 22711
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.605731479089534e-06,
+      "loss": 0.5441,
+      "step": 22712
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6050940065845677e-06,
+      "loss": 0.6139,
+      "step": 22713
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6044566496008829e-06,
+      "loss": 0.5607,
+      "step": 22714
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6038194081472536e-06,
+      "loss": 0.557,
+      "step": 22715
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.603182282232446e-06,
+      "loss": 0.5902,
+      "step": 22716
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.602545271865229e-06,
+      "loss": 0.6268,
+      "step": 22717
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6019083770543654e-06,
+      "loss": 0.4913,
+      "step": 22718
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6012715978086247e-06,
+      "loss": 0.6055,
+      "step": 22719
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6006349341367656e-06,
+      "loss": 0.5609,
+      "step": 22720
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5999983860475499e-06,
+      "loss": 0.6293,
+      "step": 22721
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.599361953549735e-06,
+      "loss": 0.6088,
+      "step": 22722
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5987256366520842e-06,
+      "loss": 0.6112,
+      "step": 22723
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5980894353633503e-06,
+      "loss": 0.5679,
+      "step": 22724
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5974533496922862e-06,
+      "loss": 0.6644,
+      "step": 22725
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5968173796476494e-06,
+      "loss": 0.5375,
+      "step": 22726
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5961815252381885e-06,
+      "loss": 0.548,
+      "step": 22727
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5955457864726553e-06,
+      "loss": 0.6229,
+      "step": 22728
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.594910163359793e-06,
+      "loss": 0.5559,
+      "step": 22729
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5942746559083556e-06,
+      "loss": 0.5821,
+      "step": 22730
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5936392641270848e-06,
+      "loss": 0.5549,
+      "step": 22731
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.593003988024724e-06,
+      "loss": 0.6755,
+      "step": 22732
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5923688276100125e-06,
+      "loss": 0.5446,
+      "step": 22733
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5917337828916969e-06,
+      "loss": 0.5862,
+      "step": 22734
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5910988538785122e-06,
+      "loss": 0.5233,
+      "step": 22735
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5904640405791927e-06,
+      "loss": 0.5621,
+      "step": 22736
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5898293430024804e-06,
+      "loss": 0.5715,
+      "step": 22737
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5891947611571067e-06,
+      "loss": 0.5116,
+      "step": 22738
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.588560295051803e-06,
+      "loss": 0.5597,
+      "step": 22739
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5879259446952978e-06,
+      "loss": 0.55,
+      "step": 22740
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5872917100963259e-06,
+      "loss": 0.605,
+      "step": 22741
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5866575912636117e-06,
+      "loss": 0.5444,
+      "step": 22742
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5860235882058827e-06,
+      "loss": 0.5967,
+      "step": 22743
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5853897009318586e-06,
+      "loss": 0.5622,
+      "step": 22744
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.584755929450269e-06,
+      "loss": 0.6148,
+      "step": 22745
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5841222737698325e-06,
+      "loss": 0.5851,
+      "step": 22746
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5834887338992655e-06,
+      "loss": 0.5695,
+      "step": 22747
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.582855309847291e-06,
+      "loss": 0.5744,
+      "step": 22748
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5822220016226242e-06,
+      "loss": 0.5928,
+      "step": 22749
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5815888092339782e-06,
+      "loss": 0.6791,
+      "step": 22750
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.580955732690065e-06,
+      "loss": 0.5256,
+      "step": 22751
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5803227719996016e-06,
+      "loss": 0.5419,
+      "step": 22752
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.579689927171294e-06,
+      "loss": 0.5774,
+      "step": 22753
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5790571982138525e-06,
+      "loss": 0.5539,
+      "step": 22754
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5784245851359813e-06,
+      "loss": 0.5434,
+      "step": 22755
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5777920879463892e-06,
+      "loss": 0.55,
+      "step": 22756
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.577159706653779e-06,
+      "loss": 0.5592,
+      "step": 22757
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5765274412668497e-06,
+      "loss": 0.5474,
+      "step": 22758
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.575895291794306e-06,
+      "loss": 0.5956,
+      "step": 22759
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5752632582448458e-06,
+      "loss": 0.6005,
+      "step": 22760
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5746313406271651e-06,
+      "loss": 0.614,
+      "step": 22761
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5739995389499586e-06,
+      "loss": 0.6475,
+      "step": 22762
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5733678532219243e-06,
+      "loss": 0.5846,
+      "step": 22763
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5727362834517534e-06,
+      "loss": 0.5357,
+      "step": 22764
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5721048296481356e-06,
+      "loss": 0.5634,
+      "step": 22765
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5714734918197572e-06,
+      "loss": 0.5877,
+      "step": 22766
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5708422699753135e-06,
+      "loss": 0.5198,
+      "step": 22767
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5702111641234863e-06,
+      "loss": 0.5773,
+      "step": 22768
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5695801742729577e-06,
+      "loss": 0.5605,
+      "step": 22769
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5689493004324163e-06,
+      "loss": 0.5959,
+      "step": 22770
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5683185426105397e-06,
+      "loss": 0.5555,
+      "step": 22771
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5676879008160095e-06,
+      "loss": 0.5646,
+      "step": 22772
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5670573750575013e-06,
+      "loss": 0.5043,
+      "step": 22773
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5664269653436947e-06,
+      "loss": 0.5389,
+      "step": 22774
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.565796671683264e-06,
+      "loss": 0.5158,
+      "step": 22775
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5651664940848798e-06,
+      "loss": 0.5858,
+      "step": 22776
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5645364325572188e-06,
+      "loss": 0.5228,
+      "step": 22777
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5639064871089472e-06,
+      "loss": 0.6238,
+      "step": 22778
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5632766577487357e-06,
+      "loss": 0.5749,
+      "step": 22779
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5626469444852488e-06,
+      "loss": 0.5526,
+      "step": 22780
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.562017347327155e-06,
+      "loss": 0.5272,
+      "step": 22781
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5613878662831172e-06,
+      "loss": 0.537,
+      "step": 22782
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5607585013617977e-06,
+      "loss": 0.5773,
+      "step": 22783
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5601292525718537e-06,
+      "loss": 0.5749,
+      "step": 22784
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5595001199219496e-06,
+      "loss": 0.5791,
+      "step": 22785
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5588711034207404e-06,
+      "loss": 0.5644,
+      "step": 22786
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5582422030768785e-06,
+      "loss": 0.6175,
+      "step": 22787
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5576134188990254e-06,
+      "loss": 0.6225,
+      "step": 22788
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.556984750895829e-06,
+      "loss": 0.5838,
+      "step": 22789
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5563561990759412e-06,
+      "loss": 0.5695,
+      "step": 22790
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5557277634480084e-06,
+      "loss": 0.6006,
+      "step": 22791
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5550994440206846e-06,
+      "loss": 0.6192,
+      "step": 22792
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.554471240802612e-06,
+      "loss": 0.6913,
+      "step": 22793
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5538431538024369e-06,
+      "loss": 0.5863,
+      "step": 22794
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5532151830287978e-06,
+      "loss": 0.5433,
+      "step": 22795
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5525873284903437e-06,
+      "loss": 0.5434,
+      "step": 22796
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5519595901957096e-06,
+      "loss": 0.5833,
+      "step": 22797
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5513319681535321e-06,
+      "loss": 0.5721,
+      "step": 22798
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5507044623724532e-06,
+      "loss": 0.5935,
+      "step": 22799
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5500770728611047e-06,
+      "loss": 0.546,
+      "step": 22800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.549449799628121e-06,
+      "loss": 0.5733,
+      "step": 22801
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5488226426821307e-06,
+      "loss": 0.6056,
+      "step": 22802
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.548195602031769e-06,
+      "loss": 0.5638,
+      "step": 22803
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5475686776856624e-06,
+      "loss": 0.539,
+      "step": 22804
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5469418696524374e-06,
+      "loss": 0.5956,
+      "step": 22805
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5463151779407182e-06,
+      "loss": 0.5773,
+      "step": 22806
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5456886025591311e-06,
+      "loss": 0.5169,
+      "step": 22807
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5450621435162994e-06,
+      "loss": 0.5589,
+      "step": 22808
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5444358008208383e-06,
+      "loss": 0.5442,
+      "step": 22809
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.543809574481372e-06,
+      "loss": 0.5826,
+      "step": 22810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5431834645065169e-06,
+      "loss": 0.5965,
+      "step": 22811
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5425574709048884e-06,
+      "loss": 0.6117,
+      "step": 22812
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5419315936850976e-06,
+      "loss": 0.5104,
+      "step": 22813
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5413058328557628e-06,
+      "loss": 0.5615,
+      "step": 22814
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5406801884254918e-06,
+      "loss": 0.592,
+      "step": 22815
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5400546604028954e-06,
+      "loss": 0.6008,
+      "step": 22816
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5394292487965767e-06,
+      "loss": 0.5877,
+      "step": 22817
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5388039536151488e-06,
+      "loss": 0.573,
+      "step": 22818
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5381787748672127e-06,
+      "loss": 0.5726,
+      "step": 22819
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5375537125613704e-06,
+      "loss": 0.6561,
+      "step": 22820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.536928766706226e-06,
+      "loss": 0.5726,
+      "step": 22821
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5363039373103783e-06,
+      "loss": 0.6571,
+      "step": 22822
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5356792243824258e-06,
+      "loss": 0.5908,
+      "step": 22823
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.535054627930962e-06,
+      "loss": 0.5515,
+      "step": 22824
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5344301479645862e-06,
+      "loss": 0.616,
+      "step": 22825
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5338057844918907e-06,
+      "loss": 0.5774,
+      "step": 22826
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5331815375214676e-06,
+      "loss": 0.5408,
+      "step": 22827
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.532557407061902e-06,
+      "loss": 0.5783,
+      "step": 22828
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5319333931217906e-06,
+      "loss": 0.5636,
+      "step": 22829
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5313094957097152e-06,
+      "loss": 0.5746,
+      "step": 22830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5306857148342602e-06,
+      "loss": 0.6116,
+      "step": 22831
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5300620505040142e-06,
+      "loss": 0.5624,
+      "step": 22832
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.529438502727557e-06,
+      "loss": 0.5695,
+      "step": 22833
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5288150715134686e-06,
+      "loss": 0.5854,
+      "step": 22834
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5281917568703253e-06,
+      "loss": 0.5816,
+      "step": 22835
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5275685588067102e-06,
+      "loss": 0.5516,
+      "step": 22836
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5269454773311965e-06,
+      "loss": 0.5384,
+      "step": 22837
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5263225124523563e-06,
+      "loss": 0.584,
+      "step": 22838
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5256996641787647e-06,
+      "loss": 0.6004,
+      "step": 22839
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.525076932518993e-06,
+      "loss": 0.6301,
+      "step": 22840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5244543174816096e-06,
+      "loss": 0.5743,
+      "step": 22841
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5238318190751789e-06,
+      "loss": 0.5838,
+      "step": 22842
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5232094373082729e-06,
+      "loss": 0.5824,
+      "step": 22843
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5225871721894536e-06,
+      "loss": 0.5655,
+      "step": 22844
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.521965023727283e-06,
+      "loss": 0.6345,
+      "step": 22845
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5213429919303225e-06,
+      "loss": 0.5396,
+      "step": 22846
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5207210768071334e-06,
+      "loss": 0.6106,
+      "step": 22847
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5200992783662727e-06,
+      "loss": 0.567,
+      "step": 22848
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5194775966162955e-06,
+      "loss": 0.6146,
+      "step": 22849
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5188560315657596e-06,
+      "loss": 0.658,
+      "step": 22850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.518234583223218e-06,
+      "loss": 0.5476,
+      "step": 22851
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5176132515972209e-06,
+      "loss": 0.5838,
+      "step": 22852
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5169920366963152e-06,
+      "loss": 0.5765,
+      "step": 22853
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.516370938529057e-06,
+      "loss": 0.5418,
+      "step": 22854
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.515749957103988e-06,
+      "loss": 0.6525,
+      "step": 22855
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5151290924296547e-06,
+      "loss": 0.6196,
+      "step": 22856
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.514508344514598e-06,
+      "loss": 0.5251,
+      "step": 22857
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5138877133673656e-06,
+      "loss": 0.5623,
+      "step": 22858
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5132671989964942e-06,
+      "loss": 0.6155,
+      "step": 22859
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.512646801410521e-06,
+      "loss": 0.5653,
+      "step": 22860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5120265206179873e-06,
+      "loss": 0.5953,
+      "step": 22861
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5114063566274274e-06,
+      "loss": 0.5606,
+      "step": 22862
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.510786309447375e-06,
+      "loss": 0.5767,
+      "step": 22863
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5101663790863597e-06,
+      "loss": 0.5962,
+      "step": 22864
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5095465655529174e-06,
+      "loss": 0.5257,
+      "step": 22865
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5089268688555769e-06,
+      "loss": 0.5693,
+      "step": 22866
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5083072890028582e-06,
+      "loss": 0.5195,
+      "step": 22867
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5076878260032957e-06,
+      "loss": 0.5882,
+      "step": 22868
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.507068479865411e-06,
+      "loss": 0.5428,
+      "step": 22869
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5064492505977234e-06,
+      "loss": 0.6177,
+      "step": 22870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5058301382087603e-06,
+      "loss": 0.5883,
+      "step": 22871
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.505211142707038e-06,
+      "loss": 0.6371,
+      "step": 22872
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5045922641010746e-06,
+      "loss": 0.5823,
+      "step": 22873
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.503973502399384e-06,
+      "loss": 0.581,
+      "step": 22874
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5033548576104851e-06,
+      "loss": 0.5625,
+      "step": 22875
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5027363297428898e-06,
+      "loss": 0.5754,
+      "step": 22876
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5021179188051082e-06,
+      "loss": 0.5975,
+      "step": 22877
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.501499624805649e-06,
+      "loss": 0.6392,
+      "step": 22878
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5008814477530242e-06,
+      "loss": 0.6031,
+      "step": 22879
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5002633876557394e-06,
+      "loss": 0.6054,
+      "step": 22880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4996454445222963e-06,
+      "loss": 0.6119,
+      "step": 22881
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.499027618361204e-06,
+      "loss": 0.5972,
+      "step": 22882
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.49840990918096e-06,
+      "loss": 0.6119,
+      "step": 22883
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4977923169900665e-06,
+      "loss": 0.5311,
+      "step": 22884
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.497174841797019e-06,
+      "loss": 0.5414,
+      "step": 22885
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4965574836103192e-06,
+      "loss": 0.5515,
+      "step": 22886
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4959402424384607e-06,
+      "loss": 0.5591,
+      "step": 22887
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4953231182899363e-06,
+      "loss": 0.5781,
+      "step": 22888
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4947061111732363e-06,
+      "loss": 0.5201,
+      "step": 22889
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.494089221096856e-06,
+      "loss": 0.5309,
+      "step": 22890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.493472448069282e-06,
+      "loss": 0.5834,
+      "step": 22891
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4928557920989994e-06,
+      "loss": 0.5583,
+      "step": 22892
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4922392531944985e-06,
+      "loss": 0.5498,
+      "step": 22893
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4916228313642612e-06,
+      "loss": 0.6138,
+      "step": 22894
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4910065266167695e-06,
+      "loss": 0.504,
+      "step": 22895
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4903903389605024e-06,
+      "loss": 0.5357,
+      "step": 22896
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.489774268403943e-06,
+      "loss": 0.5647,
+      "step": 22897
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4891583149555688e-06,
+      "loss": 0.5687,
+      "step": 22898
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4885424786238533e-06,
+      "loss": 0.5849,
+      "step": 22899
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4879267594172708e-06,
+      "loss": 0.5671,
+      "step": 22900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4873111573442967e-06,
+      "loss": 0.5409,
+      "step": 22901
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.486695672413402e-06,
+      "loss": 0.6285,
+      "step": 22902
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4860803046330519e-06,
+      "loss": 0.6168,
+      "step": 22903
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4854650540117211e-06,
+      "loss": 0.6389,
+      "step": 22904
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4848499205578726e-06,
+      "loss": 0.6359,
+      "step": 22905
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4842349042799708e-06,
+      "loss": 0.5475,
+      "step": 22906
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.483620005186477e-06,
+      "loss": 0.5859,
+      "step": 22907
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4830052232858582e-06,
+      "loss": 0.5386,
+      "step": 22908
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4823905585865706e-06,
+      "loss": 0.5675,
+      "step": 22909
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4817760110970726e-06,
+      "loss": 0.561,
+      "step": 22910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4811615808258195e-06,
+      "loss": 0.6101,
+      "step": 22911
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4805472677812715e-06,
+      "loss": 0.6068,
+      "step": 22912
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4799330719718774e-06,
+      "loss": 0.5652,
+      "step": 22913
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4793189934060891e-06,
+      "loss": 0.5578,
+      "step": 22914
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4787050320923613e-06,
+      "loss": 0.5316,
+      "step": 22915
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4780911880391392e-06,
+      "loss": 0.5355,
+      "step": 22916
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4774774612548704e-06,
+      "loss": 0.5942,
+      "step": 22917
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4768638517479983e-06,
+      "loss": 0.5644,
+      "step": 22918
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4762503595269705e-06,
+      "loss": 0.5294,
+      "step": 22919
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4756369846002273e-06,
+      "loss": 0.5885,
+      "step": 22920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.475023726976209e-06,
+      "loss": 0.5692,
+      "step": 22921
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.474410586663354e-06,
+      "loss": 0.5979,
+      "step": 22922
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4737975636701019e-06,
+      "loss": 0.574,
+      "step": 22923
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.473184658004887e-06,
+      "loss": 0.5513,
+      "step": 22924
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4725718696761415e-06,
+      "loss": 0.6182,
+      "step": 22925
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4719591986923021e-06,
+      "loss": 0.5736,
+      "step": 22926
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4713466450617975e-06,
+      "loss": 0.5764,
+      "step": 22927
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4707342087930576e-06,
+      "loss": 0.5271,
+      "step": 22928
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4701218898945058e-06,
+      "loss": 0.5745,
+      "step": 22929
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4695096883745752e-06,
+      "loss": 0.556,
+      "step": 22930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.468897604241687e-06,
+      "loss": 0.5028,
+      "step": 22931
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4682856375042608e-06,
+      "loss": 0.5506,
+      "step": 22932
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4676737881707237e-06,
+      "loss": 0.5712,
+      "step": 22933
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.467062056249492e-06,
+      "loss": 0.5681,
+      "step": 22934
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4664504417489844e-06,
+      "loss": 0.574,
+      "step": 22935
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4658389446776145e-06,
+      "loss": 0.5588,
+      "step": 22936
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4652275650438019e-06,
+      "loss": 0.5949,
+      "step": 22937
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4646163028559568e-06,
+      "loss": 0.5759,
+      "step": 22938
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4640051581224923e-06,
+      "loss": 0.5051,
+      "step": 22939
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.463394130851814e-06,
+      "loss": 0.6668,
+      "step": 22940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4627832210523363e-06,
+      "loss": 0.5675,
+      "step": 22941
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4621724287324623e-06,
+      "loss": 0.5636,
+      "step": 22942
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4615617539005955e-06,
+      "loss": 0.6615,
+      "step": 22943
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4609511965651447e-06,
+      "loss": 0.5616,
+      "step": 22944
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4603407567345074e-06,
+      "loss": 0.6004,
+      "step": 22945
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4597304344170848e-06,
+      "loss": 0.5643,
+      "step": 22946
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4591202296212736e-06,
+      "loss": 0.554,
+      "step": 22947
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4585101423554758e-06,
+      "loss": 0.5795,
+      "step": 22948
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4579001726280828e-06,
+      "loss": 0.5452,
+      "step": 22949
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4572903204474897e-06,
+      "loss": 0.6069,
+      "step": 22950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4566805858220857e-06,
+      "loss": 0.5885,
+      "step": 22951
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4560709687602649e-06,
+      "loss": 0.6328,
+      "step": 22952
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.455461469270416e-06,
+      "loss": 0.5903,
+      "step": 22953
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4548520873609228e-06,
+      "loss": 0.541,
+      "step": 22954
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.454242823040175e-06,
+      "loss": 0.5226,
+      "step": 22955
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.453633676316556e-06,
+      "loss": 0.5191,
+      "step": 22956
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4530246471984455e-06,
+      "loss": 0.5335,
+      "step": 22957
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4524157356942249e-06,
+      "loss": 0.515,
+      "step": 22958
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.451806941812276e-06,
+      "loss": 0.5876,
+      "step": 22959
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4511982655609746e-06,
+      "loss": 0.5994,
+      "step": 22960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4505897069486973e-06,
+      "loss": 0.4635,
+      "step": 22961
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4499812659838152e-06,
+      "loss": 0.6017,
+      "step": 22962
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4493729426747062e-06,
+      "loss": 0.6203,
+      "step": 22963
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4487647370297386e-06,
+      "loss": 0.6178,
+      "step": 22964
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4481566490572797e-06,
+      "loss": 0.5668,
+      "step": 22965
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4475486787657011e-06,
+      "loss": 0.5655,
+      "step": 22966
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4469408261633689e-06,
+      "loss": 0.4848,
+      "step": 22967
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4463330912586448e-06,
+      "loss": 0.5536,
+      "step": 22968
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4457254740598925e-06,
+      "loss": 0.5912,
+      "step": 22969
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4451179745754752e-06,
+      "loss": 0.6042,
+      "step": 22970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4445105928137526e-06,
+      "loss": 0.5139,
+      "step": 22971
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4439033287830805e-06,
+      "loss": 0.5519,
+      "step": 22972
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4432961824918156e-06,
+      "loss": 0.5356,
+      "step": 22973
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4426891539483158e-06,
+      "loss": 0.5878,
+      "step": 22974
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4420822431609316e-06,
+      "loss": 0.5754,
+      "step": 22975
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4414754501380134e-06,
+      "loss": 0.5532,
+      "step": 22976
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4408687748879157e-06,
+      "loss": 0.5745,
+      "step": 22977
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4402622174189851e-06,
+      "loss": 0.5772,
+      "step": 22978
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4396557777395658e-06,
+      "loss": 0.5126,
+      "step": 22979
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4390494558580038e-06,
+      "loss": 0.5089,
+      "step": 22980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4384432517826463e-06,
+      "loss": 0.5953,
+      "step": 22981
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4378371655218314e-06,
+      "loss": 0.5803,
+      "step": 22982
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4372311970839015e-06,
+      "loss": 0.5657,
+      "step": 22983
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4366253464771906e-06,
+      "loss": 0.6123,
+      "step": 22984
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4360196137100423e-06,
+      "loss": 0.5349,
+      "step": 22985
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.43541399879079e-06,
+      "loss": 0.5478,
+      "step": 22986
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4348085017277635e-06,
+      "loss": 0.5838,
+      "step": 22987
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4342031225292995e-06,
+      "loss": 0.6145,
+      "step": 22988
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4335978612037283e-06,
+      "loss": 0.5793,
+      "step": 22989
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4329927177593771e-06,
+      "loss": 0.5517,
+      "step": 22990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.432387692204572e-06,
+      "loss": 0.5513,
+      "step": 22991
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4317827845476418e-06,
+      "loss": 0.5858,
+      "step": 22992
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4311779947969095e-06,
+      "loss": 0.6397,
+      "step": 22993
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4305733229606956e-06,
+      "loss": 0.6158,
+      "step": 22994
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4299687690473241e-06,
+      "loss": 0.5848,
+      "step": 22995
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.429364333065113e-06,
+      "loss": 0.5829,
+      "step": 22996
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4287600150223791e-06,
+      "loss": 0.5578,
+      "step": 22997
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4281558149274377e-06,
+      "loss": 0.6077,
+      "step": 22998
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4275517327886045e-06,
+      "loss": 0.6199,
+      "step": 22999
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.426947768614193e-06,
+      "loss": 0.58,
+      "step": 23000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.426343922412513e-06,
+      "loss": 0.6191,
+      "step": 23001
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4257401941918725e-06,
+      "loss": 0.5975,
+      "step": 23002
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4251365839605823e-06,
+      "loss": 0.5783,
+      "step": 23003
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4245330917269473e-06,
+      "loss": 0.62,
+      "step": 23004
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4239297174992706e-06,
+      "loss": 0.5641,
+      "step": 23005
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4233264612858577e-06,
+      "loss": 0.54,
+      "step": 23006
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4227233230950089e-06,
+      "loss": 0.6537,
+      "step": 23007
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.422120302935025e-06,
+      "loss": 0.6198,
+      "step": 23008
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4215174008141997e-06,
+      "loss": 0.5847,
+      "step": 23009
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.420914616740835e-06,
+      "loss": 0.633,
+      "step": 23010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4203119507232233e-06,
+      "loss": 0.5078,
+      "step": 23011
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.419709402769658e-06,
+      "loss": 0.5473,
+      "step": 23012
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4191069728884287e-06,
+      "loss": 0.5734,
+      "step": 23013
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4185046610878294e-06,
+      "loss": 0.5287,
+      "step": 23014
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4179024673761465e-06,
+      "loss": 0.6056,
+      "step": 23015
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4173003917616635e-06,
+      "loss": 0.528,
+      "step": 23016
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4166984342526724e-06,
+      "loss": 0.5081,
+      "step": 23017
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4160965948574513e-06,
+      "loss": 0.5687,
+      "step": 23018
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4154948735842844e-06,
+      "loss": 0.5766,
+      "step": 23019
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4148932704414475e-06,
+      "loss": 0.5406,
+      "step": 23020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4142917854372262e-06,
+      "loss": 0.5636,
+      "step": 23021
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4136904185798939e-06,
+      "loss": 0.6625,
+      "step": 23022
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.413089169877726e-06,
+      "loss": 0.5844,
+      "step": 23023
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.412488039338994e-06,
+      "loss": 0.5083,
+      "step": 23024
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4118870269719743e-06,
+      "loss": 0.5727,
+      "step": 23025
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4112861327849347e-06,
+      "loss": 0.686,
+      "step": 23026
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4106853567861435e-06,
+      "loss": 0.5395,
+      "step": 23027
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.41008469898387e-06,
+      "loss": 0.6234,
+      "step": 23028
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4094841593863794e-06,
+      "loss": 0.5734,
+      "step": 23029
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.408883738001936e-06,
+      "loss": 0.5727,
+      "step": 23030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4082834348387974e-06,
+      "loss": 0.5518,
+      "step": 23031
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4076832499052317e-06,
+      "loss": 0.566,
+      "step": 23032
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4070831832094934e-06,
+      "loss": 0.5588,
+      "step": 23033
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4064832347598412e-06,
+      "loss": 0.6357,
+      "step": 23034
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4058834045645287e-06,
+      "loss": 0.6007,
+      "step": 23035
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4052836926318136e-06,
+      "loss": 0.5345,
+      "step": 23036
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4046840989699474e-06,
+      "loss": 0.5815,
+      "step": 23037
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4040846235871775e-06,
+      "loss": 0.6393,
+      "step": 23038
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4034852664917586e-06,
+      "loss": 0.5224,
+      "step": 23039
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4028860276919365e-06,
+      "loss": 0.5924,
+      "step": 23040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4022869071959566e-06,
+      "loss": 0.6328,
+      "step": 23041
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4016879050120612e-06,
+      "loss": 0.554,
+      "step": 23042
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4010890211484983e-06,
+      "loss": 0.5519,
+      "step": 23043
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4004902556135048e-06,
+      "loss": 0.5786,
+      "step": 23044
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3998916084153225e-06,
+      "loss": 0.5848,
+      "step": 23045
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3992930795621862e-06,
+      "loss": 0.5693,
+      "step": 23046
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.398694669062336e-06,
+      "loss": 0.5382,
+      "step": 23047
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3980963769240051e-06,
+      "loss": 0.5022,
+      "step": 23048
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.397498203155424e-06,
+      "loss": 0.5834,
+      "step": 23049
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3969001477648292e-06,
+      "loss": 0.532,
+      "step": 23050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3963022107604474e-06,
+      "loss": 0.6189,
+      "step": 23051
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3957043921505075e-06,
+      "loss": 0.5184,
+      "step": 23052
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3951066919432332e-06,
+      "loss": 0.59,
+      "step": 23053
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3945091101468532e-06,
+      "loss": 0.6112,
+      "step": 23054
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3939116467695902e-06,
+      "loss": 0.5906,
+      "step": 23055
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3933143018196615e-06,
+      "loss": 0.5757,
+      "step": 23056
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3927170753052944e-06,
+      "loss": 0.6451,
+      "step": 23057
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.392119967234702e-06,
+      "loss": 0.5391,
+      "step": 23058
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3915229776161031e-06,
+      "loss": 0.6103,
+      "step": 23059
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3909261064577094e-06,
+      "loss": 0.5705,
+      "step": 23060
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3903293537677398e-06,
+      "loss": 0.6589,
+      "step": 23061
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.389732719554403e-06,
+      "loss": 0.6328,
+      "step": 23062
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3891362038259093e-06,
+      "loss": 0.5774,
+      "step": 23063
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3885398065904666e-06,
+      "loss": 0.5448,
+      "step": 23064
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3879435278562835e-06,
+      "loss": 0.585,
+      "step": 23065
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3873473676315662e-06,
+      "loss": 0.5641,
+      "step": 23066
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3867513259245135e-06,
+      "loss": 0.583,
+      "step": 23067
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3861554027433332e-06,
+      "loss": 0.5789,
+      "step": 23068
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3855595980962243e-06,
+      "loss": 0.5547,
+      "step": 23069
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3849639119913838e-06,
+      "loss": 0.5721,
+      "step": 23070
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3843683444370071e-06,
+      "loss": 0.5269,
+      "step": 23071
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3837728954412955e-06,
+      "loss": 0.5617,
+      "step": 23072
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.383177565012439e-06,
+      "loss": 0.539,
+      "step": 23073
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3825823531586312e-06,
+      "loss": 0.591,
+      "step": 23074
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3819872598880602e-06,
+      "loss": 0.5657,
+      "step": 23075
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3813922852089178e-06,
+      "loss": 0.5974,
+      "step": 23076
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3807974291293947e-06,
+      "loss": 0.6106,
+      "step": 23077
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.380202691657666e-06,
+      "loss": 0.5898,
+      "step": 23078
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3796080728019256e-06,
+      "loss": 0.5648,
+      "step": 23079
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3790135725703523e-06,
+      "loss": 0.5382,
+      "step": 23080
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3784191909711253e-06,
+      "loss": 0.521,
+      "step": 23081
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3778249280124278e-06,
+      "loss": 0.5459,
+      "step": 23082
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3772307837024356e-06,
+      "loss": 0.5346,
+      "step": 23083
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3766367580493244e-06,
+      "loss": 0.6021,
+      "step": 23084
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3760428510612667e-06,
+      "loss": 0.5171,
+      "step": 23085
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.375449062746439e-06,
+      "loss": 0.5992,
+      "step": 23086
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3748553931130103e-06,
+      "loss": 0.5885,
+      "step": 23087
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3742618421691501e-06,
+      "loss": 0.5811,
+      "step": 23088
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3736684099230247e-06,
+      "loss": 0.595,
+      "step": 23089
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3730750963828033e-06,
+      "loss": 0.5206,
+      "step": 23090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3724819015566481e-06,
+      "loss": 0.5747,
+      "step": 23091
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.371888825452722e-06,
+      "loss": 0.5341,
+      "step": 23092
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3712958680791878e-06,
+      "loss": 0.5619,
+      "step": 23093
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3707030294442048e-06,
+      "loss": 0.5817,
+      "step": 23094
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3701103095559308e-06,
+      "loss": 0.5994,
+      "step": 23095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3695177084225198e-06,
+      "loss": 0.5688,
+      "step": 23096
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3689252260521302e-06,
+      "loss": 0.5524,
+      "step": 23097
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3683328624529135e-06,
+      "loss": 0.5693,
+      "step": 23098
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3677406176330187e-06,
+      "loss": 0.528,
+      "step": 23099
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3671484916006006e-06,
+      "loss": 0.5357,
+      "step": 23100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3665564843638045e-06,
+      "loss": 0.5867,
+      "step": 23101
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3659645959307777e-06,
+      "loss": 0.5891,
+      "step": 23102
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.365372826309661e-06,
+      "loss": 0.5553,
+      "step": 23103
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3647811755086048e-06,
+      "loss": 0.5855,
+      "step": 23104
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.364189643535747e-06,
+      "loss": 0.6226,
+      "step": 23105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.363598230399228e-06,
+      "loss": 0.5781,
+      "step": 23106
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.363006936107183e-06,
+      "loss": 0.5822,
+      "step": 23107
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3624157606677534e-06,
+      "loss": 0.5893,
+      "step": 23108
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3618247040890743e-06,
+      "loss": 0.594,
+      "step": 23109
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3612337663792741e-06,
+      "loss": 0.6013,
+      "step": 23110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.36064294754649e-06,
+      "loss": 0.6239,
+      "step": 23111
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3600522475988508e-06,
+      "loss": 0.6295,
+      "step": 23112
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3594616665444837e-06,
+      "loss": 0.5734,
+      "step": 23113
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3588712043915153e-06,
+      "loss": 0.5489,
+      "step": 23114
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3582808611480735e-06,
+      "loss": 0.5533,
+      "step": 23115
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3576906368222797e-06,
+      "loss": 0.5541,
+      "step": 23116
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3571005314222574e-06,
+      "loss": 0.5726,
+      "step": 23117
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3565105449561234e-06,
+      "loss": 0.5493,
+      "step": 23118
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3559206774320012e-06,
+      "loss": 0.5457,
+      "step": 23119
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3553309288580064e-06,
+      "loss": 0.6106,
+      "step": 23120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3547412992422515e-06,
+      "loss": 0.5768,
+      "step": 23121
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3541517885928546e-06,
+      "loss": 0.5863,
+      "step": 23122
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3535623969179256e-06,
+      "loss": 0.6397,
+      "step": 23123
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3529731242255762e-06,
+      "loss": 0.5928,
+      "step": 23124
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3523839705239128e-06,
+      "loss": 0.5926,
+      "step": 23125
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3517949358210458e-06,
+      "loss": 0.5495,
+      "step": 23126
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3512060201250798e-06,
+      "loss": 0.559,
+      "step": 23127
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3506172234441184e-06,
+      "loss": 0.5921,
+      "step": 23128
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3500285457862616e-06,
+      "loss": 0.6134,
+      "step": 23129
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3494399871596143e-06,
+      "loss": 0.5872,
+      "step": 23130
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3488515475722752e-06,
+      "loss": 0.5302,
+      "step": 23131
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.348263227032337e-06,
+      "loss": 0.6489,
+      "step": 23132
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3476750255479009e-06,
+      "loss": 0.6344,
+      "step": 23133
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3470869431270606e-06,
+      "loss": 0.5431,
+      "step": 23134
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.346498979777906e-06,
+      "loss": 0.6289,
+      "step": 23135
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3459111355085263e-06,
+      "loss": 0.5459,
+      "step": 23136
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3453234103270162e-06,
+      "loss": 0.5123,
+      "step": 23137
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3447358042414617e-06,
+      "loss": 0.5309,
+      "step": 23138
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3441483172599468e-06,
+      "loss": 0.6124,
+      "step": 23139
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3435609493905544e-06,
+      "loss": 0.6154,
+      "step": 23140
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3429737006413712e-06,
+      "loss": 0.5173,
+      "step": 23141
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3423865710204776e-06,
+      "loss": 0.6093,
+      "step": 23142
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3417995605359503e-06,
+      "loss": 0.5715,
+      "step": 23143
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3412126691958692e-06,
+      "loss": 0.5934,
+      "step": 23144
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3406258970083119e-06,
+      "loss": 0.6244,
+      "step": 23145
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3400392439813504e-06,
+      "loss": 0.5601,
+      "step": 23146
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.339452710123056e-06,
+      "loss": 0.5306,
+      "step": 23147
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3388662954415032e-06,
+      "loss": 0.6018,
+      "step": 23148
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3382799999447616e-06,
+      "loss": 0.5294,
+      "step": 23149
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3376938236408986e-06,
+      "loss": 0.629,
+      "step": 23150
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3371077665379772e-06,
+      "loss": 0.5185,
+      "step": 23151
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3365218286440663e-06,
+      "loss": 0.5983,
+      "step": 23152
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.335936009967228e-06,
+      "loss": 0.5614,
+      "step": 23153
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.335350310515522e-06,
+      "loss": 0.5329,
+      "step": 23154
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.33476473029701e-06,
+      "loss": 0.5259,
+      "step": 23155
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3341792693197498e-06,
+      "loss": 0.579,
+      "step": 23156
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3335939275917964e-06,
+      "loss": 0.572,
+      "step": 23157
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3330087051212048e-06,
+      "loss": 0.5997,
+      "step": 23158
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3324236019160296e-06,
+      "loss": 0.6713,
+      "step": 23159
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3318386179843223e-06,
+      "loss": 0.5219,
+      "step": 23160
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3312537533341307e-06,
+      "loss": 0.5825,
+      "step": 23161
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3306690079735063e-06,
+      "loss": 0.5879,
+      "step": 23162
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3300843819104937e-06,
+      "loss": 0.6502,
+      "step": 23163
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3294998751531374e-06,
+      "loss": 0.5793,
+      "step": 23164
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3289154877094801e-06,
+      "loss": 0.6278,
+      "step": 23165
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3283312195875663e-06,
+      "loss": 0.5392,
+      "step": 23166
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3277470707954343e-06,
+      "loss": 0.5362,
+      "step": 23167
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3271630413411241e-06,
+      "loss": 0.5404,
+      "step": 23168
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.326579131232668e-06,
+      "loss": 0.5688,
+      "step": 23169
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3259953404781055e-06,
+      "loss": 0.5683,
+      "step": 23170
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3254116690854702e-06,
+      "loss": 0.583,
+      "step": 23171
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3248281170627897e-06,
+      "loss": 0.5707,
+      "step": 23172
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.324244684418099e-06,
+      "loss": 0.5428,
+      "step": 23173
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3236613711594248e-06,
+      "loss": 0.5694,
+      "step": 23174
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3230781772947942e-06,
+      "loss": 0.5859,
+      "step": 23175
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3224951028322296e-06,
+      "loss": 0.5719,
+      "step": 23176
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3219121477797592e-06,
+      "loss": 0.5485,
+      "step": 23177
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.321329312145403e-06,
+      "loss": 0.6212,
+      "step": 23178
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3207465959371812e-06,
+      "loss": 0.5318,
+      "step": 23179
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.320163999163111e-06,
+      "loss": 0.5558,
+      "step": 23180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3195815218312125e-06,
+      "loss": 0.5503,
+      "step": 23181
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3189991639494992e-06,
+      "loss": 0.6108,
+      "step": 23182
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3184169255259827e-06,
+      "loss": 0.583,
+      "step": 23183
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3178348065686807e-06,
+      "loss": 0.5449,
+      "step": 23184
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3172528070856006e-06,
+      "loss": 0.5519,
+      "step": 23185
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.316670927084751e-06,
+      "loss": 0.5218,
+      "step": 23186
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.316089166574136e-06,
+      "loss": 0.5641,
+      "step": 23187
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3155075255617677e-06,
+      "loss": 0.5499,
+      "step": 23188
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3149260040556467e-06,
+      "loss": 0.5426,
+      "step": 23189
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.314344602063775e-06,
+      "loss": 0.5419,
+      "step": 23190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3137633195941524e-06,
+      "loss": 0.6114,
+      "step": 23191
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3131821566547798e-06,
+      "loss": 0.6221,
+      "step": 23192
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3126011132536542e-06,
+      "loss": 0.5973,
+      "step": 23193
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3120201893987683e-06,
+      "loss": 0.5232,
+      "step": 23194
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.31143938509812e-06,
+      "loss": 0.5926,
+      "step": 23195
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3108587003597007e-06,
+      "loss": 0.5713,
+      "step": 23196
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3102781351915016e-06,
+      "loss": 0.5638,
+      "step": 23197
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3096976896015079e-06,
+      "loss": 0.5435,
+      "step": 23198
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3091173635977116e-06,
+      "loss": 0.6229,
+      "step": 23199
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.308537157188098e-06,
+      "loss": 0.6067,
+      "step": 23200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3079570703806487e-06,
+      "loss": 0.5992,
+      "step": 23201
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.307377103183346e-06,
+      "loss": 0.5236,
+      "step": 23202
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3067972556041753e-06,
+      "loss": 0.537,
+      "step": 23203
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3062175276511124e-06,
+      "loss": 0.6565,
+      "step": 23204
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3056379193321333e-06,
+      "loss": 0.6047,
+      "step": 23205
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3050584306552172e-06,
+      "loss": 0.567,
+      "step": 23206
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3044790616283376e-06,
+      "loss": 0.5526,
+      "step": 23207
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.303899812259467e-06,
+      "loss": 0.6054,
+      "step": 23208
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3033206825565747e-06,
+      "loss": 0.5272,
+      "step": 23209
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.302741672527632e-06,
+      "loss": 0.547,
+      "step": 23210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.302162782180607e-06,
+      "loss": 0.5456,
+      "step": 23211
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3015840115234645e-06,
+      "loss": 0.607,
+      "step": 23212
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.301005360564166e-06,
+      "loss": 0.5892,
+      "step": 23213
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3004268293106803e-06,
+      "loss": 0.5103,
+      "step": 23214
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2998484177709658e-06,
+      "loss": 0.5936,
+      "step": 23215
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2992701259529793e-06,
+      "loss": 0.6007,
+      "step": 23216
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2986919538646835e-06,
+      "loss": 0.555,
+      "step": 23217
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2981139015140309e-06,
+      "loss": 0.6044,
+      "step": 23218
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2975359689089785e-06,
+      "loss": 0.5341,
+      "step": 23219
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.296958156057475e-06,
+      "loss": 0.5812,
+      "step": 23220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.296380462967477e-06,
+      "loss": 0.5983,
+      "step": 23221
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.295802889646931e-06,
+      "loss": 0.5258,
+      "step": 23222
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.295225436103783e-06,
+      "loss": 0.5769,
+      "step": 23223
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2946481023459844e-06,
+      "loss": 0.5728,
+      "step": 23224
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2940708883814768e-06,
+      "loss": 0.6057,
+      "step": 23225
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2934937942182034e-06,
+      "loss": 0.5011,
+      "step": 23226
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.292916819864104e-06,
+      "loss": 0.5797,
+      "step": 23227
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2923399653271217e-06,
+      "loss": 0.5801,
+      "step": 23228
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.291763230615193e-06,
+      "loss": 0.5747,
+      "step": 23229
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.291186615736254e-06,
+      "loss": 0.5646,
+      "step": 23230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2906101206982368e-06,
+      "loss": 0.6069,
+      "step": 23231
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2900337455090793e-06,
+      "loss": 0.5727,
+      "step": 23232
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.289457490176711e-06,
+      "loss": 0.5746,
+      "step": 23233
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2888813547090585e-06,
+      "loss": 0.5713,
+      "step": 23234
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2883053391140553e-06,
+      "loss": 0.6032,
+      "step": 23235
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2877294433996256e-06,
+      "loss": 0.56,
+      "step": 23236
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.287153667573694e-06,
+      "loss": 0.5853,
+      "step": 23237
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2865780116441817e-06,
+      "loss": 0.652,
+      "step": 23238
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2860024756190149e-06,
+      "loss": 0.5511,
+      "step": 23239
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2854270595061102e-06,
+      "loss": 0.5777,
+      "step": 23240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2848517633133861e-06,
+      "loss": 0.5585,
+      "step": 23241
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2842765870487571e-06,
+      "loss": 0.5626,
+      "step": 23242
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.283701530720144e-06,
+      "loss": 0.5824,
+      "step": 23243
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2831265943354553e-06,
+      "loss": 0.5969,
+      "step": 23244
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.282551777902603e-06,
+      "loss": 0.5436,
+      "step": 23245
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2819770814294996e-06,
+      "loss": 0.5828,
+      "step": 23246
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2814025049240519e-06,
+      "loss": 0.5689,
+      "step": 23247
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2808280483941672e-06,
+      "loss": 0.5381,
+      "step": 23248
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2802537118477476e-06,
+      "loss": 0.5511,
+      "step": 23249
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2796794952927005e-06,
+      "loss": 0.6322,
+      "step": 23250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.279105398736925e-06,
+      "loss": 0.5171,
+      "step": 23251
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2785314221883238e-06,
+      "loss": 0.5393,
+      "step": 23252
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2779575656547906e-06,
+      "loss": 0.5676,
+      "step": 23253
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2773838291442276e-06,
+      "loss": 0.545,
+      "step": 23254
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2768102126645266e-06,
+      "loss": 0.5424,
+      "step": 23255
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.27623671622358e-06,
+      "loss": 0.594,
+      "step": 23256
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.275663339829285e-06,
+      "loss": 0.582,
+      "step": 23257
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2750900834895274e-06,
+      "loss": 0.5915,
+      "step": 23258
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2745169472121966e-06,
+      "loss": 0.6109,
+      "step": 23259
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.273943931005177e-06,
+      "loss": 0.5589,
+      "step": 23260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2733710348763584e-06,
+      "loss": 0.5372,
+      "step": 23261
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2727982588336229e-06,
+      "loss": 0.5565,
+      "step": 23262
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2722256028848513e-06,
+      "loss": 0.535,
+      "step": 23263
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2716530670379224e-06,
+      "loss": 0.5515,
+      "step": 23264
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.271080651300719e-06,
+      "loss": 0.6222,
+      "step": 23265
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.270508355681116e-06,
+      "loss": 0.6037,
+      "step": 23266
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2699361801869858e-06,
+      "loss": 0.5523,
+      "step": 23267
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2693641248262067e-06,
+      "loss": 0.5843,
+      "step": 23268
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.268792189606649e-06,
+      "loss": 0.6006,
+      "step": 23269
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.268220374536182e-06,
+      "loss": 0.5717,
+      "step": 23270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2676486796226738e-06,
+      "loss": 0.6406,
+      "step": 23271
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2670771048739939e-06,
+      "loss": 0.5965,
+      "step": 23272
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2665056502980078e-06,
+      "loss": 0.5095,
+      "step": 23273
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2659343159025762e-06,
+      "loss": 0.574,
+      "step": 23274
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2653631016955615e-06,
+      "loss": 0.5939,
+      "step": 23275
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2647920076848275e-06,
+      "loss": 0.599,
+      "step": 23276
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2642210338782314e-06,
+      "loss": 0.5399,
+      "step": 23277
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2636501802836277e-06,
+      "loss": 0.5429,
+      "step": 23278
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2630794469088759e-06,
+      "loss": 0.5445,
+      "step": 23279
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2625088337618275e-06,
+      "loss": 0.6016,
+      "step": 23280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.261938340850335e-06,
+      "loss": 0.5106,
+      "step": 23281
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2613679681822465e-06,
+      "loss": 0.5427,
+      "step": 23282
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2607977157654162e-06,
+      "loss": 0.556,
+      "step": 23283
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.260227583607687e-06,
+      "loss": 0.5569,
+      "step": 23284
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2596575717169047e-06,
+      "loss": 0.5908,
+      "step": 23285
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2590876801009154e-06,
+      "loss": 0.5352,
+      "step": 23286
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2585179087675603e-06,
+      "loss": 0.5168,
+      "step": 23287
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2579482577246804e-06,
+      "loss": 0.5296,
+      "step": 23288
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.257378726980113e-06,
+      "loss": 0.5902,
+      "step": 23289
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2568093165416972e-06,
+      "loss": 0.5256,
+      "step": 23290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.256240026417267e-06,
+      "loss": 0.5395,
+      "step": 23291
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2556708566146548e-06,
+      "loss": 0.57,
+      "step": 23292
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2551018071416975e-06,
+      "loss": 0.5919,
+      "step": 23293
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2545328780062238e-06,
+      "loss": 0.5887,
+      "step": 23294
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2539640692160616e-06,
+      "loss": 0.5918,
+      "step": 23295
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2533953807790356e-06,
+      "loss": 0.5721,
+      "step": 23296
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2528268127029786e-06,
+      "loss": 0.6019,
+      "step": 23297
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2522583649957088e-06,
+      "loss": 0.5804,
+      "step": 23298
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2516900376650498e-06,
+      "loss": 0.6153,
+      "step": 23299
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2511218307188233e-06,
+      "loss": 0.6054,
+      "step": 23300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2505537441648496e-06,
+      "loss": 0.6117,
+      "step": 23301
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2499857780109426e-06,
+      "loss": 0.5568,
+      "step": 23302
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2494179322649192e-06,
+      "loss": 0.592,
+      "step": 23303
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2488502069345954e-06,
+      "loss": 0.5499,
+      "step": 23304
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2482826020277816e-06,
+      "loss": 0.6235,
+      "step": 23305
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2477151175522894e-06,
+      "loss": 0.5811,
+      "step": 23306
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.247147753515926e-06,
+      "loss": 0.5326,
+      "step": 23307
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2465805099265026e-06,
+      "loss": 0.5879,
+      "step": 23308
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2460133867918224e-06,
+      "loss": 0.6218,
+      "step": 23309
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2454463841196884e-06,
+      "loss": 0.6066,
+      "step": 23310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.244879501917906e-06,
+      "loss": 0.5521,
+      "step": 23311
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2443127401942746e-06,
+      "loss": 0.5937,
+      "step": 23312
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.243746098956594e-06,
+      "loss": 0.6093,
+      "step": 23313
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2431795782126576e-06,
+      "loss": 0.6292,
+      "step": 23314
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2426131779702678e-06,
+      "loss": 0.6062,
+      "step": 23315
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2420468982372158e-06,
+      "loss": 0.5873,
+      "step": 23316
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2414807390212912e-06,
+      "loss": 0.6037,
+      "step": 23317
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2409147003302901e-06,
+      "loss": 0.5718,
+      "step": 23318
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2403487821719984e-06,
+      "loss": 0.5467,
+      "step": 23319
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2397829845542053e-06,
+      "loss": 0.5531,
+      "step": 23320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2392173074846925e-06,
+      "loss": 0.5935,
+      "step": 23321
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2386517509712503e-06,
+      "loss": 0.5723,
+      "step": 23322
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2380863150216583e-06,
+      "loss": 0.6136,
+      "step": 23323
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2375209996436977e-06,
+      "loss": 0.6274,
+      "step": 23324
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2369558048451447e-06,
+      "loss": 0.6241,
+      "step": 23325
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.236390730633783e-06,
+      "loss": 0.5999,
+      "step": 23326
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2358257770173842e-06,
+      "loss": 0.5327,
+      "step": 23327
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.235260944003721e-06,
+      "loss": 0.573,
+      "step": 23328
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2346962316005718e-06,
+      "loss": 0.5736,
+      "step": 23329
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2341316398157044e-06,
+      "loss": 0.5465,
+      "step": 23330
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2335671686568872e-06,
+      "loss": 0.5642,
+      "step": 23331
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2330028181318876e-06,
+      "loss": 0.5234,
+      "step": 23332
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2324385882484736e-06,
+      "loss": 0.5573,
+      "step": 23333
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2318744790144087e-06,
+      "loss": 0.526,
+      "step": 23334
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2313104904374562e-06,
+      "loss": 0.58,
+      "step": 23335
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2307466225253729e-06,
+      "loss": 0.564,
+      "step": 23336
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2301828752859246e-06,
+      "loss": 0.5694,
+      "step": 23337
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2296192487268644e-06,
+      "loss": 0.5323,
+      "step": 23338
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.229055742855949e-06,
+      "loss": 0.5058,
+      "step": 23339
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.228492357680935e-06,
+      "loss": 0.5168,
+      "step": 23340
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2279290932095734e-06,
+      "loss": 0.5724,
+      "step": 23341
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2273659494496148e-06,
+      "loss": 0.5608,
+      "step": 23342
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2268029264088076e-06,
+      "loss": 0.5402,
+      "step": 23343
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2262400240949023e-06,
+      "loss": 0.571,
+      "step": 23344
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2256772425156449e-06,
+      "loss": 0.625,
+      "step": 23345
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2251145816787768e-06,
+      "loss": 0.5701,
+      "step": 23346
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.224552041592042e-06,
+      "loss": 0.5965,
+      "step": 23347
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.223989622263183e-06,
+      "loss": 0.5648,
+      "step": 23348
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2234273236999371e-06,
+      "loss": 0.594,
+      "step": 23349
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2228651459100427e-06,
+      "loss": 0.5715,
+      "step": 23350
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2223030889012377e-06,
+      "loss": 0.6448,
+      "step": 23351
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2217411526812539e-06,
+      "loss": 0.5605,
+      "step": 23352
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2211793372578262e-06,
+      "loss": 0.5226,
+      "step": 23353
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2206176426386828e-06,
+      "loss": 0.6205,
+      "step": 23354
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2200560688315567e-06,
+      "loss": 0.5661,
+      "step": 23355
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2194946158441733e-06,
+      "loss": 0.6026,
+      "step": 23356
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2189332836842593e-06,
+      "loss": 0.5874,
+      "step": 23357
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.218372072359537e-06,
+      "loss": 0.5789,
+      "step": 23358
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2178109818777327e-06,
+      "loss": 0.5868,
+      "step": 23359
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.217250012246567e-06,
+      "loss": 0.5336,
+      "step": 23360
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2166891634737564e-06,
+      "loss": 0.6235,
+      "step": 23361
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2161284355670223e-06,
+      "loss": 0.5686,
+      "step": 23362
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2155678285340789e-06,
+      "loss": 0.5939,
+      "step": 23363
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2150073423826403e-06,
+      "loss": 0.5405,
+      "step": 23364
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2144469771204193e-06,
+      "loss": 0.6043,
+      "step": 23365
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2138867327551285e-06,
+      "loss": 0.5587,
+      "step": 23366
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2133266092944763e-06,
+      "loss": 0.5846,
+      "step": 23367
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.212766606746172e-06,
+      "loss": 0.6321,
+      "step": 23368
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.212206725117917e-06,
+      "loss": 0.5773,
+      "step": 23369
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.211646964417421e-06,
+      "loss": 0.599,
+      "step": 23370
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2110873246523857e-06,
+      "loss": 0.4896,
+      "step": 23371
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.210527805830508e-06,
+      "loss": 0.6106,
+      "step": 23372
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2099684079594943e-06,
+      "loss": 0.6128,
+      "step": 23373
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2094091310470391e-06,
+      "loss": 0.5778,
+      "step": 23374
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2088499751008377e-06,
+      "loss": 0.5817,
+      "step": 23375
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2082909401285825e-06,
+      "loss": 0.5598,
+      "step": 23376
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2077320261379722e-06,
+      "loss": 0.5692,
+      "step": 23377
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2071732331366936e-06,
+      "loss": 0.5221,
+      "step": 23378
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2066145611324366e-06,
+      "loss": 0.6072,
+      "step": 23379
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.206056010132891e-06,
+      "loss": 0.5782,
+      "step": 23380
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2054975801457424e-06,
+      "loss": 0.5196,
+      "step": 23381
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2049392711786733e-06,
+      "loss": 0.5759,
+      "step": 23382
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2043810832393676e-06,
+      "loss": 0.5684,
+      "step": 23383
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.203823016335508e-06,
+      "loss": 0.5924,
+      "step": 23384
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2032650704747716e-06,
+      "loss": 0.5767,
+      "step": 23385
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2027072456648392e-06,
+      "loss": 0.5421,
+      "step": 23386
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2021495419133822e-06,
+      "loss": 0.5552,
+      "step": 23387
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.20159195922808e-06,
+      "loss": 0.5536,
+      "step": 23388
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2010344976166033e-06,
+      "loss": 0.6301,
+      "step": 23389
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2004771570866214e-06,
+      "loss": 0.5661,
+      "step": 23390
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.199919937645807e-06,
+      "loss": 0.5981,
+      "step": 23391
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1993628393018276e-06,
+      "loss": 0.5639,
+      "step": 23392
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1988058620623488e-06,
+      "loss": 0.5779,
+      "step": 23393
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1982490059350316e-06,
+      "loss": 0.5642,
+      "step": 23394
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1976922709275452e-06,
+      "loss": 0.5273,
+      "step": 23395
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1971356570475467e-06,
+      "loss": 0.5713,
+      "step": 23396
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1965791643026969e-06,
+      "loss": 0.6016,
+      "step": 23397
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1960227927006508e-06,
+      "loss": 0.5234,
+      "step": 23398
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1954665422490686e-06,
+      "loss": 0.6013,
+      "step": 23399
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1949104129556034e-06,
+      "loss": 0.5527,
+      "step": 23400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1943544048279065e-06,
+      "loss": 0.5585,
+      "step": 23401
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1937985178736312e-06,
+      "loss": 0.5667,
+      "step": 23402
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1932427521004275e-06,
+      "loss": 0.632,
+      "step": 23403
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1926871075159406e-06,
+      "loss": 0.6373,
+      "step": 23404
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1921315841278158e-06,
+      "loss": 0.5524,
+      "step": 23405
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1915761819437021e-06,
+      "loss": 0.5794,
+      "step": 23406
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1910209009712403e-06,
+      "loss": 0.6215,
+      "step": 23407
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.190465741218071e-06,
+      "loss": 0.6084,
+      "step": 23408
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.189910702691831e-06,
+      "loss": 0.5792,
+      "step": 23409
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1893557854001636e-06,
+      "loss": 0.5862,
+      "step": 23410
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1888009893507024e-06,
+      "loss": 0.6055,
+      "step": 23411
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1882463145510792e-06,
+      "loss": 0.6107,
+      "step": 23412
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1876917610089323e-06,
+      "loss": 0.6363,
+      "step": 23413
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.18713732873189e-06,
+      "loss": 0.5597,
+      "step": 23414
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1865830177275805e-06,
+      "loss": 0.5511,
+      "step": 23415
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1860288280036314e-06,
+      "loss": 0.4935,
+      "step": 23416
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.185474759567673e-06,
+      "loss": 0.601,
+      "step": 23417
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.184920812427326e-06,
+      "loss": 0.6172,
+      "step": 23418
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1843669865902152e-06,
+      "loss": 0.5769,
+      "step": 23419
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.183813282063958e-06,
+      "loss": 0.5651,
+      "step": 23420
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1832596988561785e-06,
+      "loss": 0.555,
+      "step": 23421
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1827062369744924e-06,
+      "loss": 0.5298,
+      "step": 23422
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.182152896426515e-06,
+      "loss": 0.5451,
+      "step": 23423
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1815996772198635e-06,
+      "loss": 0.5575,
+      "step": 23424
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1810465793621484e-06,
+      "loss": 0.5633,
+      "step": 23425
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1804936028609826e-06,
+      "loss": 0.5295,
+      "step": 23426
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.179940747723971e-06,
+      "loss": 0.6044,
+      "step": 23427
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1793880139587266e-06,
+      "loss": 0.5894,
+      "step": 23428
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1788354015728543e-06,
+      "loss": 0.5433,
+      "step": 23429
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1782829105739578e-06,
+      "loss": 0.5894,
+      "step": 23430
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1777305409696382e-06,
+      "loss": 0.5788,
+      "step": 23431
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1771782927675001e-06,
+      "loss": 0.6559,
+      "step": 23432
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.176626165975141e-06,
+      "loss": 0.6076,
+      "step": 23433
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1760741606001558e-06,
+      "loss": 0.5834,
+      "step": 23434
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1755222766501473e-06,
+      "loss": 0.5368,
+      "step": 23435
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.174970514132705e-06,
+      "loss": 0.5983,
+      "step": 23436
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.174418873055423e-06,
+      "loss": 0.5343,
+      "step": 23437
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1738673534258894e-06,
+      "loss": 0.5047,
+      "step": 23438
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1733159552516992e-06,
+      "loss": 0.5852,
+      "step": 23439
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.172764678540438e-06,
+      "loss": 0.5184,
+      "step": 23440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1722135232996879e-06,
+      "loss": 0.5748,
+      "step": 23441
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1716624895370387e-06,
+      "loss": 0.5193,
+      "step": 23442
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1711115772600712e-06,
+      "loss": 0.6011,
+      "step": 23443
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1705607864763657e-06,
+      "loss": 0.6209,
+      "step": 23444
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1700101171935008e-06,
+      "loss": 0.5188,
+      "step": 23445
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1694595694190557e-06,
+      "loss": 0.5851,
+      "step": 23446
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1689091431606082e-06,
+      "loss": 0.6059,
+      "step": 23447
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1683588384257294e-06,
+      "loss": 0.528,
+      "step": 23448
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1678086552219915e-06,
+      "loss": 0.6053,
+      "step": 23449
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1672585935569691e-06,
+      "loss": 0.5435,
+      "step": 23450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1667086534382287e-06,
+      "loss": 0.5438,
+      "step": 23451
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1661588348733377e-06,
+      "loss": 0.5851,
+      "step": 23452
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1656091378698653e-06,
+      "loss": 0.5938,
+      "step": 23453
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1650595624353744e-06,
+      "loss": 0.5629,
+      "step": 23454
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1645101085774258e-06,
+      "loss": 0.6186,
+      "step": 23455
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1639607763035798e-06,
+      "loss": 0.6217,
+      "step": 23456
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1634115656213995e-06,
+      "loss": 0.5776,
+      "step": 23457
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.162862476538441e-06,
+      "loss": 0.5943,
+      "step": 23458
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1623135090622606e-06,
+      "loss": 0.5186,
+      "step": 23459
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1617646632004087e-06,
+      "loss": 0.6217,
+      "step": 23460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1612159389604426e-06,
+      "loss": 0.5267,
+      "step": 23461
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1606673363499121e-06,
+      "loss": 0.5953,
+      "step": 23462
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1601188553763643e-06,
+      "loss": 0.5758,
+      "step": 23463
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1595704960473498e-06,
+      "loss": 0.645,
+      "step": 23464
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1590222583704135e-06,
+      "loss": 0.6358,
+      "step": 23465
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1584741423530999e-06,
+      "loss": 0.589,
+      "step": 23466
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.157926148002948e-06,
+      "loss": 0.6143,
+      "step": 23467
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1573782753275042e-06,
+      "loss": 0.6017,
+      "step": 23468
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1568305243343047e-06,
+      "loss": 0.5576,
+      "step": 23469
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.156282895030888e-06,
+      "loss": 0.5651,
+      "step": 23470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1557353874247868e-06,
+      "loss": 0.5203,
+      "step": 23471
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1551880015235406e-06,
+      "loss": 0.5874,
+      "step": 23472
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.154640737334679e-06,
+      "loss": 0.5713,
+      "step": 23473
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1540935948657305e-06,
+      "loss": 0.5414,
+      "step": 23474
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.153546574124229e-06,
+      "loss": 0.547,
+      "step": 23475
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1529996751177007e-06,
+      "loss": 0.5923,
+      "step": 23476
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1524528978536698e-06,
+      "loss": 0.5634,
+      "step": 23477
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.15190624233966e-06,
+      "loss": 0.5925,
+      "step": 23478
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1513597085831952e-06,
+      "loss": 0.6015,
+      "step": 23479
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1508132965917974e-06,
+      "loss": 0.5705,
+      "step": 23480
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1502670063729837e-06,
+      "loss": 0.6248,
+      "step": 23481
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1497208379342706e-06,
+      "loss": 0.5804,
+      "step": 23482
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1491747912831763e-06,
+      "loss": 0.609,
+      "step": 23483
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1486288664272138e-06,
+      "loss": 0.5994,
+      "step": 23484
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1480830633738949e-06,
+      "loss": 0.5047,
+      "step": 23485
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1475373821307322e-06,
+      "loss": 0.5695,
+      "step": 23486
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1469918227052335e-06,
+      "loss": 0.5717,
+      "step": 23487
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.146446385104907e-06,
+      "loss": 0.627,
+      "step": 23488
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.145901069337254e-06,
+      "loss": 0.5852,
+      "step": 23489
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.145355875409786e-06,
+      "loss": 0.5795,
+      "step": 23490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.144810803330001e-06,
+      "loss": 0.5966,
+      "step": 23491
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1442658531053996e-06,
+      "loss": 0.5659,
+      "step": 23492
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1437210247434793e-06,
+      "loss": 0.583,
+      "step": 23493
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.143176318251743e-06,
+      "loss": 0.6003,
+      "step": 23494
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1426317336376812e-06,
+      "loss": 0.5874,
+      "step": 23495
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1420872709087882e-06,
+      "loss": 0.6181,
+      "step": 23496
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.14154293007256e-06,
+      "loss": 0.6033,
+      "step": 23497
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1409987111364863e-06,
+      "loss": 0.5627,
+      "step": 23498
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1404546141080497e-06,
+      "loss": 0.609,
+      "step": 23499
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1399106389947457e-06,
+      "loss": 0.6158,
+      "step": 23500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1393667858040557e-06,
+      "loss": 0.501,
+      "step": 23501
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1388230545434652e-06,
+      "loss": 0.5691,
+      "step": 23502
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1382794452204527e-06,
+      "loss": 0.6212,
+      "step": 23503
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.137735957842504e-06,
+      "loss": 0.6502,
+      "step": 23504
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1371925924170957e-06,
+      "loss": 0.5614,
+      "step": 23505
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1366493489517028e-06,
+      "loss": 0.5643,
+      "step": 23506
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1361062274538037e-06,
+      "loss": 0.6405,
+      "step": 23507
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1355632279308726e-06,
+      "loss": 0.593,
+      "step": 23508
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1350203503903812e-06,
+      "loss": 0.6092,
+      "step": 23509
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1344775948397956e-06,
+      "loss": 0.5101,
+      "step": 23510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1339349612865912e-06,
+      "loss": 0.588,
+      "step": 23511
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.133392449738232e-06,
+      "loss": 0.5836,
+      "step": 23512
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.132850060202183e-06,
+      "loss": 0.5389,
+      "step": 23513
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1323077926859072e-06,
+      "loss": 0.5901,
+      "step": 23514
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1317656471968696e-06,
+      "loss": 0.5909,
+      "step": 23515
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.13122362374253e-06,
+      "loss": 0.5604,
+      "step": 23516
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1306817223303424e-06,
+      "loss": 0.5103,
+      "step": 23517
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1301399429677705e-06,
+      "loss": 0.5704,
+      "step": 23518
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1295982856622677e-06,
+      "loss": 0.6153,
+      "step": 23519
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1290567504212857e-06,
+      "loss": 0.6038,
+      "step": 23520
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1285153372522751e-06,
+      "loss": 0.568,
+      "step": 23521
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1279740461626908e-06,
+      "loss": 0.5744,
+      "step": 23522
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1274328771599797e-06,
+      "loss": 0.613,
+      "step": 23523
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1268918302515884e-06,
+      "loss": 0.6779,
+      "step": 23524
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1263509054449596e-06,
+      "loss": 0.4949,
+      "step": 23525
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.125810102747542e-06,
+      "loss": 0.5506,
+      "step": 23526
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1252694221667747e-06,
+      "loss": 0.5426,
+      "step": 23527
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.124728863710096e-06,
+      "loss": 0.5681,
+      "step": 23528
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.124188427384949e-06,
+      "loss": 0.601,
+      "step": 23529
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1236481131987676e-06,
+      "loss": 0.5439,
+      "step": 23530
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.123107921158988e-06,
+      "loss": 0.552,
+      "step": 23531
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1225678512730409e-06,
+      "loss": 0.568,
+      "step": 23532
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.122027903548364e-06,
+      "loss": 0.5519,
+      "step": 23533
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.121488077992382e-06,
+      "loss": 0.5756,
+      "step": 23534
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.120948374612527e-06,
+      "loss": 0.5642,
+      "step": 23535
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1204087934162211e-06,
+      "loss": 0.5574,
+      "step": 23536
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1198693344108957e-06,
+      "loss": 0.5137,
+      "step": 23537
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1193299976039695e-06,
+      "loss": 0.6345,
+      "step": 23538
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1187907830028644e-06,
+      "loss": 0.5217,
+      "step": 23539
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1182516906150042e-06,
+      "loss": 0.4869,
+      "step": 23540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1177127204478044e-06,
+      "loss": 0.5233,
+      "step": 23541
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1171738725086833e-06,
+      "loss": 0.5247,
+      "step": 23542
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1166351468050517e-06,
+      "loss": 0.563,
+      "step": 23543
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1160965433443283e-06,
+      "loss": 0.5829,
+      "step": 23544
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1155580621339234e-06,
+      "loss": 0.5886,
+      "step": 23545
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1150197031812427e-06,
+      "loss": 0.5547,
+      "step": 23546
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.114481466493701e-06,
+      "loss": 0.5577,
+      "step": 23547
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1139433520787014e-06,
+      "loss": 0.5544,
+      "step": 23548
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.113405359943649e-06,
+      "loss": 0.5804,
+      "step": 23549
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1128674900959457e-06,
+      "loss": 0.5525,
+      "step": 23550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1123297425429968e-06,
+      "loss": 0.6311,
+      "step": 23551
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1117921172921997e-06,
+      "loss": 0.6072,
+      "step": 23552
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1112546143509528e-06,
+      "loss": 0.5165,
+      "step": 23553
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1107172337266515e-06,
+      "loss": 0.5851,
+      "step": 23554
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.110179975426693e-06,
+      "loss": 0.5474,
+      "step": 23555
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1096428394584701e-06,
+      "loss": 0.5795,
+      "step": 23556
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1091058258293708e-06,
+      "loss": 0.5556,
+      "step": 23557
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1085689345467899e-06,
+      "loss": 0.6031,
+      "step": 23558
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1080321656181125e-06,
+      "loss": 0.5639,
+      "step": 23559
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1074955190507264e-06,
+      "loss": 0.5472,
+      "step": 23560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1069589948520131e-06,
+      "loss": 0.6287,
+      "step": 23561
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1064225930293616e-06,
+      "loss": 0.6042,
+      "step": 23562
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1058863135901476e-06,
+      "loss": 0.5563,
+      "step": 23563
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1053501565417545e-06,
+      "loss": 0.5569,
+      "step": 23564
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1048141218915565e-06,
+      "loss": 0.5838,
+      "step": 23565
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.104278209646934e-06,
+      "loss": 0.6421,
+      "step": 23566
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.103742419815259e-06,
+      "loss": 0.5761,
+      "step": 23567
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1032067524039047e-06,
+      "loss": 0.4927,
+      "step": 23568
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1026712074202439e-06,
+      "loss": 0.5549,
+      "step": 23569
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1021357848716462e-06,
+      "loss": 0.4882,
+      "step": 23570
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.101600484765477e-06,
+      "loss": 0.5563,
+      "step": 23571
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1010653071091027e-06,
+      "loss": 0.6237,
+      "step": 23572
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1005302519098915e-06,
+      "loss": 0.5821,
+      "step": 23573
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0999953191752034e-06,
+      "loss": 0.5348,
+      "step": 23574
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0994605089123999e-06,
+      "loss": 0.5885,
+      "step": 23575
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0989258211288379e-06,
+      "loss": 0.5839,
+      "step": 23576
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.098391255831881e-06,
+      "loss": 0.5774,
+      "step": 23577
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0978568130288813e-06,
+      "loss": 0.568,
+      "step": 23578
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0973224927271919e-06,
+      "loss": 0.582,
+      "step": 23579
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0967882949341701e-06,
+      "loss": 0.5786,
+      "step": 23580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0962542196571636e-06,
+      "loss": 0.529,
+      "step": 23581
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0957202669035228e-06,
+      "loss": 0.6054,
+      "step": 23582
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0951864366805931e-06,
+      "loss": 0.6143,
+      "step": 23583
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0946527289957255e-06,
+      "loss": 0.5627,
+      "step": 23584
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0941191438562604e-06,
+      "loss": 0.586,
+      "step": 23585
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0935856812695412e-06,
+      "loss": 0.6422,
+      "step": 23586
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0930523412429072e-06,
+      "loss": 0.5467,
+      "step": 23587
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0925191237837007e-06,
+      "loss": 0.5569,
+      "step": 23588
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0919860288992578e-06,
+      "loss": 0.5421,
+      "step": 23589
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0914530565969139e-06,
+      "loss": 0.5301,
+      "step": 23590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.090920206884004e-06,
+      "loss": 0.5483,
+      "step": 23591
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0903874797678616e-06,
+      "loss": 0.5505,
+      "step": 23592
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0898548752558148e-06,
+      "loss": 0.6095,
+      "step": 23593
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0893223933551932e-06,
+      "loss": 0.5821,
+      "step": 23594
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0887900340733271e-06,
+      "loss": 0.4873,
+      "step": 23595
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.088257797417539e-06,
+      "loss": 0.635,
+      "step": 23596
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0877256833951556e-06,
+      "loss": 0.5215,
+      "step": 23597
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.087193692013494e-06,
+      "loss": 0.5317,
+      "step": 23598
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.086661823279882e-06,
+      "loss": 0.565,
+      "step": 23599
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0861300772016347e-06,
+      "loss": 0.616,
+      "step": 23600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0855984537860686e-06,
+      "loss": 0.5405,
+      "step": 23601
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0850669530405022e-06,
+      "loss": 0.5453,
+      "step": 23602
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0845355749722486e-06,
+      "loss": 0.564,
+      "step": 23603
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0840043195886185e-06,
+      "loss": 0.5788,
+      "step": 23604
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0834731868969217e-06,
+      "loss": 0.6,
+      "step": 23605
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0829421769044702e-06,
+      "loss": 0.5632,
+      "step": 23606
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.082411289618569e-06,
+      "loss": 0.579,
+      "step": 23607
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0818805250465237e-06,
+      "loss": 0.6243,
+      "step": 23608
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0813498831956393e-06,
+      "loss": 0.5858,
+      "step": 23609
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0808193640732178e-06,
+      "loss": 0.5173,
+      "step": 23610
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.080288967686558e-06,
+      "loss": 0.5915,
+      "step": 23611
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0797586940429582e-06,
+      "loss": 0.6523,
+      "step": 23612
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0792285431497184e-06,
+      "loss": 0.6261,
+      "step": 23613
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0786985150141327e-06,
+      "loss": 0.5673,
+      "step": 23614
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0781686096434952e-06,
+      "loss": 0.5435,
+      "step": 23615
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0776388270450933e-06,
+      "loss": 0.6231,
+      "step": 23616
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0771091672262246e-06,
+      "loss": 0.5521,
+      "step": 23617
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0765796301941734e-06,
+      "loss": 0.5327,
+      "step": 23618
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0760502159562247e-06,
+      "loss": 0.6272,
+      "step": 23619
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0755209245196695e-06,
+      "loss": 0.5282,
+      "step": 23620
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0749917558917867e-06,
+      "loss": 0.566,
+      "step": 23621
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0744627100798611e-06,
+      "loss": 0.5383,
+      "step": 23622
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0739337870911681e-06,
+      "loss": 0.6029,
+      "step": 23623
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.073404986932992e-06,
+      "loss": 0.5442,
+      "step": 23624
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.072876309612606e-06,
+      "loss": 0.5527,
+      "step": 23625
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0723477551372873e-06,
+      "loss": 0.5664,
+      "step": 23626
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0718193235143048e-06,
+      "loss": 0.5978,
+      "step": 23627
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0712910147509358e-06,
+      "loss": 0.561,
+      "step": 23628
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0707628288544481e-06,
+      "loss": 0.554,
+      "step": 23629
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0702347658321077e-06,
+      "loss": 0.5565,
+      "step": 23630
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0697068256911859e-06,
+      "loss": 0.5602,
+      "step": 23631
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0691790084389443e-06,
+      "loss": 0.5178,
+      "step": 23632
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0686513140826472e-06,
+      "loss": 0.5428,
+      "step": 23633
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0681237426295531e-06,
+      "loss": 0.5419,
+      "step": 23634
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0675962940869266e-06,
+      "loss": 0.5978,
+      "step": 23635
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.067068968462025e-06,
+      "loss": 0.5665,
+      "step": 23636
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0665417657621025e-06,
+      "loss": 0.6054,
+      "step": 23637
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0660146859944131e-06,
+      "loss": 0.5062,
+      "step": 23638
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0654877291662136e-06,
+      "loss": 0.4893,
+      "step": 23639
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0649608952847535e-06,
+      "loss": 0.6254,
+      "step": 23640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0644341843572815e-06,
+      "loss": 0.5867,
+      "step": 23641
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0639075963910473e-06,
+      "loss": 0.5034,
+      "step": 23642
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0633811313932974e-06,
+      "loss": 0.6223,
+      "step": 23643
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.062854789371275e-06,
+      "loss": 0.5555,
+      "step": 23644
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0623285703322216e-06,
+      "loss": 0.5508,
+      "step": 23645
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.061802474283382e-06,
+      "loss": 0.5641,
+      "step": 23646
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0612765012319948e-06,
+      "loss": 0.5615,
+      "step": 23647
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0607506511852972e-06,
+      "loss": 0.5522,
+      "step": 23648
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0602249241505224e-06,
+      "loss": 0.5286,
+      "step": 23649
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0596993201349115e-06,
+      "loss": 0.6098,
+      "step": 23650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0591738391456918e-06,
+      "loss": 0.5714,
+      "step": 23651
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0586484811900955e-06,
+      "loss": 0.549,
+      "step": 23652
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0581232462753544e-06,
+      "loss": 0.5558,
+      "step": 23653
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0575981344086939e-06,
+      "loss": 0.6217,
+      "step": 23654
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0570731455973415e-06,
+      "loss": 0.6059,
+      "step": 23655
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0565482798485172e-06,
+      "loss": 0.5487,
+      "step": 23656
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0560235371694505e-06,
+      "loss": 0.5926,
+      "step": 23657
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.055498917567358e-06,
+      "loss": 0.5741,
+      "step": 23658
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0549744210494573e-06,
+      "loss": 0.5321,
+      "step": 23659
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0544500476229713e-06,
+      "loss": 0.6189,
+      "step": 23660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0539257972951123e-06,
+      "loss": 0.5822,
+      "step": 23661
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0534016700730954e-06,
+      "loss": 0.6446,
+      "step": 23662
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0528776659641305e-06,
+      "loss": 0.6048,
+      "step": 23663
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0523537849754316e-06,
+      "loss": 0.5982,
+      "step": 23664
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.051830027114208e-06,
+      "loss": 0.5226,
+      "step": 23665
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0513063923876655e-06,
+      "loss": 0.5732,
+      "step": 23666
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0507828808030075e-06,
+      "loss": 0.614,
+      "step": 23667
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0502594923674437e-06,
+      "loss": 0.5701,
+      "step": 23668
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0497362270881718e-06,
+      "loss": 0.6306,
+      "step": 23669
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0492130849723924e-06,
+      "loss": 0.5813,
+      "step": 23670
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0486900660273081e-06,
+      "loss": 0.5456,
+      "step": 23671
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0481671702601138e-06,
+      "loss": 0.5817,
+      "step": 23672
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.047644397678005e-06,
+      "loss": 0.6268,
+      "step": 23673
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0471217482881723e-06,
+      "loss": 0.5821,
+      "step": 23674
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.046599222097815e-06,
+      "loss": 0.6132,
+      "step": 23675
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.046076819114118e-06,
+      "loss": 0.6253,
+      "step": 23676
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0455545393442722e-06,
+      "loss": 0.5311,
+      "step": 23677
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0450323827954623e-06,
+      "loss": 0.5418,
+      "step": 23678
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0445103494748766e-06,
+      "loss": 0.6161,
+      "step": 23679
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0439884393896982e-06,
+      "loss": 0.5844,
+      "step": 23680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.043466652547106e-06,
+      "loss": 0.5145,
+      "step": 23681
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0429449889542842e-06,
+      "loss": 0.5513,
+      "step": 23682
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0424234486184104e-06,
+      "loss": 0.575,
+      "step": 23683
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.04190203154666e-06,
+      "loss": 0.5663,
+      "step": 23684
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.041380737746207e-06,
+      "loss": 0.5885,
+      "step": 23685
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0408595672242293e-06,
+      "loss": 0.59,
+      "step": 23686
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0403385199878957e-06,
+      "loss": 0.5688,
+      "step": 23687
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.039817596044378e-06,
+      "loss": 0.6768,
+      "step": 23688
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0392967954008394e-06,
+      "loss": 0.5476,
+      "step": 23689
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.038776118064454e-06,
+      "loss": 0.5915,
+      "step": 23690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.038255564042382e-06,
+      "loss": 0.6149,
+      "step": 23691
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0377351333417862e-06,
+      "loss": 0.5169,
+      "step": 23692
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0372148259698323e-06,
+      "loss": 0.6636,
+      "step": 23693
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0366946419336775e-06,
+      "loss": 0.6233,
+      "step": 23694
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.036174581240481e-06,
+      "loss": 0.5435,
+      "step": 23695
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0356546438973958e-06,
+      "loss": 0.5791,
+      "step": 23696
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0351348299115816e-06,
+      "loss": 0.5936,
+      "step": 23697
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0346151392901892e-06,
+      "loss": 0.5459,
+      "step": 23698
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.034095572040371e-06,
+      "loss": 0.5477,
+      "step": 23699
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0335761281692736e-06,
+      "loss": 0.5725,
+      "step": 23700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0330568076840486e-06,
+      "loss": 0.6039,
+      "step": 23701
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0325376105918417e-06,
+      "loss": 0.6424,
+      "step": 23702
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0320185368997937e-06,
+      "loss": 0.6372,
+      "step": 23703
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0314995866150535e-06,
+      "loss": 0.5628,
+      "step": 23704
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0309807597447585e-06,
+      "loss": 0.562,
+      "step": 23705
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0304620562960498e-06,
+      "loss": 0.5423,
+      "step": 23706
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0299434762760629e-06,
+      "loss": 0.5054,
+      "step": 23707
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0294250196919375e-06,
+      "loss": 0.5473,
+      "step": 23708
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0289066865508079e-06,
+      "loss": 0.5975,
+      "step": 23709
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0283884768598008e-06,
+      "loss": 0.626,
+      "step": 23710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0278703906260534e-06,
+      "loss": 0.6186,
+      "step": 23711
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0273524278566927e-06,
+      "loss": 0.584,
+      "step": 23712
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.026834588558845e-06,
+      "loss": 0.6011,
+      "step": 23713
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.02631687273964e-06,
+      "loss": 0.5269,
+      "step": 23714
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0257992804062e-06,
+      "loss": 0.6406,
+      "step": 23715
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.025281811565646e-06,
+      "loss": 0.497,
+      "step": 23716
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0247644662251e-06,
+      "loss": 0.5638,
+      "step": 23717
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0242472443916817e-06,
+      "loss": 0.5828,
+      "step": 23718
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0237301460725091e-06,
+      "loss": 0.5618,
+      "step": 23719
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0232131712746973e-06,
+      "loss": 0.5646,
+      "step": 23720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0226963200053564e-06,
+      "loss": 0.5753,
+      "step": 23721
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.022179592271606e-06,
+      "loss": 0.6383,
+      "step": 23722
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0216629880805517e-06,
+      "loss": 0.594,
+      "step": 23723
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0211465074393023e-06,
+      "loss": 0.4842,
+      "step": 23724
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0206301503549688e-06,
+      "loss": 0.6363,
+      "step": 23725
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.020113916834654e-06,
+      "loss": 0.5669,
+      "step": 23726
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0195978068854618e-06,
+      "loss": 0.5951,
+      "step": 23727
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0190818205144938e-06,
+      "loss": 0.5338,
+      "step": 23728
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.018565957728852e-06,
+      "loss": 0.5751,
+      "step": 23729
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0180502185356345e-06,
+      "loss": 0.5324,
+      "step": 23730
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0175346029419387e-06,
+      "loss": 0.5615,
+      "step": 23731
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0170191109548577e-06,
+      "loss": 0.5403,
+      "step": 23732
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0165037425814883e-06,
+      "loss": 0.5343,
+      "step": 23733
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0159884978289215e-06,
+      "loss": 0.5407,
+      "step": 23734
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0154733767042446e-06,
+      "loss": 0.546,
+      "step": 23735
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0149583792145502e-06,
+      "loss": 0.6153,
+      "step": 23736
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0144435053669244e-06,
+      "loss": 0.5604,
+      "step": 23737
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0139287551684507e-06,
+      "loss": 0.6163,
+      "step": 23738
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0134141286262112e-06,
+      "loss": 0.5311,
+      "step": 23739
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0128996257472924e-06,
+      "loss": 0.543,
+      "step": 23740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0123852465387706e-06,
+      "loss": 0.56,
+      "step": 23741
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.011870991007725e-06,
+      "loss": 0.5718,
+      "step": 23742
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.011356859161231e-06,
+      "loss": 0.583,
+      "step": 23743
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0108428510063661e-06,
+      "loss": 0.498,
+      "step": 23744
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0103289665502025e-06,
+      "loss": 0.5354,
+      "step": 23745
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.009815205799809e-06,
+      "loss": 0.5283,
+      "step": 23746
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0093015687622598e-06,
+      "loss": 0.6608,
+      "step": 23747
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0087880554446205e-06,
+      "loss": 0.6253,
+      "step": 23748
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0082746658539588e-06,
+      "loss": 0.5675,
+      "step": 23749
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0077613999973357e-06,
+      "loss": 0.4951,
+      "step": 23750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0072482578818198e-06,
+      "loss": 0.5608,
+      "step": 23751
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0067352395144692e-06,
+      "loss": 0.5703,
+      "step": 23752
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0062223449023446e-06,
+      "loss": 0.6423,
+      "step": 23753
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0057095740525003e-06,
+      "loss": 0.6292,
+      "step": 23754
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0051969269719974e-06,
+      "loss": 0.5396,
+      "step": 23755
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0046844036678882e-06,
+      "loss": 0.6047,
+      "step": 23756
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0041720041472236e-06,
+      "loss": 0.5422,
+      "step": 23757
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.003659728417059e-06,
+      "loss": 0.6322,
+      "step": 23758
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0031475764844412e-06,
+      "loss": 0.5735,
+      "step": 23759
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0026355483564176e-06,
+      "loss": 0.6021,
+      "step": 23760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0021236440400339e-06,
+      "loss": 0.5342,
+      "step": 23761
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0016118635423366e-06,
+      "loss": 0.5509,
+      "step": 23762
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0011002068703657e-06,
+      "loss": 0.5841,
+      "step": 23763
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0005886740311621e-06,
+      "loss": 0.5028,
+      "step": 23764
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0000772650317682e-06,
+      "loss": 0.5579,
+      "step": 23765
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.99565979879218e-07,
+      "loss": 0.5941,
+      "step": 23766
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.990548185805493e-07,
+      "loss": 0.5451,
+      "step": 23767
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.985437811427934e-07,
+      "loss": 0.5498,
+      "step": 23768
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.980328675729855e-07,
+      "loss": 0.5079,
+      "step": 23769
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.975220778781558e-07,
+      "loss": 0.5894,
+      "step": 23770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.97011412065333e-07,
+      "loss": 0.5664,
+      "step": 23771
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.965008701415412e-07,
+      "loss": 0.5847,
+      "step": 23772
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.95990452113811e-07,
+      "loss": 0.6458,
+      "step": 23773
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.954801579891627e-07,
+      "loss": 0.5183,
+      "step": 23774
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.94969987774619e-07,
+      "loss": 0.5437,
+      "step": 23775
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.944599414772017e-07,
+      "loss": 0.5908,
+      "step": 23776
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.939500191039286e-07,
+      "loss": 0.6379,
+      "step": 23777
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.934402206618176e-07,
+      "loss": 0.6061,
+      "step": 23778
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.929305461578799e-07,
+      "loss": 0.5878,
+      "step": 23779
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.924209955991348e-07,
+      "loss": 0.5174,
+      "step": 23780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.919115689925918e-07,
+      "loss": 0.5596,
+      "step": 23781
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.914022663452605e-07,
+      "loss": 0.6001,
+      "step": 23782
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.908930876641477e-07,
+      "loss": 0.5672,
+      "step": 23783
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.903840329562642e-07,
+      "loss": 0.5365,
+      "step": 23784
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.898751022286124e-07,
+      "loss": 0.5206,
+      "step": 23785
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.893662954881945e-07,
+      "loss": 0.5912,
+      "step": 23786
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.888576127420158e-07,
+      "loss": 0.5614,
+      "step": 23787
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.883490539970741e-07,
+      "loss": 0.6106,
+      "step": 23788
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.878406192603684e-07,
+      "loss": 0.4881,
+      "step": 23789
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.87332308538892e-07,
+      "loss": 0.561,
+      "step": 23790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.868241218396447e-07,
+      "loss": 0.5945,
+      "step": 23791
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.863160591696163e-07,
+      "loss": 0.5632,
+      "step": 23792
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.858081205357993e-07,
+      "loss": 0.579,
+      "step": 23793
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.853003059451827e-07,
+      "loss": 0.5986,
+      "step": 23794
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.84792615404755e-07,
+      "loss": 0.6158,
+      "step": 23795
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.842850489215028e-07,
+      "loss": 0.5379,
+      "step": 23796
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.837776065024086e-07,
+      "loss": 0.5935,
+      "step": 23797
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.832702881544588e-07,
+      "loss": 0.5291,
+      "step": 23798
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.827630938846323e-07,
+      "loss": 0.5661,
+      "step": 23799
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.82256023699909e-07,
+      "loss": 0.5165,
+      "step": 23800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.817490776072647e-07,
+      "loss": 0.5953,
+      "step": 23801
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.81242255613679e-07,
+      "loss": 0.5119,
+      "step": 23802
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.807355577261246e-07,
+      "loss": 0.5985,
+      "step": 23803
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.802289839515743e-07,
+      "loss": 0.5952,
+      "step": 23804
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.797225342969963e-07,
+      "loss": 0.5746,
+      "step": 23805
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.792162087693634e-07,
+      "loss": 0.5004,
+      "step": 23806
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.787100073756428e-07,
+      "loss": 0.6538,
+      "step": 23807
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.782039301227974e-07,
+      "loss": 0.5637,
+      "step": 23808
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.776979770177942e-07,
+      "loss": 0.6041,
+      "step": 23809
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.77192148067595e-07,
+      "loss": 0.5784,
+      "step": 23810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.7668644327916e-07,
+      "loss": 0.602,
+      "step": 23811
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.76180862659446e-07,
+      "loss": 0.5416,
+      "step": 23812
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.75675406215415e-07,
+      "loss": 0.6071,
+      "step": 23813
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.751700739540193e-07,
+      "loss": 0.5905,
+      "step": 23814
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.746648658822133e-07,
+      "loss": 0.5515,
+      "step": 23815
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.74159782006947e-07,
+      "loss": 0.5231,
+      "step": 23816
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.736548223351749e-07,
+      "loss": 0.6092,
+      "step": 23817
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.731499868738448e-07,
+      "loss": 0.5324,
+      "step": 23818
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.726452756298988e-07,
+      "loss": 0.5964,
+      "step": 23819
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.721406886102891e-07,
+      "loss": 0.6128,
+      "step": 23820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.716362258219559e-07,
+      "loss": 0.5215,
+      "step": 23821
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.71131887271841e-07,
+      "loss": 0.5612,
+      "step": 23822
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.70627672966883e-07,
+      "loss": 0.5428,
+      "step": 23823
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.701235829140244e-07,
+      "loss": 0.562,
+      "step": 23824
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.696196171201987e-07,
+      "loss": 0.5401,
+      "step": 23825
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.691157755923408e-07,
+      "loss": 0.5657,
+      "step": 23826
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.68612058337387e-07,
+      "loss": 0.5834,
+      "step": 23827
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.681084653622662e-07,
+      "loss": 0.6187,
+      "step": 23828
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.676049966739098e-07,
+      "loss": 0.5668,
+      "step": 23829
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.67101652279242e-07,
+      "loss": 0.6052,
+      "step": 23830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.665984321851961e-07,
+      "loss": 0.5491,
+      "step": 23831
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.660953363986925e-07,
+      "loss": 0.5437,
+      "step": 23832
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.65592364926654e-07,
+      "loss": 0.6438,
+      "step": 23833
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.650895177760023e-07,
+      "loss": 0.5862,
+      "step": 23834
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.645867949536591e-07,
+      "loss": 0.5706,
+      "step": 23835
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.640841964665403e-07,
+      "loss": 0.6098,
+      "step": 23836
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.635817223215616e-07,
+      "loss": 0.5231,
+      "step": 23837
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.630793725256403e-07,
+      "loss": 0.5463,
+      "step": 23838
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.62577147085687e-07,
+      "loss": 0.6046,
+      "step": 23839
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.620750460086126e-07,
+      "loss": 0.5564,
+      "step": 23840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.61573069301326e-07,
+      "loss": 0.577,
+      "step": 23841
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.61071216970737e-07,
+      "loss": 0.5686,
+      "step": 23842
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.605694890237516e-07,
+      "loss": 0.5422,
+      "step": 23843
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.600678854672719e-07,
+      "loss": 0.5613,
+      "step": 23844
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.595664063081989e-07,
+      "loss": 0.5849,
+      "step": 23845
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.590650515534382e-07,
+      "loss": 0.6168,
+      "step": 23846
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.58563821209887e-07,
+      "loss": 0.5823,
+      "step": 23847
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.580627152844391e-07,
+      "loss": 0.5438,
+      "step": 23848
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.57561733783996e-07,
+      "loss": 0.5551,
+      "step": 23849
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.570608767154477e-07,
+      "loss": 0.6062,
+      "step": 23850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.565601440856887e-07,
+      "loss": 0.591,
+      "step": 23851
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.560595359016057e-07,
+      "loss": 0.5428,
+      "step": 23852
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.55559052170093e-07,
+      "loss": 0.6189,
+      "step": 23853
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.550586928980343e-07,
+      "loss": 0.4973,
+      "step": 23854
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.545584580923152e-07,
+      "loss": 0.5051,
+      "step": 23855
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.540583477598187e-07,
+      "loss": 0.5739,
+      "step": 23856
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.535583619074296e-07,
+      "loss": 0.6006,
+      "step": 23857
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.530585005420256e-07,
+      "loss": 0.622,
+      "step": 23858
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.525587636704847e-07,
+      "loss": 0.6342,
+      "step": 23859
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.520591512996858e-07,
+      "loss": 0.5508,
+      "step": 23860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.515596634365043e-07,
+      "loss": 0.587,
+      "step": 23861
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.510603000878127e-07,
+      "loss": 0.5204,
+      "step": 23862
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.505610612604798e-07,
+      "loss": 0.5645,
+      "step": 23863
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.500619469613803e-07,
+      "loss": 0.5724,
+      "step": 23864
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.495629571973808e-07,
+      "loss": 0.5916,
+      "step": 23865
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.490640919753469e-07,
+      "loss": 0.4787,
+      "step": 23866
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.48565351302142e-07,
+      "loss": 0.5718,
+      "step": 23867
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.48066735184634e-07,
+      "loss": 0.5341,
+      "step": 23868
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.475682436296807e-07,
+      "loss": 0.5978,
+      "step": 23869
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.4706987664414e-07,
+      "loss": 0.604,
+      "step": 23870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.465716342348752e-07,
+      "loss": 0.594,
+      "step": 23871
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.460735164087398e-07,
+      "loss": 0.5862,
+      "step": 23872
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.455755231725883e-07,
+      "loss": 0.5465,
+      "step": 23873
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.450776545332707e-07,
+      "loss": 0.5563,
+      "step": 23874
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.445799104976438e-07,
+      "loss": 0.5176,
+      "step": 23875
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.440822910725534e-07,
+      "loss": 0.5178,
+      "step": 23876
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.435847962648493e-07,
+      "loss": 0.585,
+      "step": 23877
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.430874260813727e-07,
+      "loss": 0.6248,
+      "step": 23878
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.425901805289739e-07,
+      "loss": 0.5683,
+      "step": 23879
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.420930596144917e-07,
+      "loss": 0.5536,
+      "step": 23880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.415960633447674e-07,
+      "loss": 0.5957,
+      "step": 23881
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.410991917266421e-07,
+      "loss": 0.5212,
+      "step": 23882
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.406024447669515e-07,
+      "loss": 0.5713,
+      "step": 23883
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.401058224725313e-07,
+      "loss": 0.5937,
+      "step": 23884
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.396093248502147e-07,
+      "loss": 0.5582,
+      "step": 23885
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.391129519068364e-07,
+      "loss": 0.5378,
+      "step": 23886
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.386167036492255e-07,
+      "loss": 0.5658,
+      "step": 23887
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.381205800842086e-07,
+      "loss": 0.5466,
+      "step": 23888
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.376245812186169e-07,
+      "loss": 0.5692,
+      "step": 23889
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.371287070592738e-07,
+      "loss": 0.5626,
+      "step": 23890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.36632957613004e-07,
+      "loss": 0.5532,
+      "step": 23891
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.361373328866252e-07,
+      "loss": 0.6065,
+      "step": 23892
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.356418328869643e-07,
+      "loss": 0.5599,
+      "step": 23893
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.351464576208346e-07,
+      "loss": 0.6419,
+      "step": 23894
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.346512070950564e-07,
+      "loss": 0.622,
+      "step": 23895
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.341560813164407e-07,
+      "loss": 0.5492,
+      "step": 23896
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.336610802918044e-07,
+      "loss": 0.6292,
+      "step": 23897
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.331662040279588e-07,
+      "loss": 0.5876,
+      "step": 23898
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.326714525317104e-07,
+      "loss": 0.6151,
+      "step": 23899
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.321768258098729e-07,
+      "loss": 0.5168,
+      "step": 23900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.316823238692496e-07,
+      "loss": 0.5406,
+      "step": 23901
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.311879467166463e-07,
+      "loss": 0.6062,
+      "step": 23902
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.306936943588629e-07,
+      "loss": 0.5218,
+      "step": 23903
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.301995668027041e-07,
+      "loss": 0.5812,
+      "step": 23904
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.297055640549701e-07,
+      "loss": 0.6251,
+      "step": 23905
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.292116861224565e-07,
+      "loss": 0.5131,
+      "step": 23906
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.2871793301196e-07,
+      "loss": 0.5799,
+      "step": 23907
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.282243047302764e-07,
+      "loss": 0.577,
+      "step": 23908
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.277308012841968e-07,
+      "loss": 0.5702,
+      "step": 23909
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.272374226805115e-07,
+      "loss": 0.511,
+      "step": 23910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.267441689260137e-07,
+      "loss": 0.5901,
+      "step": 23911
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.262510400274882e-07,
+      "loss": 0.6244,
+      "step": 23912
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.257580359917218e-07,
+      "loss": 0.5551,
+      "step": 23913
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.252651568254955e-07,
+      "loss": 0.5592,
+      "step": 23914
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.247724025355964e-07,
+      "loss": 0.5701,
+      "step": 23915
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.242797731288045e-07,
+      "loss": 0.5084,
+      "step": 23916
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.237872686118965e-07,
+      "loss": 0.5381,
+      "step": 23917
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.232948889916483e-07,
+      "loss": 0.5442,
+      "step": 23918
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.228026342748408e-07,
+      "loss": 0.6193,
+      "step": 23919
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.223105044682457e-07,
+      "loss": 0.6373,
+      "step": 23920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.218184995786317e-07,
+      "loss": 0.5556,
+      "step": 23921
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.213266196127735e-07,
+      "loss": 0.533,
+      "step": 23922
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.20834864577439e-07,
+      "loss": 0.5271,
+      "step": 23923
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.203432344793917e-07,
+      "loss": 0.6354,
+      "step": 23924
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.198517293254006e-07,
+      "loss": 0.5402,
+      "step": 23925
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.193603491222292e-07,
+      "loss": 0.6288,
+      "step": 23926
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.188690938766387e-07,
+      "loss": 0.5804,
+      "step": 23927
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.183779635953872e-07,
+      "loss": 0.5416,
+      "step": 23928
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.178869582852356e-07,
+      "loss": 0.5542,
+      "step": 23929
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.173960779529401e-07,
+      "loss": 0.6701,
+      "step": 23930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.169053226052537e-07,
+      "loss": 0.5482,
+      "step": 23931
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.164146922489326e-07,
+      "loss": 0.6027,
+      "step": 23932
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.159241868907265e-07,
+      "loss": 0.4913,
+      "step": 23933
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.154338065373857e-07,
+      "loss": 0.5862,
+      "step": 23934
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.149435511956572e-07,
+      "loss": 0.5779,
+      "step": 23935
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.1445342087229e-07,
+      "loss": 0.5628,
+      "step": 23936
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.139634155740262e-07,
+      "loss": 0.5421,
+      "step": 23937
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.134735353076096e-07,
+      "loss": 0.565,
+      "step": 23938
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.129837800797792e-07,
+      "loss": 0.6214,
+      "step": 23939
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.124941498972795e-07,
+      "loss": 0.6466,
+      "step": 23940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.12004644766844e-07,
+      "loss": 0.5902,
+      "step": 23941
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.115152646952086e-07,
+      "loss": 0.6157,
+      "step": 23942
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.11026009689111e-07,
+      "loss": 0.5363,
+      "step": 23943
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.105368797552827e-07,
+      "loss": 0.5576,
+      "step": 23944
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.100478749004527e-07,
+      "loss": 0.5879,
+      "step": 23945
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.095589951313488e-07,
+      "loss": 0.5429,
+      "step": 23946
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.090702404547025e-07,
+      "loss": 0.5576,
+      "step": 23947
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.085816108772383e-07,
+      "loss": 0.5676,
+      "step": 23948
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.080931064056797e-07,
+      "loss": 0.5842,
+      "step": 23949
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.07604727046747e-07,
+      "loss": 0.5636,
+      "step": 23950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.071164728071635e-07,
+      "loss": 0.5219,
+      "step": 23951
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.066283436936474e-07,
+      "loss": 0.5767,
+      "step": 23952
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.061403397129143e-07,
+      "loss": 0.5514,
+      "step": 23953
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.056524608716821e-07,
+      "loss": 0.5487,
+      "step": 23954
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.051647071766623e-07,
+      "loss": 0.5669,
+      "step": 23955
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.046770786345682e-07,
+      "loss": 0.6702,
+      "step": 23956
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.04189575252108e-07,
+      "loss": 0.5143,
+      "step": 23957
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.037021970359927e-07,
+      "loss": 0.5533,
+      "step": 23958
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.032149439929283e-07,
+      "loss": 0.5477,
+      "step": 23959
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.027278161296194e-07,
+      "loss": 0.6039,
+      "step": 23960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.022408134527671e-07,
+      "loss": 0.6361,
+      "step": 23961
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.017539359690764e-07,
+      "loss": 0.61,
+      "step": 23962
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.012671836852471e-07,
+      "loss": 0.6543,
+      "step": 23963
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.007805566079741e-07,
+      "loss": 0.5723,
+      "step": 23964
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.002940547439576e-07,
+      "loss": 0.6373,
+      "step": 23965
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.99807678099891e-07,
+      "loss": 0.5961,
+      "step": 23966
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.993214266824657e-07,
+      "loss": 0.6244,
+      "step": 23967
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.98835300498373e-07,
+      "loss": 0.6114,
+      "step": 23968
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.983492995543063e-07,
+      "loss": 0.4956,
+      "step": 23969
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.978634238569495e-07,
+      "loss": 0.562,
+      "step": 23970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.973776734129902e-07,
+      "loss": 0.5484,
+      "step": 23971
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.968920482291099e-07,
+      "loss": 0.548,
+      "step": 23972
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.964065483119966e-07,
+      "loss": 0.6241,
+      "step": 23973
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.959211736683282e-07,
+      "loss": 0.5408,
+      "step": 23974
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.954359243047816e-07,
+      "loss": 0.5376,
+      "step": 23975
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.949508002280383e-07,
+      "loss": 0.6084,
+      "step": 23976
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.944658014447727e-07,
+      "loss": 0.5377,
+      "step": 23977
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.939809279616596e-07,
+      "loss": 0.5688,
+      "step": 23978
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.934961797853669e-07,
+      "loss": 0.5277,
+      "step": 23979
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.930115569225706e-07,
+      "loss": 0.6016,
+      "step": 23980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.925270593799373e-07,
+      "loss": 0.6018,
+      "step": 23981
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.92042687164133e-07,
+      "loss": 0.4975,
+      "step": 23982
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.915584402818256e-07,
+      "loss": 0.5197,
+      "step": 23983
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.910743187396775e-07,
+      "loss": 0.5428,
+      "step": 23984
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.905903225443502e-07,
+      "loss": 0.6327,
+      "step": 23985
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.901064517025026e-07,
+      "loss": 0.558,
+      "step": 23986
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.896227062207974e-07,
+      "loss": 0.5453,
+      "step": 23987
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.891390861058879e-07,
+      "loss": 0.5731,
+      "step": 23988
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.886555913644313e-07,
+      "loss": 0.6145,
+      "step": 23989
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.881722220030764e-07,
+      "loss": 0.6598,
+      "step": 23990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.876889780284814e-07,
+      "loss": 0.5614,
+      "step": 23991
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.872058594472921e-07,
+      "loss": 0.6056,
+      "step": 23992
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.867228662661553e-07,
+      "loss": 0.5287,
+      "step": 23993
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.862399984917214e-07,
+      "loss": 0.5895,
+      "step": 23994
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.857572561306327e-07,
+      "loss": 0.5943,
+      "step": 23995
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.852746391895339e-07,
+      "loss": 0.6344,
+      "step": 23996
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.84792147675062e-07,
+      "loss": 0.57,
+      "step": 23997
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.843097815938617e-07,
+      "loss": 0.5623,
+      "step": 23998
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.838275409525688e-07,
+      "loss": 0.5929,
+      "step": 23999
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.833454257578189e-07,
+      "loss": 0.5876,
+      "step": 24000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.828634360162447e-07,
+      "loss": 0.5983,
+      "step": 24001
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.82381571734483e-07,
+      "loss": 0.5371,
+      "step": 24002
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.818998329191619e-07,
+      "loss": 0.5195,
+      "step": 24003
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.814182195769083e-07,
+      "loss": 0.6044,
+      "step": 24004
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.809367317143558e-07,
+      "loss": 0.5498,
+      "step": 24005
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.804553693381257e-07,
+      "loss": 0.6537,
+      "step": 24006
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.799741324548427e-07,
+      "loss": 0.5195,
+      "step": 24007
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.794930210711273e-07,
+      "loss": 0.5914,
+      "step": 24008
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.790120351936038e-07,
+      "loss": 0.564,
+      "step": 24009
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.785311748288894e-07,
+      "loss": 0.4998,
+      "step": 24010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.780504399835999e-07,
+      "loss": 0.5357,
+      "step": 24011
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.775698306643509e-07,
+      "loss": 0.5037,
+      "step": 24012
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.770893468777586e-07,
+      "loss": 0.547,
+      "step": 24013
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.766089886304319e-07,
+      "loss": 0.5103,
+      "step": 24014
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.76128755928981e-07,
+      "loss": 0.578,
+      "step": 24015
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.756486487800175e-07,
+      "loss": 0.479,
+      "step": 24016
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.75168667190145e-07,
+      "loss": 0.5682,
+      "step": 24017
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.746888111659701e-07,
+      "loss": 0.5952,
+      "step": 24018
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.742090807140935e-07,
+      "loss": 0.6055,
+      "step": 24019
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.737294758411197e-07,
+      "loss": 0.6233,
+      "step": 24020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.732499965536478e-07,
+      "loss": 0.5697,
+      "step": 24021
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.727706428582761e-07,
+      "loss": 0.6111,
+      "step": 24022
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.722914147615968e-07,
+      "loss": 0.605,
+      "step": 24023
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.718123122702116e-07,
+      "loss": 0.6442,
+      "step": 24024
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.713333353907083e-07,
+      "loss": 0.5636,
+      "step": 24025
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.708544841296785e-07,
+      "loss": 0.5769,
+      "step": 24026
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.703757584937144e-07,
+      "loss": 0.5873,
+      "step": 24027
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.698971584894011e-07,
+      "loss": 0.5689,
+      "step": 24028
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.694186841233255e-07,
+      "loss": 0.5216,
+      "step": 24029
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.689403354020709e-07,
+      "loss": 0.6326,
+      "step": 24030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.684621123322213e-07,
+      "loss": 0.5784,
+      "step": 24031
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.679840149203567e-07,
+      "loss": 0.6205,
+      "step": 24032
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.675060431730564e-07,
+      "loss": 0.5774,
+      "step": 24033
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.670281970968952e-07,
+      "loss": 0.5513,
+      "step": 24034
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.665504766984533e-07,
+      "loss": 0.5601,
+      "step": 24035
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.660728819843023e-07,
+      "loss": 0.5747,
+      "step": 24036
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.655954129610122e-07,
+      "loss": 0.6464,
+      "step": 24037
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.651180696351569e-07,
+      "loss": 0.5932,
+      "step": 24038
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.646408520133043e-07,
+      "loss": 0.4925,
+      "step": 24039
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.641637601020203e-07,
+      "loss": 0.5491,
+      "step": 24040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.636867939078686e-07,
+      "loss": 0.559,
+      "step": 24041
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.632099534374161e-07,
+      "loss": 0.5904,
+      "step": 24042
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.627332386972231e-07,
+      "loss": 0.5379,
+      "step": 24043
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.622566496938467e-07,
+      "loss": 0.5688,
+      "step": 24044
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.617801864338515e-07,
+      "loss": 0.6063,
+      "step": 24045
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.613038489237891e-07,
+      "loss": 0.6041,
+      "step": 24046
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.608276371702162e-07,
+      "loss": 0.553,
+      "step": 24047
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.603515511796834e-07,
+      "loss": 0.582,
+      "step": 24048
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.598755909587464e-07,
+      "loss": 0.6723,
+      "step": 24049
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.593997565139522e-07,
+      "loss": 0.6225,
+      "step": 24050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.589240478518479e-07,
+      "loss": 0.5165,
+      "step": 24051
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.584484649789804e-07,
+      "loss": 0.5562,
+      "step": 24052
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.579730079018955e-07,
+      "loss": 0.5943,
+      "step": 24053
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.574976766271359e-07,
+      "loss": 0.5295,
+      "step": 24054
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.570224711612385e-07,
+      "loss": 0.5329,
+      "step": 24055
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.565473915107481e-07,
+      "loss": 0.5631,
+      "step": 24056
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.560724376821994e-07,
+      "loss": 0.5766,
+      "step": 24057
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.555976096821294e-07,
+      "loss": 0.5406,
+      "step": 24058
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.551229075170686e-07,
+      "loss": 0.5284,
+      "step": 24059
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.546483311935538e-07,
+      "loss": 0.5206,
+      "step": 24060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.541738807181143e-07,
+      "loss": 0.5361,
+      "step": 24061
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.536995560972783e-07,
+      "loss": 0.5839,
+      "step": 24062
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.532253573375704e-07,
+      "loss": 0.5543,
+      "step": 24063
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.52751284445521e-07,
+      "loss": 0.6218,
+      "step": 24064
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.522773374276516e-07,
+      "loss": 0.5743,
+      "step": 24065
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.518035162904825e-07,
+      "loss": 0.5212,
+      "step": 24066
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.513298210405363e-07,
+      "loss": 0.5446,
+      "step": 24067
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.50856251684331e-07,
+      "loss": 0.5313,
+      "step": 24068
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.503828082283827e-07,
+      "loss": 0.5595,
+      "step": 24069
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.499094906792049e-07,
+      "loss": 0.5422,
+      "step": 24070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.494362990433136e-07,
+      "loss": 0.5824,
+      "step": 24071
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.489632333272202e-07,
+      "loss": 0.5846,
+      "step": 24072
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.484902935374328e-07,
+      "loss": 0.6068,
+      "step": 24073
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.480174796804586e-07,
+      "loss": 0.4955,
+      "step": 24074
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.475447917628065e-07,
+      "loss": 0.5329,
+      "step": 24075
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.470722297909817e-07,
+      "loss": 0.544,
+      "step": 24076
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.46599793771482e-07,
+      "loss": 0.5871,
+      "step": 24077
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.461274837108158e-07,
+      "loss": 0.5538,
+      "step": 24078
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.456552996154765e-07,
+      "loss": 0.5619,
+      "step": 24079
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.451832414919658e-07,
+      "loss": 0.6244,
+      "step": 24080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.447113093467752e-07,
+      "loss": 0.5892,
+      "step": 24081
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.442395031864037e-07,
+      "loss": 0.5659,
+      "step": 24082
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.437678230173407e-07,
+      "loss": 0.6084,
+      "step": 24083
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.432962688460788e-07,
+      "loss": 0.5572,
+      "step": 24084
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.42824840679104e-07,
+      "loss": 0.5897,
+      "step": 24085
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.423535385229076e-07,
+      "loss": 0.5995,
+      "step": 24086
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.418823623839722e-07,
+      "loss": 0.6054,
+      "step": 24087
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.414113122687817e-07,
+      "loss": 0.5868,
+      "step": 24088
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.409403881838207e-07,
+      "loss": 0.5853,
+      "step": 24089
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.404695901355675e-07,
+      "loss": 0.6102,
+      "step": 24090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.399989181305013e-07,
+      "loss": 0.5476,
+      "step": 24091
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.395283721750958e-07,
+      "loss": 0.5623,
+      "step": 24092
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.390579522758324e-07,
+      "loss": 0.5453,
+      "step": 24093
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.385876584391794e-07,
+      "loss": 0.6371,
+      "step": 24094
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.381174906716117e-07,
+      "loss": 0.5777,
+      "step": 24095
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.37647448979595e-07,
+      "loss": 0.5537,
+      "step": 24096
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.371775333696019e-07,
+      "loss": 0.6105,
+      "step": 24097
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.367077438480975e-07,
+      "loss": 0.5531,
+      "step": 24098
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.362380804215442e-07,
+      "loss": 0.5678,
+      "step": 24099
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.357685430964091e-07,
+      "loss": 0.6366,
+      "step": 24100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.352991318791503e-07,
+      "loss": 0.5665,
+      "step": 24101
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.348298467762284e-07,
+      "loss": 0.5047,
+      "step": 24102
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.343606877941002e-07,
+      "loss": 0.5726,
+      "step": 24103
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.33891654939224e-07,
+      "loss": 0.5938,
+      "step": 24104
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.334227482180524e-07,
+      "loss": 0.6099,
+      "step": 24105
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.329539676370357e-07,
+      "loss": 0.531,
+      "step": 24106
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.3248531320263e-07,
+      "loss": 0.5605,
+      "step": 24107
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.320167849212823e-07,
+      "loss": 0.5798,
+      "step": 24108
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.315483827994387e-07,
+      "loss": 0.5873,
+      "step": 24109
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.310801068435426e-07,
+      "loss": 0.6216,
+      "step": 24110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.306119570600435e-07,
+      "loss": 0.522,
+      "step": 24111
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.301439334553807e-07,
+      "loss": 0.5679,
+      "step": 24112
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.296760360359957e-07,
+      "loss": 0.5519,
+      "step": 24113
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.292082648083233e-07,
+      "loss": 0.5038,
+      "step": 24114
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.28740619778805e-07,
+      "loss": 0.5107,
+      "step": 24115
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.282731009538747e-07,
+      "loss": 0.632,
+      "step": 24116
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.278057083399638e-07,
+      "loss": 0.5662,
+      "step": 24117
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.273384419435071e-07,
+      "loss": 0.5631,
+      "step": 24118
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.26871301770934e-07,
+      "loss": 0.6048,
+      "step": 24119
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.264042878286715e-07,
+      "loss": 0.5699,
+      "step": 24120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.259374001231447e-07,
+      "loss": 0.5019,
+      "step": 24121
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.254706386607825e-07,
+      "loss": 0.5649,
+      "step": 24122
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.250040034480055e-07,
+      "loss": 0.5921,
+      "step": 24123
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.245374944912354e-07,
+      "loss": 0.6233,
+      "step": 24124
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.240711117968903e-07,
+      "loss": 0.5402,
+      "step": 24125
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.236048553713905e-07,
+      "loss": 0.5859,
+      "step": 24126
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.231387252211509e-07,
+      "loss": 0.5312,
+      "step": 24127
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.226727213525843e-07,
+      "loss": 0.5425,
+      "step": 24128
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.222068437721076e-07,
+      "loss": 0.5508,
+      "step": 24129
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.21741092486128e-07,
+      "loss": 0.5334,
+      "step": 24130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.212754675010559e-07,
+      "loss": 0.6054,
+      "step": 24131
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.208099688232984e-07,
+      "loss": 0.564,
+      "step": 24132
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.203445964592616e-07,
+      "loss": 0.5472,
+      "step": 24133
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.198793504153491e-07,
+      "loss": 0.5789,
+      "step": 24134
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.194142306979613e-07,
+      "loss": 0.5465,
+      "step": 24135
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.189492373135021e-07,
+      "loss": 0.5353,
+      "step": 24136
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.184843702683687e-07,
+      "loss": 0.5868,
+      "step": 24137
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.18019629568958e-07,
+      "loss": 0.5896,
+      "step": 24138
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.175550152216627e-07,
+      "loss": 0.5551,
+      "step": 24139
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.17090527232881e-07,
+      "loss": 0.5524,
+      "step": 24140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.166261656090035e-07,
+      "loss": 0.6058,
+      "step": 24141
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.161619303564161e-07,
+      "loss": 0.5799,
+      "step": 24142
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.156978214815125e-07,
+      "loss": 0.5372,
+      "step": 24143
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.152338389906767e-07,
+      "loss": 0.5666,
+      "step": 24144
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.147699828902933e-07,
+      "loss": 0.6547,
+      "step": 24145
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.143062531867451e-07,
+      "loss": 0.5884,
+      "step": 24146
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.13842649886415e-07,
+      "loss": 0.5069,
+      "step": 24147
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.133791729956819e-07,
+      "loss": 0.6725,
+      "step": 24148
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.129158225209222e-07,
+      "loss": 0.5861,
+      "step": 24149
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.124525984685139e-07,
+      "loss": 0.5789,
+      "step": 24150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.11989500844832e-07,
+      "loss": 0.515,
+      "step": 24151
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.115265296562469e-07,
+      "loss": 0.5307,
+      "step": 24152
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.110636849091291e-07,
+      "loss": 0.5191,
+      "step": 24153
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.106009666098514e-07,
+      "loss": 0.5423,
+      "step": 24154
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.101383747647784e-07,
+      "loss": 0.6054,
+      "step": 24155
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.096759093802752e-07,
+      "loss": 0.5572,
+      "step": 24156
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.092135704627069e-07,
+      "loss": 0.5129,
+      "step": 24157
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.087513580184358e-07,
+      "loss": 0.6187,
+      "step": 24158
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.082892720538238e-07,
+      "loss": 0.6272,
+      "step": 24159
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.078273125752245e-07,
+      "loss": 0.6361,
+      "step": 24160
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.073654795890007e-07,
+      "loss": 0.5318,
+      "step": 24161
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.069037731015062e-07,
+      "loss": 0.6162,
+      "step": 24162
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.064421931190925e-07,
+      "loss": 0.5825,
+      "step": 24163
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.059807396481112e-07,
+      "loss": 0.5276,
+      "step": 24164
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.055194126949151e-07,
+      "loss": 0.5606,
+      "step": 24165
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.050582122658501e-07,
+      "loss": 0.5898,
+      "step": 24166
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.045971383672646e-07,
+      "loss": 0.5921,
+      "step": 24167
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.041361910055002e-07,
+      "loss": 0.6076,
+      "step": 24168
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.036753701869026e-07,
+      "loss": 0.516,
+      "step": 24169
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.032146759178138e-07,
+      "loss": 0.5812,
+      "step": 24170
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.027541082045698e-07,
+      "loss": 0.5741,
+      "step": 24171
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.02293667053513e-07,
+      "loss": 0.6003,
+      "step": 24172
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.018333524709765e-07,
+      "loss": 0.5821,
+      "step": 24173
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.01373164463296e-07,
+      "loss": 0.5838,
+      "step": 24174
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.009131030368012e-07,
+      "loss": 0.4977,
+      "step": 24175
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.004531681978267e-07,
+      "loss": 0.5157,
+      "step": 24176
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.999933599526999e-07,
+      "loss": 0.5801,
+      "step": 24177
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.99533678307749e-07,
+      "loss": 0.5775,
+      "step": 24178
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.990741232692967e-07,
+      "loss": 0.5729,
+      "step": 24179
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.986146948436713e-07,
+      "loss": 0.622,
+      "step": 24180
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.981553930371922e-07,
+      "loss": 0.5053,
+      "step": 24181
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.976962178561787e-07,
+      "loss": 0.5426,
+      "step": 24182
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.972371693069525e-07,
+      "loss": 0.539,
+      "step": 24183
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.967782473958296e-07,
+      "loss": 0.5759,
+      "step": 24184
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.963194521291251e-07,
+      "loss": 0.5553,
+      "step": 24185
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.958607835131493e-07,
+      "loss": 0.6093,
+      "step": 24186
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.954022415542195e-07,
+      "loss": 0.5406,
+      "step": 24187
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.949438262586418e-07,
+      "loss": 0.5091,
+      "step": 24188
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.944855376327254e-07,
+      "loss": 0.5483,
+      "step": 24189
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.940273756827754e-07,
+      "loss": 0.5512,
+      "step": 24190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.93569340415099e-07,
+      "loss": 0.5351,
+      "step": 24191
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.931114318359978e-07,
+      "loss": 0.5394,
+      "step": 24192
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.926536499517723e-07,
+      "loss": 0.6116,
+      "step": 24193
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.921959947687241e-07,
+      "loss": 0.546,
+      "step": 24194
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.917384662931493e-07,
+      "loss": 0.6243,
+      "step": 24195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.912810645313451e-07,
+      "loss": 0.5957,
+      "step": 24196
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.90823789489602e-07,
+      "loss": 0.577,
+      "step": 24197
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.903666411742184e-07,
+      "loss": 0.6118,
+      "step": 24198
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.899096195914813e-07,
+      "loss": 0.5638,
+      "step": 24199
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.894527247476813e-07,
+      "loss": 0.5675,
+      "step": 24200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.889959566491023e-07,
+      "loss": 0.6506,
+      "step": 24201
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.885393153020348e-07,
+      "loss": 0.5571,
+      "step": 24202
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.880828007127605e-07,
+      "loss": 0.5169,
+      "step": 24203
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.876264128875588e-07,
+      "loss": 0.6289,
+      "step": 24204
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.871701518327146e-07,
+      "loss": 0.5008,
+      "step": 24205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.86714017554503e-07,
+      "loss": 0.598,
+      "step": 24206
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.862580100592032e-07,
+      "loss": 0.6361,
+      "step": 24207
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.858021293530871e-07,
+      "loss": 0.5781,
+      "step": 24208
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.853463754424318e-07,
+      "loss": 0.5406,
+      "step": 24209
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.848907483335066e-07,
+      "loss": 0.6086,
+      "step": 24210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.844352480325813e-07,
+      "loss": 0.517,
+      "step": 24211
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.839798745459249e-07,
+      "loss": 0.5716,
+      "step": 24212
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.835246278798037e-07,
+      "loss": 0.5629,
+      "step": 24213
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.830695080404815e-07,
+      "loss": 0.5492,
+      "step": 24214
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.826145150342213e-07,
+      "loss": 0.6197,
+      "step": 24215
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.821596488672845e-07,
+      "loss": 0.545,
+      "step": 24216
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.817049095459317e-07,
+      "loss": 0.5895,
+      "step": 24217
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.81250297076419e-07,
+      "loss": 0.6169,
+      "step": 24218
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.807958114650005e-07,
+      "loss": 0.556,
+      "step": 24219
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.803414527179343e-07,
+      "loss": 0.5654,
+      "step": 24220
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.798872208414698e-07,
+      "loss": 0.5713,
+      "step": 24221
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.794331158418577e-07,
+      "loss": 0.4944,
+      "step": 24222
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.789791377253497e-07,
+      "loss": 0.632,
+      "step": 24223
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.785252864981907e-07,
+      "loss": 0.6235,
+      "step": 24224
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.780715621666268e-07,
+      "loss": 0.5828,
+      "step": 24225
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.776179647368997e-07,
+      "loss": 0.5399,
+      "step": 24226
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.771644942152546e-07,
+      "loss": 0.5344,
+      "step": 24227
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.767111506079294e-07,
+      "loss": 0.5616,
+      "step": 24228
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.76257933921164e-07,
+      "loss": 0.5903,
+      "step": 24229
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.758048441611921e-07,
+      "loss": 0.5682,
+      "step": 24230
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.75351881334252e-07,
+      "loss": 0.5256,
+      "step": 24231
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.748990454465755e-07,
+      "loss": 0.5722,
+      "step": 24232
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.744463365043919e-07,
+      "loss": 0.5521,
+      "step": 24233
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.739937545139354e-07,
+      "loss": 0.6561,
+      "step": 24234
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.735412994814318e-07,
+      "loss": 0.5939,
+      "step": 24235
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.730889714131062e-07,
+      "loss": 0.5569,
+      "step": 24236
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.726367703151816e-07,
+      "loss": 0.6018,
+      "step": 24237
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.721846961938861e-07,
+      "loss": 0.5225,
+      "step": 24238
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.717327490554372e-07,
+      "loss": 0.6289,
+      "step": 24239
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.71280928906053e-07,
+      "loss": 0.5697,
+      "step": 24240
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.70829235751952e-07,
+      "loss": 0.5395,
+      "step": 24241
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.703776695993515e-07,
+      "loss": 0.5657,
+      "step": 24242
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.699262304544641e-07,
+      "loss": 0.5804,
+      "step": 24243
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.694749183234995e-07,
+      "loss": 0.5639,
+      "step": 24244
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.690237332126737e-07,
+      "loss": 0.5551,
+      "step": 24245
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.685726751281919e-07,
+      "loss": 0.5926,
+      "step": 24246
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.681217440762611e-07,
+      "loss": 0.5639,
+      "step": 24247
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.676709400630855e-07,
+      "loss": 0.5976,
+      "step": 24248
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.672202630948722e-07,
+      "loss": 0.4991,
+      "step": 24249
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.667697131778196e-07,
+      "loss": 0.5882,
+      "step": 24250
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.663192903181293e-07,
+      "loss": 0.543,
+      "step": 24251
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.658689945219977e-07,
+      "loss": 0.5496,
+      "step": 24252
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.654188257956241e-07,
+      "loss": 0.5633,
+      "step": 24253
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.649687841452014e-07,
+      "loss": 0.6042,
+      "step": 24254
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.645188695769212e-07,
+      "loss": 0.6162,
+      "step": 24255
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.640690820969776e-07,
+      "loss": 0.526,
+      "step": 24256
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.6361942171156e-07,
+      "loss": 0.5824,
+      "step": 24257
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.631698884268546e-07,
+      "loss": 0.5309,
+      "step": 24258
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.627204822490453e-07,
+      "loss": 0.5495,
+      "step": 24259
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.622712031843204e-07,
+      "loss": 0.5499,
+      "step": 24260
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.618220512388619e-07,
+      "loss": 0.6243,
+      "step": 24261
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.61373026418849e-07,
+      "loss": 0.5225,
+      "step": 24262
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.609241287304581e-07,
+      "loss": 0.5585,
+      "step": 24263
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.604753581798718e-07,
+      "loss": 0.5276,
+      "step": 24264
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.600267147732643e-07,
+      "loss": 0.5545,
+      "step": 24265
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.595781985168049e-07,
+      "loss": 0.56,
+      "step": 24266
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.591298094166711e-07,
+      "loss": 0.5532,
+      "step": 24267
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.58681547479031e-07,
+      "loss": 0.6381,
+      "step": 24268
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.582334127100521e-07,
+      "loss": 0.5765,
+      "step": 24269
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.577854051159017e-07,
+      "loss": 0.5438,
+      "step": 24270
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.573375247027459e-07,
+      "loss": 0.5466,
+      "step": 24271
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.568897714767464e-07,
+      "loss": 0.5186,
+      "step": 24272
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.564421454440651e-07,
+      "loss": 0.595,
+      "step": 24273
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.559946466108625e-07,
+      "loss": 0.5323,
+      "step": 24274
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.55547274983297e-07,
+      "loss": 0.5632,
+      "step": 24275
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.551000305675238e-07,
+      "loss": 0.6301,
+      "step": 24276
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.546529133696944e-07,
+      "loss": 0.5386,
+      "step": 24277
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.542059233959676e-07,
+      "loss": 0.5904,
+      "step": 24278
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.537590606524914e-07,
+      "loss": 0.6112,
+      "step": 24279
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.533123251454145e-07,
+      "loss": 0.5426,
+      "step": 24280
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.528657168808829e-07,
+      "loss": 0.6118,
+      "step": 24281
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.524192358650461e-07,
+      "loss": 0.5725,
+      "step": 24282
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.519728821040473e-07,
+      "loss": 0.596,
+      "step": 24283
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.515266556040246e-07,
+      "loss": 0.51,
+      "step": 24284
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.510805563711243e-07,
+      "loss": 0.5773,
+      "step": 24285
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.506345844114816e-07,
+      "loss": 0.57,
+      "step": 24286
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.501887397312346e-07,
+      "loss": 0.5461,
+      "step": 24287
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.497430223365165e-07,
+      "loss": 0.5322,
+      "step": 24288
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.492974322334634e-07,
+      "loss": 0.5119,
+      "step": 24289
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.48851969428207e-07,
+      "loss": 0.5763,
+      "step": 24290
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.484066339268759e-07,
+      "loss": 0.5199,
+      "step": 24291
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.479614257355972e-07,
+      "loss": 0.5996,
+      "step": 24292
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.475163448604994e-07,
+      "loss": 0.5287,
+      "step": 24293
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.470713913077077e-07,
+      "loss": 0.6004,
+      "step": 24294
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.466265650833416e-07,
+      "loss": 0.5742,
+      "step": 24295
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.461818661935271e-07,
+      "loss": 0.5854,
+      "step": 24296
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.457372946443808e-07,
+      "loss": 0.5424,
+      "step": 24297
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.45292850442021e-07,
+      "loss": 0.552,
+      "step": 24298
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.448485335925626e-07,
+      "loss": 0.532,
+      "step": 24299
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.444043441021209e-07,
+      "loss": 0.5052,
+      "step": 24300
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.439602819768088e-07,
+      "loss": 0.5505,
+      "step": 24301
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.435163472227369e-07,
+      "loss": 0.5755,
+      "step": 24302
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.430725398460103e-07,
+      "loss": 0.5786,
+      "step": 24303
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.42628859852742e-07,
+      "loss": 0.6127,
+      "step": 24304
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.421853072490348e-07,
+      "loss": 0.6377,
+      "step": 24305
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.417418820409917e-07,
+      "loss": 0.5443,
+      "step": 24306
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.412985842347153e-07,
+      "loss": 0.6316,
+      "step": 24307
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.408554138363067e-07,
+      "loss": 0.5838,
+      "step": 24308
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.404123708518629e-07,
+      "loss": 0.5664,
+      "step": 24309
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.399694552874803e-07,
+      "loss": 0.5877,
+      "step": 24310
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.395266671492552e-07,
+      "loss": 0.6075,
+      "step": 24311
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.390840064432802e-07,
+      "loss": 0.6482,
+      "step": 24312
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.386414731756475e-07,
+      "loss": 0.6437,
+      "step": 24313
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.381990673524431e-07,
+      "loss": 0.582,
+      "step": 24314
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.377567889797598e-07,
+      "loss": 0.5238,
+      "step": 24315
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.37314638063682e-07,
+      "loss": 0.5319,
+      "step": 24316
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.368726146102923e-07,
+      "loss": 0.5641,
+      "step": 24317
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.364307186256758e-07,
+      "loss": 0.5287,
+      "step": 24318
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.359889501159123e-07,
+      "loss": 0.5274,
+      "step": 24319
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.355473090870813e-07,
+      "loss": 0.5622,
+      "step": 24320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.351057955452567e-07,
+      "loss": 0.5831,
+      "step": 24321
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.346644094965205e-07,
+      "loss": 0.6034,
+      "step": 24322
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.342231509469422e-07,
+      "loss": 0.5779,
+      "step": 24323
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.337820199025958e-07,
+      "loss": 0.5489,
+      "step": 24324
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.333410163695487e-07,
+      "loss": 0.6162,
+      "step": 24325
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.329001403538738e-07,
+      "loss": 0.5724,
+      "step": 24326
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.324593918616352e-07,
+      "loss": 0.4753,
+      "step": 24327
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.320187708988968e-07,
+      "loss": 0.5645,
+      "step": 24328
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.315782774717262e-07,
+      "loss": 0.6128,
+      "step": 24329
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.311379115861816e-07,
+      "loss": 0.5394,
+      "step": 24330
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.30697673248324e-07,
+      "loss": 0.5686,
+      "step": 24331
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.302575624642094e-07,
+      "loss": 0.5787,
+      "step": 24332
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.298175792398976e-07,
+      "loss": 0.5408,
+      "step": 24333
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.293777235814415e-07,
+      "loss": 0.5755,
+      "step": 24334
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.289379954948916e-07,
+      "loss": 0.5252,
+      "step": 24335
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.284983949863034e-07,
+      "loss": 0.5609,
+      "step": 24336
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.280589220617229e-07,
+      "loss": 0.5299,
+      "step": 24337
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.276195767271998e-07,
+      "loss": 0.5438,
+      "step": 24338
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.271803589887771e-07,
+      "loss": 0.5673,
+      "step": 24339
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.26741268852501e-07,
+      "loss": 0.5642,
+      "step": 24340
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.263023063244146e-07,
+      "loss": 0.5396,
+      "step": 24341
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.258634714105561e-07,
+      "loss": 0.5308,
+      "step": 24342
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.254247641169654e-07,
+      "loss": 0.511,
+      "step": 24343
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.249861844496797e-07,
+      "loss": 0.5783,
+      "step": 24344
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.245477324147343e-07,
+      "loss": 0.606,
+      "step": 24345
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.241094080181599e-07,
+      "loss": 0.5757,
+      "step": 24346
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.236712112659916e-07,
+      "loss": 0.5656,
+      "step": 24347
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.232331421642591e-07,
+      "loss": 0.6022,
+      "step": 24348
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.227952007189876e-07,
+      "loss": 0.6348,
+      "step": 24349
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.223573869362077e-07,
+      "loss": 0.5911,
+      "step": 24350
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.219197008219425e-07,
+      "loss": 0.5907,
+      "step": 24351
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.21482142382215e-07,
+      "loss": 0.5545,
+      "step": 24352
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.210447116230435e-07,
+      "loss": 0.5565,
+      "step": 24353
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.206074085504522e-07,
+      "loss": 0.5274,
+      "step": 24354
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.201702331704563e-07,
+      "loss": 0.5312,
+      "step": 24355
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.197331854890721e-07,
+      "loss": 0.5853,
+      "step": 24356
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.192962655123126e-07,
+      "loss": 0.5697,
+      "step": 24357
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.188594732461918e-07,
+      "loss": 0.552,
+      "step": 24358
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.184228086967204e-07,
+      "loss": 0.5634,
+      "step": 24359
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.179862718699049e-07,
+      "loss": 0.5978,
+      "step": 24360
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.175498627717559e-07,
+      "loss": 0.6402,
+      "step": 24361
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.171135814082764e-07,
+      "loss": 0.5543,
+      "step": 24362
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.166774277854716e-07,
+      "loss": 0.5622,
+      "step": 24363
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.1624140190934e-07,
+      "loss": 0.5734,
+      "step": 24364
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.158055037858858e-07,
+      "loss": 0.592,
+      "step": 24365
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.153697334211063e-07,
+      "loss": 0.5894,
+      "step": 24366
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.149340908209956e-07,
+      "loss": 0.5791,
+      "step": 24367
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.144985759915524e-07,
+      "loss": 0.5391,
+      "step": 24368
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.140631889387672e-07,
+      "loss": 0.5686,
+      "step": 24369
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.136279296686322e-07,
+      "loss": 0.5241,
+      "step": 24370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.131927981871345e-07,
+      "loss": 0.5506,
+      "step": 24371
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.127577945002662e-07,
+      "loss": 0.6,
+      "step": 24372
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.123229186140101e-07,
+      "loss": 0.5686,
+      "step": 24373
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.118881705343527e-07,
+      "loss": 0.5849,
+      "step": 24374
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.114535502672726e-07,
+      "loss": 0.6297,
+      "step": 24375
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.11019057818756e-07,
+      "loss": 0.5507,
+      "step": 24376
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.105846931947791e-07,
+      "loss": 0.5658,
+      "step": 24377
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.101504564013162e-07,
+      "loss": 0.5721,
+      "step": 24378
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.097163474443491e-07,
+      "loss": 0.5258,
+      "step": 24379
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.092823663298476e-07,
+      "loss": 0.5337,
+      "step": 24380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.088485130637835e-07,
+      "loss": 0.5823,
+      "step": 24381
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.084147876521274e-07,
+      "loss": 0.5144,
+      "step": 24382
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.079811901008493e-07,
+      "loss": 0.56,
+      "step": 24383
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.075477204159142e-07,
+      "loss": 0.576,
+      "step": 24384
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.071143786032886e-07,
+      "loss": 0.5606,
+      "step": 24385
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.06681164668932e-07,
+      "loss": 0.5562,
+      "step": 24386
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.062480786188097e-07,
+      "loss": 0.5922,
+      "step": 24387
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.058151204588815e-07,
+      "loss": 0.5934,
+      "step": 24388
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.053822901951013e-07,
+      "loss": 0.5862,
+      "step": 24389
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.04949587833429e-07,
+      "loss": 0.5571,
+      "step": 24390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.045170133798184e-07,
+      "loss": 0.5452,
+      "step": 24391
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.040845668402218e-07,
+      "loss": 0.5599,
+      "step": 24392
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.036522482205866e-07,
+      "loss": 0.5907,
+      "step": 24393
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.032200575268677e-07,
+      "loss": 0.5284,
+      "step": 24394
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.027879947650096e-07,
+      "loss": 0.6178,
+      "step": 24395
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.023560599409574e-07,
+      "loss": 0.5791,
+      "step": 24396
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.01924253060654e-07,
+      "loss": 0.5534,
+      "step": 24397
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.014925741300449e-07,
+      "loss": 0.5827,
+      "step": 24398
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.010610231550686e-07,
+      "loss": 0.5593,
+      "step": 24399
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.006296001416613e-07,
+      "loss": 0.5424,
+      "step": 24400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.001983050957628e-07,
+      "loss": 0.5479,
+      "step": 24401
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.997671380233084e-07,
+      "loss": 0.5703,
+      "step": 24402
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.993360989302289e-07,
+      "loss": 0.5941,
+      "step": 24403
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.989051878224562e-07,
+      "loss": 0.5158,
+      "step": 24404
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.984744047059222e-07,
+      "loss": 0.5725,
+      "step": 24405
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.980437495865533e-07,
+      "loss": 0.5969,
+      "step": 24406
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.976132224702759e-07,
+      "loss": 0.5346,
+      "step": 24407
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.971828233630129e-07,
+      "loss": 0.5542,
+      "step": 24408
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.967525522706897e-07,
+      "loss": 0.5641,
+      "step": 24409
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.96322409199226e-07,
+      "loss": 0.5112,
+      "step": 24410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.958923941545393e-07,
+      "loss": 0.5423,
+      "step": 24411
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.954625071425502e-07,
+      "loss": 0.6385,
+      "step": 24412
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.950327481691732e-07,
+      "loss": 0.5608,
+      "step": 24413
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.946031172403211e-07,
+      "loss": 0.5488,
+      "step": 24414
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.941736143619048e-07,
+      "loss": 0.5623,
+      "step": 24415
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.937442395398375e-07,
+      "loss": 0.5751,
+      "step": 24416
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.933149927800275e-07,
+      "loss": 0.6212,
+      "step": 24417
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.928858740883803e-07,
+      "loss": 0.6459,
+      "step": 24418
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.924568834708001e-07,
+      "loss": 0.5731,
+      "step": 24419
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.920280209331931e-07,
+      "loss": 0.5654,
+      "step": 24420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.915992864814602e-07,
+      "loss": 0.5618,
+      "step": 24421
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.911706801214968e-07,
+      "loss": 0.5727,
+      "step": 24422
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.907422018592069e-07,
+      "loss": 0.5492,
+      "step": 24423
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.903138517004837e-07,
+      "loss": 0.6075,
+      "step": 24424
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.898856296512225e-07,
+      "loss": 0.5644,
+      "step": 24425
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.894575357173139e-07,
+      "loss": 0.6017,
+      "step": 24426
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.890295699046524e-07,
+      "loss": 0.5492,
+      "step": 24427
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.886017322191252e-07,
+      "loss": 0.563,
+      "step": 24428
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.881740226666178e-07,
+      "loss": 0.5385,
+      "step": 24429
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.8774644125302e-07,
+      "loss": 0.6308,
+      "step": 24430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.873189879842134e-07,
+      "loss": 0.5477,
+      "step": 24431
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.868916628660815e-07,
+      "loss": 0.6186,
+      "step": 24432
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.864644659045006e-07,
+      "loss": 0.5732,
+      "step": 24433
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.860373971053535e-07,
+      "loss": 0.6348,
+      "step": 24434
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.856104564745169e-07,
+      "loss": 0.5705,
+      "step": 24435
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.85183644017865e-07,
+      "loss": 0.5709,
+      "step": 24436
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.847569597412684e-07,
+      "loss": 0.5578,
+      "step": 24437
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.843304036506026e-07,
+      "loss": 0.5755,
+      "step": 24438
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.839039757517363e-07,
+      "loss": 0.5131,
+      "step": 24439
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.834776760505357e-07,
+      "loss": 0.5929,
+      "step": 24440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.830515045528696e-07,
+      "loss": 0.6056,
+      "step": 24441
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.826254612646011e-07,
+      "loss": 0.5545,
+      "step": 24442
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.821995461915942e-07,
+      "loss": 0.5667,
+      "step": 24443
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.817737593397067e-07,
+      "loss": 0.5518,
+      "step": 24444
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.813481007148015e-07,
+      "loss": 0.5818,
+      "step": 24445
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.809225703227352e-07,
+      "loss": 0.5899,
+      "step": 24446
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.804971681693618e-07,
+      "loss": 0.5728,
+      "step": 24447
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.800718942605355e-07,
+      "loss": 0.6265,
+      "step": 24448
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.796467486021108e-07,
+      "loss": 0.5677,
+      "step": 24449
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.79221731199936e-07,
+      "loss": 0.5937,
+      "step": 24450
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.787968420598579e-07,
+      "loss": 0.5123,
+      "step": 24451
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.783720811877281e-07,
+      "loss": 0.5483,
+      "step": 24452
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.779474485893889e-07,
+      "loss": 0.5817,
+      "step": 24453
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.775229442706843e-07,
+      "loss": 0.5541,
+      "step": 24454
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.770985682374531e-07,
+      "loss": 0.5704,
+      "step": 24455
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.766743204955395e-07,
+      "loss": 0.5653,
+      "step": 24456
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.762502010507788e-07,
+      "loss": 0.5744,
+      "step": 24457
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.758262099090085e-07,
+      "loss": 0.6076,
+      "step": 24458
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.754023470760607e-07,
+      "loss": 0.5594,
+      "step": 24459
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.749786125577717e-07,
+      "loss": 0.5462,
+      "step": 24460
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.745550063599704e-07,
+      "loss": 0.5626,
+      "step": 24461
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.741315284884842e-07,
+      "loss": 0.5652,
+      "step": 24462
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.737081789491451e-07,
+      "loss": 0.6145,
+      "step": 24463
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.732849577477763e-07,
+      "loss": 0.61,
+      "step": 24464
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.728618648902008e-07,
+      "loss": 0.5525,
+      "step": 24465
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.724389003822406e-07,
+      "loss": 0.5659,
+      "step": 24466
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.72016064229718e-07,
+      "loss": 0.6105,
+      "step": 24467
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.715933564384514e-07,
+      "loss": 0.5898,
+      "step": 24468
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.711707770142573e-07,
+      "loss": 0.5818,
+      "step": 24469
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.707483259629477e-07,
+      "loss": 0.5468,
+      "step": 24470
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.703260032903403e-07,
+      "loss": 0.5354,
+      "step": 24471
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.699038090022447e-07,
+      "loss": 0.5336,
+      "step": 24472
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.694817431044698e-07,
+      "loss": 0.584,
+      "step": 24473
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.690598056028263e-07,
+      "loss": 0.5708,
+      "step": 24474
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.686379965031186e-07,
+      "loss": 0.5822,
+      "step": 24475
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.68216315811151e-07,
+      "loss": 0.5437,
+      "step": 24476
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.677947635327253e-07,
+      "loss": 0.5537,
+      "step": 24477
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.673733396736448e-07,
+      "loss": 0.5841,
+      "step": 24478
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.669520442397081e-07,
+      "loss": 0.5886,
+      "step": 24479
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.665308772367118e-07,
+      "loss": 0.5933,
+      "step": 24480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.66109838670449e-07,
+      "loss": 0.528,
+      "step": 24481
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.656889285467194e-07,
+      "loss": 0.536,
+      "step": 24482
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.652681468713118e-07,
+      "loss": 0.5717,
+      "step": 24483
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.648474936500149e-07,
+      "loss": 0.5475,
+      "step": 24484
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.644269688886196e-07,
+      "loss": 0.544,
+      "step": 24485
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.640065725929135e-07,
+      "loss": 0.6153,
+      "step": 24486
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.635863047686797e-07,
+      "loss": 0.5498,
+      "step": 24487
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.631661654217014e-07,
+      "loss": 0.6481,
+      "step": 24488
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.627461545577607e-07,
+      "loss": 0.5646,
+      "step": 24489
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.623262721826384e-07,
+      "loss": 0.5835,
+      "step": 24490
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.619065183021101e-07,
+      "loss": 0.5992,
+      "step": 24491
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.614868929219542e-07,
+      "loss": 0.4814,
+      "step": 24492
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.610673960479441e-07,
+      "loss": 0.5912,
+      "step": 24493
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.606480276858529e-07,
+      "loss": 0.536,
+      "step": 24494
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.602287878414493e-07,
+      "loss": 0.5881,
+      "step": 24495
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.598096765205065e-07,
+      "loss": 0.5466,
+      "step": 24496
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.593906937287886e-07,
+      "loss": 0.611,
+      "step": 24497
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.589718394720623e-07,
+      "loss": 0.6367,
+      "step": 24498
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.585531137560897e-07,
+      "loss": 0.5777,
+      "step": 24499
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.581345165866371e-07,
+      "loss": 0.5952,
+      "step": 24500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.5771604796946e-07,
+      "loss": 0.5625,
+      "step": 24501
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.572977079103193e-07,
+      "loss": 0.6793,
+      "step": 24502
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.568794964149716e-07,
+      "loss": 0.5797,
+      "step": 24503
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.564614134891722e-07,
+      "loss": 0.6275,
+      "step": 24504
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.560434591386733e-07,
+      "loss": 0.5304,
+      "step": 24505
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.556256333692257e-07,
+      "loss": 0.5262,
+      "step": 24506
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.552079361865815e-07,
+      "loss": 0.5845,
+      "step": 24507
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.547903675964873e-07,
+      "loss": 0.5847,
+      "step": 24508
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.543729276046884e-07,
+      "loss": 0.6014,
+      "step": 24509
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.53955616216928e-07,
+      "loss": 0.5792,
+      "step": 24510
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.535384334389527e-07,
+      "loss": 0.6477,
+      "step": 24511
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.531213792765012e-07,
+      "loss": 0.586,
+      "step": 24512
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.527044537353111e-07,
+      "loss": 0.5688,
+      "step": 24513
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.522876568211223e-07,
+      "loss": 0.5267,
+      "step": 24514
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.518709885396679e-07,
+      "loss": 0.5421,
+      "step": 24515
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.514544488966845e-07,
+      "loss": 0.5633,
+      "step": 24516
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.510380378978987e-07,
+      "loss": 0.5154,
+      "step": 24517
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.506217555490469e-07,
+      "loss": 0.6029,
+      "step": 24518
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.502056018558556e-07,
+      "loss": 0.611,
+      "step": 24519
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.497895768240492e-07,
+      "loss": 0.5921,
+      "step": 24520
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.493736804593531e-07,
+      "loss": 0.543,
+      "step": 24521
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.489579127674928e-07,
+      "loss": 0.6051,
+      "step": 24522
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.48542273754188e-07,
+      "loss": 0.5464,
+      "step": 24523
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.481267634251576e-07,
+      "loss": 0.5907,
+      "step": 24524
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.477113817861214e-07,
+      "loss": 0.5385,
+      "step": 24525
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.472961288427937e-07,
+      "loss": 0.5138,
+      "step": 24526
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.4688100460089e-07,
+      "loss": 0.602,
+      "step": 24527
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.464660090661202e-07,
+      "loss": 0.557,
+      "step": 24528
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.460511422441984e-07,
+      "loss": 0.5697,
+      "step": 24529
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.456364041408314e-07,
+      "loss": 0.5826,
+      "step": 24530
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.452217947617279e-07,
+      "loss": 0.5521,
+      "step": 24531
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.448073141125899e-07,
+      "loss": 0.5741,
+      "step": 24532
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.443929621991252e-07,
+      "loss": 0.6038,
+      "step": 24533
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.439787390270347e-07,
+      "loss": 0.5253,
+      "step": 24534
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.435646446020149e-07,
+      "loss": 0.4957,
+      "step": 24535
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.431506789297693e-07,
+      "loss": 0.6044,
+      "step": 24536
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.427368420159918e-07,
+      "loss": 0.5729,
+      "step": 24537
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.423231338663783e-07,
+      "loss": 0.5781,
+      "step": 24538
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.419095544866183e-07,
+      "loss": 0.5605,
+      "step": 24539
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.414961038824064e-07,
+      "loss": 0.5663,
+      "step": 24540
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.410827820594323e-07,
+      "loss": 0.5274,
+      "step": 24541
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.406695890233827e-07,
+      "loss": 0.5383,
+      "step": 24542
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.402565247799408e-07,
+      "loss": 0.5467,
+      "step": 24543
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.398435893347954e-07,
+      "loss": 0.5498,
+      "step": 24544
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.394307826936275e-07,
+      "loss": 0.6487,
+      "step": 24545
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.390181048621147e-07,
+      "loss": 0.514,
+      "step": 24546
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.386055558459392e-07,
+      "loss": 0.6004,
+      "step": 24547
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.381931356507776e-07,
+      "loss": 0.5936,
+      "step": 24548
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.377808442823041e-07,
+      "loss": 0.5409,
+      "step": 24549
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.373686817461911e-07,
+      "loss": 0.5373,
+      "step": 24550
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.369566480481138e-07,
+      "loss": 0.5388,
+      "step": 24551
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.365447431937388e-07,
+      "loss": 0.55,
+      "step": 24552
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.361329671887373e-07,
+      "loss": 0.531,
+      "step": 24553
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.357213200387724e-07,
+      "loss": 0.6183,
+      "step": 24554
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.353098017495119e-07,
+      "loss": 0.5259,
+      "step": 24555
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.348984123266144e-07,
+      "loss": 0.612,
+      "step": 24556
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.344871517757456e-07,
+      "loss": 0.657,
+      "step": 24557
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.34076020102562e-07,
+      "loss": 0.6667,
+      "step": 24558
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.336650173127224e-07,
+      "loss": 0.6057,
+      "step": 24559
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.33254143411881e-07,
+      "loss": 0.5611,
+      "step": 24560
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.328433984056948e-07,
+      "loss": 0.6517,
+      "step": 24561
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.324327822998133e-07,
+      "loss": 0.5959,
+      "step": 24562
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.320222950998889e-07,
+      "loss": 0.5535,
+      "step": 24563
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.316119368115659e-07,
+      "loss": 0.5614,
+      "step": 24564
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.312017074404975e-07,
+      "loss": 0.5936,
+      "step": 24565
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.307916069923248e-07,
+      "loss": 0.6378,
+      "step": 24566
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.303816354726911e-07,
+      "loss": 0.5962,
+      "step": 24567
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.299717928872406e-07,
+      "loss": 0.648,
+      "step": 24568
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.295620792416112e-07,
+      "loss": 0.5413,
+      "step": 24569
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.291524945414407e-07,
+      "loss": 0.5321,
+      "step": 24570
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.287430387923654e-07,
+      "loss": 0.5594,
+      "step": 24571
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.28333712000021e-07,
+      "loss": 0.5276,
+      "step": 24572
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.279245141700397e-07,
+      "loss": 0.5483,
+      "step": 24573
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.275154453080512e-07,
+      "loss": 0.4662,
+      "step": 24574
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.271065054196846e-07,
+      "loss": 0.5581,
+      "step": 24575
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.266976945105707e-07,
+      "loss": 0.5711,
+      "step": 24576
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.262890125863319e-07,
+      "loss": 0.5422,
+      "step": 24577
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.258804596525902e-07,
+      "loss": 0.487,
+      "step": 24578
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.254720357149735e-07,
+      "loss": 0.5918,
+      "step": 24579
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.250637407790971e-07,
+      "loss": 0.5811,
+      "step": 24580
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.246555748505822e-07,
+      "loss": 0.5348,
+      "step": 24581
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.242475379350432e-07,
+      "loss": 0.5627,
+      "step": 24582
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.238396300380978e-07,
+      "loss": 0.5953,
+      "step": 24583
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.23431851165357e-07,
+      "loss": 0.5425,
+      "step": 24584
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.230242013224331e-07,
+      "loss": 0.6141,
+      "step": 24585
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.226166805149348e-07,
+      "loss": 0.5455,
+      "step": 24586
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.22209288748471e-07,
+      "loss": 0.5669,
+      "step": 24587
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.218020260286484e-07,
+      "loss": 0.591,
+      "step": 24588
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.21394892361068e-07,
+      "loss": 0.612,
+      "step": 24589
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.209878877513364e-07,
+      "loss": 0.5712,
+      "step": 24590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.205810122050526e-07,
+      "loss": 0.5412,
+      "step": 24591
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.201742657278153e-07,
+      "loss": 0.5046,
+      "step": 24592
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.197676483252213e-07,
+      "loss": 0.5507,
+      "step": 24593
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.193611600028681e-07,
+      "loss": 0.5315,
+      "step": 24594
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.189548007663471e-07,
+      "loss": 0.6123,
+      "step": 24595
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.185485706212502e-07,
+      "loss": 0.5494,
+      "step": 24596
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.181424695731686e-07,
+      "loss": 0.5954,
+      "step": 24597
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.177364976276912e-07,
+      "loss": 0.5811,
+      "step": 24598
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.173306547904034e-07,
+      "loss": 0.4887,
+      "step": 24599
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.169249410668887e-07,
+      "loss": 0.6275,
+      "step": 24600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.165193564627326e-07,
+      "loss": 0.4764,
+      "step": 24601
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.161139009835159e-07,
+      "loss": 0.5491,
+      "step": 24602
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.157085746348169e-07,
+      "loss": 0.6177,
+      "step": 24603
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.153033774222117e-07,
+      "loss": 0.5908,
+      "step": 24604
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.148983093512806e-07,
+      "loss": 0.5844,
+      "step": 24605
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.144933704275946e-07,
+      "loss": 0.5789,
+      "step": 24606
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.140885606567248e-07,
+      "loss": 0.6441,
+      "step": 24607
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.136838800442457e-07,
+      "loss": 0.4759,
+      "step": 24608
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.132793285957239e-07,
+      "loss": 0.5506,
+      "step": 24609
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.128749063167272e-07,
+      "loss": 0.5739,
+      "step": 24610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.124706132128177e-07,
+      "loss": 0.5286,
+      "step": 24611
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.120664492895622e-07,
+      "loss": 0.6397,
+      "step": 24612
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.116624145525218e-07,
+      "loss": 0.6069,
+      "step": 24613
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.112585090072565e-07,
+      "loss": 0.5916,
+      "step": 24614
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.108547326593206e-07,
+      "loss": 0.6107,
+      "step": 24615
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.104510855142764e-07,
+      "loss": 0.6076,
+      "step": 24616
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.100475675776762e-07,
+      "loss": 0.5966,
+      "step": 24617
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.0964417885507e-07,
+      "loss": 0.5801,
+      "step": 24618
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.092409193520121e-07,
+      "loss": 0.5729,
+      "step": 24619
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.088377890740504e-07,
+      "loss": 0.6122,
+      "step": 24620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.084347880267338e-07,
+      "loss": 0.5806,
+      "step": 24621
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.080319162156045e-07,
+      "loss": 0.5908,
+      "step": 24622
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.076291736462092e-07,
+      "loss": 0.552,
+      "step": 24623
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.072265603240901e-07,
+      "loss": 0.6186,
+      "step": 24624
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.068240762547872e-07,
+      "loss": 0.5455,
+      "step": 24625
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.064217214438361e-07,
+      "loss": 0.5554,
+      "step": 24626
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.06019495896778e-07,
+      "loss": 0.5713,
+      "step": 24627
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.056173996191461e-07,
+      "loss": 0.5814,
+      "step": 24628
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.052154326164717e-07,
+      "loss": 0.612,
+      "step": 24629
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.048135948942902e-07,
+      "loss": 0.5758,
+      "step": 24630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.044118864581284e-07,
+      "loss": 0.5757,
+      "step": 24631
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.04010307313514e-07,
+      "loss": 0.5263,
+      "step": 24632
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.036088574659726e-07,
+      "loss": 0.5604,
+      "step": 24633
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.032075369210322e-07,
+      "loss": 0.6533,
+      "step": 24634
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.028063456842115e-07,
+      "loss": 0.5588,
+      "step": 24635
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.02405283761034e-07,
+      "loss": 0.5588,
+      "step": 24636
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.02004351157014e-07,
+      "loss": 0.5631,
+      "step": 24637
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.016035478776749e-07,
+      "loss": 0.5424,
+      "step": 24638
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.01202873928528e-07,
+      "loss": 0.5745,
+      "step": 24639
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.008023293150878e-07,
+      "loss": 0.5412,
+      "step": 24640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.004019140428674e-07,
+      "loss": 0.5603,
+      "step": 24641
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.00001628117376e-07,
+      "loss": 0.5925,
+      "step": 24642
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.996014715441211e-07,
+      "loss": 0.547,
+      "step": 24643
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.992014443286087e-07,
+      "loss": 0.6244,
+      "step": 24644
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.988015464763463e-07,
+      "loss": 0.5111,
+      "step": 24645
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.984017779928353e-07,
+      "loss": 0.4952,
+      "step": 24646
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.980021388835767e-07,
+      "loss": 0.5737,
+      "step": 24647
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.976026291540682e-07,
+      "loss": 0.567,
+      "step": 24648
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.972032488098111e-07,
+      "loss": 0.5102,
+      "step": 24649
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.968039978563001e-07,
+      "loss": 0.5794,
+      "step": 24650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.96404876299026e-07,
+      "loss": 0.5308,
+      "step": 24651
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.960058841434846e-07,
+      "loss": 0.5383,
+      "step": 24652
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.956070213951659e-07,
+      "loss": 0.5724,
+      "step": 24653
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.952082880595578e-07,
+      "loss": 0.5679,
+      "step": 24654
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.948096841421458e-07,
+      "loss": 0.6155,
+      "step": 24655
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.944112096484189e-07,
+      "loss": 0.5458,
+      "step": 24656
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.940128645838572e-07,
+      "loss": 0.5403,
+      "step": 24657
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.936146489539429e-07,
+      "loss": 0.6167,
+      "step": 24658
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.932165627641573e-07,
+      "loss": 0.5812,
+      "step": 24659
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.92818606019977e-07,
+      "loss": 0.4763,
+      "step": 24660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.924207787268788e-07,
+      "loss": 0.5789,
+      "step": 24661
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.920230808903348e-07,
+      "loss": 0.5581,
+      "step": 24662
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.916255125158221e-07,
+      "loss": 0.5272,
+      "step": 24663
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.912280736088094e-07,
+      "loss": 0.5729,
+      "step": 24664
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.908307641747657e-07,
+      "loss": 0.5261,
+      "step": 24665
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.904335842191555e-07,
+      "loss": 0.6168,
+      "step": 24666
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.900365337474501e-07,
+      "loss": 0.5958,
+      "step": 24667
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.896396127651093e-07,
+      "loss": 0.5541,
+      "step": 24668
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.892428212775958e-07,
+      "loss": 0.6697,
+      "step": 24669
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.888461592903704e-07,
+      "loss": 0.5602,
+      "step": 24670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.884496268088912e-07,
+      "loss": 0.6568,
+      "step": 24671
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.880532238386161e-07,
+      "loss": 0.5752,
+      "step": 24672
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.876569503849961e-07,
+      "loss": 0.5731,
+      "step": 24673
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.87260806453489e-07,
+      "loss": 0.563,
+      "step": 24674
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.86864792049543e-07,
+      "loss": 0.5835,
+      "step": 24675
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.8646890717861e-07,
+      "loss": 0.5629,
+      "step": 24676
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.860731518461338e-07,
+      "loss": 0.5716,
+      "step": 24677
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.856775260575642e-07,
+      "loss": 0.5771,
+      "step": 24678
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.852820298183449e-07,
+      "loss": 0.5345,
+      "step": 24679
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.848866631339145e-07,
+      "loss": 0.5512,
+      "step": 24680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.84491426009719e-07,
+      "loss": 0.5521,
+      "step": 24681
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.840963184511938e-07,
+      "loss": 0.4896,
+      "step": 24682
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.837013404637782e-07,
+      "loss": 0.5969,
+      "step": 24683
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.833064920529031e-07,
+      "loss": 0.5618,
+      "step": 24684
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.829117732240075e-07,
+      "loss": 0.6397,
+      "step": 24685
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.825171839825195e-07,
+      "loss": 0.6309,
+      "step": 24686
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.821227243338712e-07,
+      "loss": 0.583,
+      "step": 24687
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.817283942834873e-07,
+      "loss": 0.5593,
+      "step": 24688
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.813341938367978e-07,
+      "loss": 0.5716,
+      "step": 24689
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.809401229992251e-07,
+      "loss": 0.5362,
+      "step": 24690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.805461817761926e-07,
+      "loss": 0.5806,
+      "step": 24691
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.801523701731216e-07,
+      "loss": 0.6007,
+      "step": 24692
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.797586881954309e-07,
+      "loss": 0.6224,
+      "step": 24693
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.793651358485386e-07,
+      "loss": 0.5817,
+      "step": 24694
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.789717131378581e-07,
+      "loss": 0.5883,
+      "step": 24695
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.785784200688061e-07,
+      "loss": 0.6021,
+      "step": 24696
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.781852566467938e-07,
+      "loss": 0.6004,
+      "step": 24697
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.777922228772314e-07,
+      "loss": 0.5701,
+      "step": 24698
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.773993187655247e-07,
+      "loss": 0.5588,
+      "step": 24699
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.770065443170847e-07,
+      "loss": 0.5695,
+      "step": 24700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.76613899537315e-07,
+      "loss": 0.5635,
+      "step": 24701
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.762213844316155e-07,
+      "loss": 0.5899,
+      "step": 24702
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.758289990053934e-07,
+      "loss": 0.5483,
+      "step": 24703
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.75436743264044e-07,
+      "loss": 0.5269,
+      "step": 24704
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.750446172129653e-07,
+      "loss": 0.5821,
+      "step": 24705
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.746526208575542e-07,
+      "loss": 0.5716,
+      "step": 24706
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.742607542032053e-07,
+      "loss": 0.5409,
+      "step": 24707
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.738690172553107e-07,
+      "loss": 0.5504,
+      "step": 24708
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.73477410019262e-07,
+      "loss": 0.5715,
+      "step": 24709
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.730859325004445e-07,
+      "loss": 0.5541,
+      "step": 24710
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.726945847042497e-07,
+      "loss": 0.5527,
+      "step": 24711
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.723033666360611e-07,
+      "loss": 0.6073,
+      "step": 24712
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.719122783012598e-07,
+      "loss": 0.5719,
+      "step": 24713
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.715213197052316e-07,
+      "loss": 0.5391,
+      "step": 24714
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.711304908533555e-07,
+      "loss": 0.5602,
+      "step": 24715
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.707397917510072e-07,
+      "loss": 0.5894,
+      "step": 24716
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.703492224035645e-07,
+      "loss": 0.5346,
+      "step": 24717
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.699587828164033e-07,
+      "loss": 0.5602,
+      "step": 24718
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.695684729948947e-07,
+      "loss": 0.5522,
+      "step": 24719
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.691782929444089e-07,
+      "loss": 0.5971,
+      "step": 24720
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.687882426703185e-07,
+      "loss": 0.5495,
+      "step": 24721
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.683983221779887e-07,
+      "loss": 0.5926,
+      "step": 24722
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.680085314727846e-07,
+      "loss": 0.6401,
+      "step": 24723
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.676188705600693e-07,
+      "loss": 0.5995,
+      "step": 24724
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.672293394452088e-07,
+      "loss": 0.6248,
+      "step": 24725
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.66839938133561e-07,
+      "loss": 0.5797,
+      "step": 24726
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.664506666304837e-07,
+      "loss": 0.5524,
+      "step": 24727
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.660615249413326e-07,
+      "loss": 0.5353,
+      "step": 24728
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.656725130714657e-07,
+      "loss": 0.5305,
+      "step": 24729
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.652836310262355e-07,
+      "loss": 0.5795,
+      "step": 24730
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.648948788109909e-07,
+      "loss": 0.598,
+      "step": 24731
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.645062564310844e-07,
+      "loss": 0.6046,
+      "step": 24732
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.641177638918627e-07,
+      "loss": 0.5737,
+      "step": 24733
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.637294011986716e-07,
+      "loss": 0.5149,
+      "step": 24734
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.633411683568535e-07,
+      "loss": 0.57,
+      "step": 24735
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.629530653717552e-07,
+      "loss": 0.5627,
+      "step": 24736
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.625650922487136e-07,
+      "loss": 0.579,
+      "step": 24737
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.621772489930699e-07,
+      "loss": 0.5904,
+      "step": 24738
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.617895356101577e-07,
+      "loss": 0.5369,
+      "step": 24739
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.614019521053171e-07,
+      "loss": 0.516,
+      "step": 24740
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.610144984838783e-07,
+      "loss": 0.5552,
+      "step": 24741
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.606271747511727e-07,
+      "loss": 0.5362,
+      "step": 24742
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.602399809125325e-07,
+      "loss": 0.5951,
+      "step": 24743
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.598529169732847e-07,
+      "loss": 0.6094,
+      "step": 24744
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.594659829387561e-07,
+      "loss": 0.5804,
+      "step": 24745
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.590791788142691e-07,
+      "loss": 0.5707,
+      "step": 24746
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.586925046051494e-07,
+      "loss": 0.5188,
+      "step": 24747
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.583059603167162e-07,
+      "loss": 0.5539,
+      "step": 24748
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.579195459542896e-07,
+      "loss": 0.4861,
+      "step": 24749
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.575332615231855e-07,
+      "loss": 0.6599,
+      "step": 24750
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.571471070287215e-07,
+      "loss": 0.5861,
+      "step": 24751
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.567610824762115e-07,
+      "loss": 0.5744,
+      "step": 24752
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.563751878709645e-07,
+      "loss": 0.5474,
+      "step": 24753
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.559894232182939e-07,
+      "loss": 0.588,
+      "step": 24754
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.556037885235078e-07,
+      "loss": 0.5329,
+      "step": 24755
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.552182837919106e-07,
+      "loss": 0.6338,
+      "step": 24756
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.548329090288085e-07,
+      "loss": 0.563,
+      "step": 24757
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.544476642395069e-07,
+      "loss": 0.6614,
+      "step": 24758
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.540625494293039e-07,
+      "loss": 0.5607,
+      "step": 24759
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.536775646035008e-07,
+      "loss": 0.5804,
+      "step": 24760
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.532927097673923e-07,
+      "loss": 0.5666,
+      "step": 24761
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.529079849262797e-07,
+      "loss": 0.627,
+      "step": 24762
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.525233900854532e-07,
+      "loss": 0.5316,
+      "step": 24763
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.521389252502063e-07,
+      "loss": 0.584,
+      "step": 24764
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.517545904258293e-07,
+      "loss": 0.5253,
+      "step": 24765
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.513703856176112e-07,
+      "loss": 0.5332,
+      "step": 24766
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.509863108308366e-07,
+      "loss": 0.555,
+      "step": 24767
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.506023660707949e-07,
+      "loss": 0.5663,
+      "step": 24768
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.502185513427683e-07,
+      "loss": 0.4849,
+      "step": 24769
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.49834866652037e-07,
+      "loss": 0.5571,
+      "step": 24770
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.494513120038814e-07,
+      "loss": 0.5836,
+      "step": 24771
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.490678874035793e-07,
+      "loss": 0.5108,
+      "step": 24772
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.486845928564088e-07,
+      "loss": 0.614,
+      "step": 24773
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.483014283676413e-07,
+      "loss": 0.5969,
+      "step": 24774
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.479183939425525e-07,
+      "loss": 0.5485,
+      "step": 24775
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.475354895864115e-07,
+      "loss": 0.6008,
+      "step": 24776
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.471527153044875e-07,
+      "loss": 0.5609,
+      "step": 24777
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.467700711020474e-07,
+      "loss": 0.572,
+      "step": 24778
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.46387556984359e-07,
+      "loss": 0.5953,
+      "step": 24779
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.46005172956684e-07,
+      "loss": 0.5225,
+      "step": 24780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.456229190242856e-07,
+      "loss": 0.5856,
+      "step": 24781
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.452407951924211e-07,
+      "loss": 0.582,
+      "step": 24782
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.448588014663525e-07,
+      "loss": 0.5254,
+      "step": 24783
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.444769378513348e-07,
+      "loss": 0.5393,
+      "step": 24784
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.440952043526215e-07,
+      "loss": 0.5767,
+      "step": 24785
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.437136009754674e-07,
+      "loss": 0.5922,
+      "step": 24786
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.433321277251247e-07,
+      "loss": 0.5991,
+      "step": 24787
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.429507846068405e-07,
+      "loss": 0.5609,
+      "step": 24788
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.425695716258617e-07,
+      "loss": 0.5874,
+      "step": 24789
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.421884887874373e-07,
+      "loss": 0.5231,
+      "step": 24790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.418075360968089e-07,
+      "loss": 0.4963,
+      "step": 24791
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.4142671355922e-07,
+      "loss": 0.6105,
+      "step": 24792
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.410460211799095e-07,
+      "loss": 0.6346,
+      "step": 24793
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.40665458964118e-07,
+      "loss": 0.5711,
+      "step": 24794
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.402850269170823e-07,
+      "loss": 0.5875,
+      "step": 24795
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.399047250440348e-07,
+      "loss": 0.6053,
+      "step": 24796
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.395245533502114e-07,
+      "loss": 0.5734,
+      "step": 24797
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.391445118408434e-07,
+      "loss": 0.5643,
+      "step": 24798
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.387646005211599e-07,
+      "loss": 0.5857,
+      "step": 24799
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.38384819396387e-07,
+      "loss": 0.5264,
+      "step": 24800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.380051684717547e-07,
+      "loss": 0.6044,
+      "step": 24801
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.376256477524855e-07,
+      "loss": 0.6316,
+      "step": 24802
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.37246257243802e-07,
+      "loss": 0.558,
+      "step": 24803
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.368669969509222e-07,
+      "loss": 0.635,
+      "step": 24804
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.364878668790696e-07,
+      "loss": 0.5231,
+      "step": 24805
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.361088670334591e-07,
+      "loss": 0.6031,
+      "step": 24806
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.357299974193053e-07,
+      "loss": 0.5412,
+      "step": 24807
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.35351258041824e-07,
+      "loss": 0.5111,
+      "step": 24808
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.349726489062257e-07,
+      "loss": 0.5975,
+      "step": 24809
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.345941700177204e-07,
+      "loss": 0.5956,
+      "step": 24810
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.342158213815152e-07,
+      "loss": 0.5553,
+      "step": 24811
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.338376030028181e-07,
+      "loss": 0.5535,
+      "step": 24812
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.334595148868349e-07,
+      "loss": 0.5632,
+      "step": 24813
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.330815570387638e-07,
+      "loss": 0.7003,
+      "step": 24814
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.327037294638115e-07,
+      "loss": 0.579,
+      "step": 24815
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.323260321671742e-07,
+      "loss": 0.5911,
+      "step": 24816
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.319484651540497e-07,
+      "loss": 0.6033,
+      "step": 24817
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.315710284296316e-07,
+      "loss": 0.564,
+      "step": 24818
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.311937219991181e-07,
+      "loss": 0.59,
+      "step": 24819
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.308165458676984e-07,
+      "loss": 0.5973,
+      "step": 24820
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.304395000405637e-07,
+      "loss": 0.6134,
+      "step": 24821
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.300625845229013e-07,
+      "loss": 0.6319,
+      "step": 24822
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.29685799319899e-07,
+      "loss": 0.5832,
+      "step": 24823
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.293091444367415e-07,
+      "loss": 0.5844,
+      "step": 24824
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.289326198786116e-07,
+      "loss": 0.6339,
+      "step": 24825
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.285562256506915e-07,
+      "loss": 0.5724,
+      "step": 24826
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.281799617581607e-07,
+      "loss": 0.538,
+      "step": 24827
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.27803828206196e-07,
+      "loss": 0.4869,
+      "step": 24828
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.27427824999972e-07,
+      "loss": 0.591,
+      "step": 24829
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.270519521446648e-07,
+      "loss": 0.5155,
+      "step": 24830
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.266762096454481e-07,
+      "loss": 0.5693,
+      "step": 24831
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.263005975074897e-07,
+      "loss": 0.5508,
+      "step": 24832
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.25925115735959e-07,
+      "loss": 0.5678,
+      "step": 24833
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.255497643360241e-07,
+      "loss": 0.5312,
+      "step": 24834
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.251745433128497e-07,
+      "loss": 0.6122,
+      "step": 24835
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.247994526715971e-07,
+      "loss": 0.5871,
+      "step": 24836
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.244244924174313e-07,
+      "loss": 0.5889,
+      "step": 24837
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.240496625555103e-07,
+      "loss": 0.6207,
+      "step": 24838
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.236749630909932e-07,
+      "loss": 0.5737,
+      "step": 24839
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.233003940290327e-07,
+      "loss": 0.6107,
+      "step": 24840
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.229259553747867e-07,
+      "loss": 0.6041,
+      "step": 24841
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.22551647133408e-07,
+      "loss": 0.5548,
+      "step": 24842
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.221774693100457e-07,
+      "loss": 0.5871,
+      "step": 24843
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.218034219098489e-07,
+      "loss": 0.5355,
+      "step": 24844
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.214295049379658e-07,
+      "loss": 0.5215,
+      "step": 24845
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.210557183995413e-07,
+      "loss": 0.535,
+      "step": 24846
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.206820622997177e-07,
+      "loss": 0.6155,
+      "step": 24847
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.2030853664364e-07,
+      "loss": 0.606,
+      "step": 24848
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.19935141436445e-07,
+      "loss": 0.6102,
+      "step": 24849
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.195618766832744e-07,
+      "loss": 0.5338,
+      "step": 24850
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.191887423892595e-07,
+      "loss": 0.5952,
+      "step": 24851
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.188157385595405e-07,
+      "loss": 0.5663,
+      "step": 24852
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.184428651992479e-07,
+      "loss": 0.5876,
+      "step": 24853
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.180701223135121e-07,
+      "loss": 0.5912,
+      "step": 24854
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.176975099074622e-07,
+      "loss": 0.5913,
+      "step": 24855
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.173250279862274e-07,
+      "loss": 0.5055,
+      "step": 24856
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.169526765549326e-07,
+      "loss": 0.5836,
+      "step": 24857
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.165804556186993e-07,
+      "loss": 0.5016,
+      "step": 24858
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.162083651826544e-07,
+      "loss": 0.5521,
+      "step": 24859
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.158364052519149e-07,
+      "loss": 0.5955,
+      "step": 24860
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.154645758316001e-07,
+      "loss": 0.5356,
+      "step": 24861
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.150928769268259e-07,
+      "loss": 0.6402,
+      "step": 24862
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.147213085427083e-07,
+      "loss": 0.5353,
+      "step": 24863
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.143498706843609e-07,
+      "loss": 0.6089,
+      "step": 24864
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.139785633568928e-07,
+      "loss": 0.5351,
+      "step": 24865
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.136073865654146e-07,
+      "loss": 0.6174,
+      "step": 24866
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.132363403150353e-07,
+      "loss": 0.6016,
+      "step": 24867
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.128654246108589e-07,
+      "loss": 0.5708,
+      "step": 24868
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.1249463945799e-07,
+      "loss": 0.6506,
+      "step": 24869
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.121239848615323e-07,
+      "loss": 0.5753,
+      "step": 24870
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.117534608265862e-07,
+      "loss": 0.615,
+      "step": 24871
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.113830673582487e-07,
+      "loss": 0.5858,
+      "step": 24872
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.110128044616158e-07,
+      "loss": 0.536,
+      "step": 24873
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.106426721417868e-07,
+      "loss": 0.5625,
+      "step": 24874
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.102726704038518e-07,
+      "loss": 0.5813,
+      "step": 24875
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.099027992529026e-07,
+      "loss": 0.5345,
+      "step": 24876
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.095330586940305e-07,
+      "loss": 0.5157,
+      "step": 24877
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.091634487323238e-07,
+      "loss": 0.5235,
+      "step": 24878
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.087939693728661e-07,
+      "loss": 0.6164,
+      "step": 24879
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.084246206207422e-07,
+      "loss": 0.6373,
+      "step": 24880
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.08055402481037e-07,
+      "loss": 0.6124,
+      "step": 24881
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.076863149588296e-07,
+      "loss": 0.5596,
+      "step": 24882
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.073173580591994e-07,
+      "loss": 0.569,
+      "step": 24883
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.069485317872213e-07,
+      "loss": 0.5917,
+      "step": 24884
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.065798361479746e-07,
+      "loss": 0.6144,
+      "step": 24885
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.062112711465317e-07,
+      "loss": 0.5551,
+      "step": 24886
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.05842836787962e-07,
+      "loss": 0.5706,
+      "step": 24887
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.05474533077338e-07,
+      "loss": 0.5884,
+      "step": 24888
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.05106360019727e-07,
+      "loss": 0.5438,
+      "step": 24889
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.047383176201959e-07,
+      "loss": 0.5826,
+      "step": 24890
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.043704058838073e-07,
+      "loss": 0.5139,
+      "step": 24891
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.04002624815626e-07,
+      "loss": 0.5601,
+      "step": 24892
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.036349744207136e-07,
+      "loss": 0.6106,
+      "step": 24893
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.032674547041283e-07,
+      "loss": 0.592,
+      "step": 24894
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.029000656709249e-07,
+      "loss": 0.6141,
+      "step": 24895
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.025328073261638e-07,
+      "loss": 0.5388,
+      "step": 24896
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.021656796748964e-07,
+      "loss": 0.6198,
+      "step": 24897
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.017986827221733e-07,
+      "loss": 0.6359,
+      "step": 24898
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.014318164730469e-07,
+      "loss": 0.6183,
+      "step": 24899
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.010650809325657e-07,
+      "loss": 0.5709,
+      "step": 24900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.006984761057754e-07,
+      "loss": 0.6077,
+      "step": 24901
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.003320019977187e-07,
+      "loss": 0.6352,
+      "step": 24902
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.999656586134438e-07,
+      "loss": 0.5763,
+      "step": 24903
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.995994459579879e-07,
+      "loss": 0.5477,
+      "step": 24904
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.992333640363911e-07,
+      "loss": 0.6179,
+      "step": 24905
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.988674128536896e-07,
+      "loss": 0.5731,
+      "step": 24906
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.985015924149239e-07,
+      "loss": 0.5615,
+      "step": 24907
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.981359027251231e-07,
+      "loss": 0.5652,
+      "step": 24908
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.977703437893211e-07,
+      "loss": 0.5348,
+      "step": 24909
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.974049156125504e-07,
+      "loss": 0.582,
+      "step": 24910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.970396181998372e-07,
+      "loss": 0.5549,
+      "step": 24911
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.966744515562083e-07,
+      "loss": 0.5494,
+      "step": 24912
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.963094156866887e-07,
+      "loss": 0.5872,
+      "step": 24913
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.959445105963034e-07,
+      "loss": 0.5663,
+      "step": 24914
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.955797362900727e-07,
+      "loss": 0.6103,
+      "step": 24915
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.95215092773016e-07,
+      "loss": 0.6302,
+      "step": 24916
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.948505800501502e-07,
+      "loss": 0.6096,
+      "step": 24917
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.944861981264937e-07,
+      "loss": 0.5287,
+      "step": 24918
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.94121947007059e-07,
+      "loss": 0.6148,
+      "step": 24919
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.937578266968568e-07,
+      "loss": 0.5981,
+      "step": 24920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.933938372009029e-07,
+      "loss": 0.5175,
+      "step": 24921
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.930299785242021e-07,
+      "loss": 0.5559,
+      "step": 24922
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.926662506717628e-07,
+      "loss": 0.5573,
+      "step": 24923
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.923026536485876e-07,
+      "loss": 0.6086,
+      "step": 24924
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.919391874596835e-07,
+      "loss": 0.5933,
+      "step": 24925
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.915758521100511e-07,
+      "loss": 0.6143,
+      "step": 24926
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.912126476046897e-07,
+      "loss": 0.6144,
+      "step": 24927
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.908495739485963e-07,
+      "loss": 0.5624,
+      "step": 24928
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.904866311467693e-07,
+      "loss": 0.6034,
+      "step": 24929
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.901238192042024e-07,
+      "loss": 0.5884,
+      "step": 24930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.897611381258871e-07,
+      "loss": 0.5047,
+      "step": 24931
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.89398587916815e-07,
+      "loss": 0.5941,
+      "step": 24932
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.890361685819756e-07,
+      "loss": 0.5817,
+      "step": 24933
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.886738801263557e-07,
+      "loss": 0.5439,
+      "step": 24934
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.883117225549395e-07,
+      "loss": 0.5664,
+      "step": 24935
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.879496958727125e-07,
+      "loss": 0.5984,
+      "step": 24936
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.875878000846556e-07,
+      "loss": 0.6034,
+      "step": 24937
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.872260351957481e-07,
+      "loss": 0.5819,
+      "step": 24938
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.868644012109703e-07,
+      "loss": 0.6686,
+      "step": 24939
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.86502898135297e-07,
+      "loss": 0.6262,
+      "step": 24940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.861415259737035e-07,
+      "loss": 0.5753,
+      "step": 24941
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.857802847311599e-07,
+      "loss": 0.5818,
+      "step": 24942
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.854191744126424e-07,
+      "loss": 0.5974,
+      "step": 24943
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.850581950231159e-07,
+      "loss": 0.5146,
+      "step": 24944
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.846973465675498e-07,
+      "loss": 0.6126,
+      "step": 24945
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.843366290509077e-07,
+      "loss": 0.6022,
+      "step": 24946
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.839760424781559e-07,
+      "loss": 0.5468,
+      "step": 24947
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.836155868542558e-07,
+      "loss": 0.5588,
+      "step": 24948
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.832552621841657e-07,
+      "loss": 0.5709,
+      "step": 24949
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.828950684728462e-07,
+      "loss": 0.5611,
+      "step": 24950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.825350057252532e-07,
+      "loss": 0.6026,
+      "step": 24951
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.821750739463415e-07,
+      "loss": 0.5352,
+      "step": 24952
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.818152731410619e-07,
+      "loss": 0.5895,
+      "step": 24953
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.814556033143691e-07,
+      "loss": 0.5976,
+      "step": 24954
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.810960644712115e-07,
+      "loss": 0.6056,
+      "step": 24955
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.807366566165361e-07,
+      "loss": 0.5464,
+      "step": 24956
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.803773797552869e-07,
+      "loss": 0.5597,
+      "step": 24957
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.800182338924109e-07,
+      "loss": 0.5651,
+      "step": 24958
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.796592190328486e-07,
+      "loss": 0.6306,
+      "step": 24959
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.793003351815406e-07,
+      "loss": 0.6096,
+      "step": 24960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.789415823434263e-07,
+      "loss": 0.509,
+      "step": 24961
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.785829605234416e-07,
+      "loss": 0.5755,
+      "step": 24962
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.782244697265215e-07,
+      "loss": 0.625,
+      "step": 24963
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.778661099575976e-07,
+      "loss": 0.5869,
+      "step": 24964
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.77507881221605e-07,
+      "loss": 0.6259,
+      "step": 24965
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.771497835234718e-07,
+      "loss": 0.5344,
+      "step": 24966
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7679181686812406e-07,
+      "loss": 0.553,
+      "step": 24967
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.764339812604868e-07,
+      "loss": 0.6091,
+      "step": 24968
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.760762767054883e-07,
+      "loss": 0.5581,
+      "step": 24969
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.75718703208049e-07,
+      "loss": 0.6083,
+      "step": 24970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.753612607730873e-07,
+      "loss": 0.5795,
+      "step": 24971
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7500394940552585e-07,
+      "loss": 0.5223,
+      "step": 24972
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.746467691102785e-07,
+      "loss": 0.6231,
+      "step": 24973
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7428971989226136e-07,
+      "loss": 0.5356,
+      "step": 24974
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7393280175638825e-07,
+      "loss": 0.5976,
+      "step": 24975
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7357601470756964e-07,
+      "loss": 0.5362,
+      "step": 24976
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7321935875071614e-07,
+      "loss": 0.5442,
+      "step": 24977
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.728628338907326e-07,
+      "loss": 0.5339,
+      "step": 24978
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.725064401325308e-07,
+      "loss": 0.6023,
+      "step": 24979
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.721501774810111e-07,
+      "loss": 0.5878,
+      "step": 24980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7179404594107523e-07,
+      "loss": 0.5643,
+      "step": 24981
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.714380455176282e-07,
+      "loss": 0.6287,
+      "step": 24982
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7108217621556487e-07,
+      "loss": 0.5684,
+      "step": 24983
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.707264380397836e-07,
+      "loss": 0.5605,
+      "step": 24984
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.703708309951793e-07,
+      "loss": 0.5304,
+      "step": 24985
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.700153550866471e-07,
+      "loss": 0.5634,
+      "step": 24986
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.696600103190763e-07,
+      "loss": 0.5465,
+      "step": 24987
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.693047966973585e-07,
+      "loss": 0.5803,
+      "step": 24988
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6894971422637993e-07,
+      "loss": 0.592,
+      "step": 24989
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.685947629110299e-07,
+      "loss": 0.6393,
+      "step": 24990
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.682399427561901e-07,
+      "loss": 0.501,
+      "step": 24991
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.678852537667422e-07,
+      "loss": 0.535,
+      "step": 24992
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.675306959475712e-07,
+      "loss": 0.5878,
+      "step": 24993
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.671762693035531e-07,
+      "loss": 0.5587,
+      "step": 24994
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.668219738395663e-07,
+      "loss": 0.5831,
+      "step": 24995
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6646780956048463e-07,
+      "loss": 0.5742,
+      "step": 24996
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6611377647118317e-07,
+      "loss": 0.5823,
+      "step": 24997
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.657598745765346e-07,
+      "loss": 0.6049,
+      "step": 24998
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6540610388140616e-07,
+      "loss": 0.5978,
+      "step": 24999
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.650524643906673e-07,
+      "loss": 0.5183,
+      "step": 25000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6469895610918636e-07,
+      "loss": 0.5578,
+      "step": 25001
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6434557904182495e-07,
+      "loss": 0.5876,
+      "step": 25002
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6399233319344703e-07,
+      "loss": 0.6061,
+      "step": 25003
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6363921856891424e-07,
+      "loss": 0.5456,
+      "step": 25004
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.632862351730849e-07,
+      "loss": 0.5851,
+      "step": 25005
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6293338301081737e-07,
+      "loss": 0.6259,
+      "step": 25006
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.625806620869655e-07,
+      "loss": 0.5317,
+      "step": 25007
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6222807240638436e-07,
+      "loss": 0.5937,
+      "step": 25008
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6187561397392555e-07,
+      "loss": 0.5969,
+      "step": 25009
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.615232867944397e-07,
+      "loss": 0.6323,
+      "step": 25010
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.61171090872774e-07,
+      "loss": 0.6035,
+      "step": 25011
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6081902621377574e-07,
+      "loss": 0.5517,
+      "step": 25012
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.604670928222899e-07,
+      "loss": 0.526,
+      "step": 25013
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.601152907031581e-07,
+      "loss": 0.5839,
+      "step": 25014
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.597636198612232e-07,
+      "loss": 0.5818,
+      "step": 25015
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5941208030132465e-07,
+      "loss": 0.5302,
+      "step": 25016
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.590606720282975e-07,
+      "loss": 0.563,
+      "step": 25017
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.58709395046979e-07,
+      "loss": 0.5074,
+      "step": 25018
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5835824936220297e-07,
+      "loss": 0.5266,
+      "step": 25019
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.580072349788023e-07,
+      "loss": 0.5529,
+      "step": 25020
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.576563519016064e-07,
+      "loss": 0.5345,
+      "step": 25021
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5730560013544256e-07,
+      "loss": 0.5637,
+      "step": 25022
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5695497968513915e-07,
+      "loss": 0.6046,
+      "step": 25023
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5660449055552114e-07,
+      "loss": 0.5439,
+      "step": 25024
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5625413275140915e-07,
+      "loss": 0.6075,
+      "step": 25025
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5590390627762826e-07,
+      "loss": 0.5671,
+      "step": 25026
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5555381113899454e-07,
+      "loss": 0.5551,
+      "step": 25027
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5520384734032755e-07,
+      "loss": 0.5511,
+      "step": 25028
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.548540148864411e-07,
+      "loss": 0.6111,
+      "step": 25029
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5450431378215253e-07,
+      "loss": 0.564,
+      "step": 25030
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5415474403227135e-07,
+      "loss": 0.5666,
+      "step": 25031
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5380530564160806e-07,
+      "loss": 0.6209,
+      "step": 25032
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5345599861497115e-07,
+      "loss": 0.5425,
+      "step": 25033
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.531068229571689e-07,
+      "loss": 0.4976,
+      "step": 25034
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.527577786730053e-07,
+      "loss": 0.6384,
+      "step": 25035
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.524088657672832e-07,
+      "loss": 0.6262,
+      "step": 25036
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.520600842448053e-07,
+      "loss": 0.5308,
+      "step": 25037
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.517114341103701e-07,
+      "loss": 0.5736,
+      "step": 25038
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.513629153687749e-07,
+      "loss": 0.5686,
+      "step": 25039
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.510145280248157e-07,
+      "loss": 0.5204,
+      "step": 25040
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.506662720832877e-07,
+      "loss": 0.6014,
+      "step": 25041
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.503181475489826e-07,
+      "loss": 0.6527,
+      "step": 25042
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.499701544266899e-07,
+      "loss": 0.5924,
+      "step": 25043
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.496222927212002e-07,
+      "loss": 0.6473,
+      "step": 25044
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4927456243729964e-07,
+      "loss": 0.5662,
+      "step": 25045
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.489269635797733e-07,
+      "loss": 0.586,
+      "step": 25046
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.485794961534029e-07,
+      "loss": 0.5682,
+      "step": 25047
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4823216016297246e-07,
+      "loss": 0.5582,
+      "step": 25048
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.478849556132603e-07,
+      "loss": 0.5324,
+      "step": 25049
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.475378825090448e-07,
+      "loss": 0.563,
+      "step": 25050
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.471909408551001e-07,
+      "loss": 0.5544,
+      "step": 25051
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4684413065620325e-07,
+      "loss": 0.5832,
+      "step": 25052
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4649745191712503e-07,
+      "loss": 0.5586,
+      "step": 25053
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.461509046426349e-07,
+      "loss": 0.5869,
+      "step": 25054
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.458044888375046e-07,
+      "loss": 0.6207,
+      "step": 25055
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4545820450649925e-07,
+      "loss": 0.6306,
+      "step": 25056
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4511205165438496e-07,
+      "loss": 0.5434,
+      "step": 25057
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4476603028592246e-07,
+      "loss": 0.5932,
+      "step": 25058
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.444201404058768e-07,
+      "loss": 0.5888,
+      "step": 25059
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.440743820190074e-07,
+      "loss": 0.5476,
+      "step": 25060
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4372875513006953e-07,
+      "loss": 0.5555,
+      "step": 25061
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4338325974382035e-07,
+      "loss": 0.582,
+      "step": 25062
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.430378958650161e-07,
+      "loss": 0.6155,
+      "step": 25063
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4269266349840745e-07,
+      "loss": 0.5998,
+      "step": 25064
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.42347562648745e-07,
+      "loss": 0.5949,
+      "step": 25065
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.420025933207783e-07,
+      "loss": 0.6102,
+      "step": 25066
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.416577555192558e-07,
+      "loss": 0.5737,
+      "step": 25067
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.413130492489204e-07,
+      "loss": 0.619,
+      "step": 25068
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4096847451451485e-07,
+      "loss": 0.6041,
+      "step": 25069
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4062403132078434e-07,
+      "loss": 0.5451,
+      "step": 25070
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.402797196724662e-07,
+      "loss": 0.5415,
+      "step": 25071
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.399355395742988e-07,
+      "loss": 0.591,
+      "step": 25072
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.395914910310184e-07,
+      "loss": 0.5517,
+      "step": 25073
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3924757404736005e-07,
+      "loss": 0.5807,
+      "step": 25074
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3890378862805563e-07,
+      "loss": 0.6268,
+      "step": 25075
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3856013477783564e-07,
+      "loss": 0.5753,
+      "step": 25076
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3821661250143086e-07,
+      "loss": 0.5431,
+      "step": 25077
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3787322180356753e-07,
+      "loss": 0.559,
+      "step": 25078
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.375299626889695e-07,
+      "loss": 0.594,
+      "step": 25079
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3718683516236094e-07,
+      "loss": 0.6029,
+      "step": 25080
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3684383922846574e-07,
+      "loss": 0.5423,
+      "step": 25081
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3650097489200125e-07,
+      "loss": 0.5265,
+      "step": 25082
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.361582421576871e-07,
+      "loss": 0.5859,
+      "step": 25083
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3581564103023835e-07,
+      "loss": 0.5873,
+      "step": 25084
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3547317151437117e-07,
+      "loss": 0.5974,
+      "step": 25085
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.351308336147964e-07,
+      "loss": 0.5389,
+      "step": 25086
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3478862733622564e-07,
+      "loss": 0.6353,
+      "step": 25087
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3444655268336857e-07,
+      "loss": 0.5656,
+      "step": 25088
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.341046096609325e-07,
+      "loss": 0.5633,
+      "step": 25089
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3376279827362253e-07,
+      "loss": 0.5666,
+      "step": 25090
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3342111852614056e-07,
+      "loss": 0.5913,
+      "step": 25091
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.330795704231905e-07,
+      "loss": 0.6167,
+      "step": 25092
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3273815396947306e-07,
+      "loss": 0.5747,
+      "step": 25093
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3239686916968336e-07,
+      "loss": 0.5022,
+      "step": 25094
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.320557160285199e-07,
+      "loss": 0.6337,
+      "step": 25095
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3171469455067785e-07,
+      "loss": 0.5581,
+      "step": 25096
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.313738047408489e-07,
+      "loss": 0.53,
+      "step": 25097
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.310330466037227e-07,
+      "loss": 0.5689,
+      "step": 25098
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3069242014399104e-07,
+      "loss": 0.5577,
+      "step": 25099
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.303519253663402e-07,
+      "loss": 0.5387,
+      "step": 25100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3001156227545636e-07,
+      "loss": 0.5399,
+      "step": 25101
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.296713308760203e-07,
+      "loss": 0.5382,
+      "step": 25102
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.293312311727171e-07,
+      "loss": 0.6172,
+      "step": 25103
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.289912631702253e-07,
+      "loss": 0.556,
+      "step": 25104
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2865142687322225e-07,
+      "loss": 0.5695,
+      "step": 25105
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.283117222863875e-07,
+      "loss": 0.6325,
+      "step": 25106
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2797214941439403e-07,
+      "loss": 0.5129,
+      "step": 25107
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.276327082619147e-07,
+      "loss": 0.5944,
+      "step": 25108
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2729339883361806e-07,
+      "loss": 0.6663,
+      "step": 25109
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.269542211341782e-07,
+      "loss": 0.6455,
+      "step": 25110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2661517516825903e-07,
+      "loss": 0.5628,
+      "step": 25111
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2627626094052687e-07,
+      "loss": 0.6221,
+      "step": 25112
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.259374784556447e-07,
+      "loss": 0.585,
+      "step": 25113
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2559882771827545e-07,
+      "loss": 0.5132,
+      "step": 25114
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.252603087330809e-07,
+      "loss": 0.5045,
+      "step": 25115
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.249219215047151e-07,
+      "loss": 0.537,
+      "step": 25116
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2458366603783883e-07,
+      "loss": 0.5973,
+      "step": 25117
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.24245542337105e-07,
+      "loss": 0.5344,
+      "step": 25118
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2390755040716545e-07,
+      "loss": 0.5524,
+      "step": 25119
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.23569690252672e-07,
+      "loss": 0.6098,
+      "step": 25120
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.232319618782754e-07,
+      "loss": 0.5284,
+      "step": 25121
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.228943652886219e-07,
+      "loss": 0.5605,
+      "step": 25122
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.225569004883578e-07,
+      "loss": 0.5809,
+      "step": 25123
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.222195674821239e-07,
+      "loss": 0.6049,
+      "step": 25124
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2188236627456633e-07,
+      "loss": 0.4953,
+      "step": 25125
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2154529687032375e-07,
+      "loss": 0.5992,
+      "step": 25126
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2120835927403234e-07,
+      "loss": 0.6062,
+      "step": 25127
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.208715534903318e-07,
+      "loss": 0.5734,
+      "step": 25128
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.205348795238562e-07,
+      "loss": 0.6069,
+      "step": 25129
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.201983373792373e-07,
+      "loss": 0.6125,
+      "step": 25130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1986192706110597e-07,
+      "loss": 0.6279,
+      "step": 25131
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1952564857409393e-07,
+      "loss": 0.559,
+      "step": 25132
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.191895019228265e-07,
+      "loss": 0.5551,
+      "step": 25133
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1885348711192986e-07,
+      "loss": 0.5677,
+      "step": 25134
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1851760414602706e-07,
+      "loss": 0.5714,
+      "step": 25135
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1818185302974213e-07,
+      "loss": 0.5666,
+      "step": 25136
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1784623376769475e-07,
+      "loss": 0.5604,
+      "step": 25137
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.175107463645012e-07,
+      "loss": 0.5599,
+      "step": 25138
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.171753908247811e-07,
+      "loss": 0.5777,
+      "step": 25139
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1684016715314746e-07,
+      "loss": 0.5273,
+      "step": 25140
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1650507535421435e-07,
+      "loss": 0.5691,
+      "step": 25141
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1617011543259036e-07,
+      "loss": 0.5223,
+      "step": 25142
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1583528739288723e-07,
+      "loss": 0.5853,
+      "step": 25143
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1550059123971256e-07,
+      "loss": 0.5539,
+      "step": 25144
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.151660269776725e-07,
+      "loss": 0.5602,
+      "step": 25145
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.148315946113668e-07,
+      "loss": 0.5201,
+      "step": 25146
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1449729414540283e-07,
+      "loss": 0.5196,
+      "step": 25147
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.141631255843792e-07,
+      "loss": 0.5749,
+      "step": 25148
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.138290889328922e-07,
+      "loss": 0.545,
+      "step": 25149
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1349518419554035e-07,
+      "loss": 0.5682,
+      "step": 25150
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1316141137691887e-07,
+      "loss": 0.5826,
+      "step": 25151
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.128277704816208e-07,
+      "loss": 0.6242,
+      "step": 25152
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1249426151423466e-07,
+      "loss": 0.5659,
+      "step": 25153
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1216088447935345e-07,
+      "loss": 0.4838,
+      "step": 25154
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1182763938156236e-07,
+      "loss": 0.6236,
+      "step": 25155
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.114945262254466e-07,
+      "loss": 0.5824,
+      "step": 25156
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.111615450155937e-07,
+      "loss": 0.585,
+      "step": 25157
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.108286957565821e-07,
+      "loss": 0.5734,
+      "step": 25158
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.104959784529949e-07,
+      "loss": 0.629,
+      "step": 25159
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.101633931094062e-07,
+      "loss": 0.5862,
+      "step": 25160
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.098309397303979e-07,
+      "loss": 0.6067,
+      "step": 25161
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0949861832054297e-07,
+      "loss": 0.5091,
+      "step": 25162
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0916642888441327e-07,
+      "loss": 0.5459,
+      "step": 25163
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.088343714265797e-07,
+      "loss": 0.5122,
+      "step": 25164
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0850244595161294e-07,
+      "loss": 0.5845,
+      "step": 25165
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.081706524640816e-07,
+      "loss": 0.6384,
+      "step": 25166
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0783899096854874e-07,
+      "loss": 0.537,
+      "step": 25167
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0750746146958064e-07,
+      "loss": 0.5501,
+      "step": 25168
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0717606397173926e-07,
+      "loss": 0.4845,
+      "step": 25169
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0684479847958423e-07,
+      "loss": 0.5691,
+      "step": 25170
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.06513664997672e-07,
+      "loss": 0.5487,
+      "step": 25171
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0618266353056325e-07,
+      "loss": 0.5428,
+      "step": 25172
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0585179408280996e-07,
+      "loss": 0.5525,
+      "step": 25173
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0552105665896735e-07,
+      "loss": 0.5427,
+      "step": 25174
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0519045126358293e-07,
+      "loss": 0.5369,
+      "step": 25175
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0485997790121077e-07,
+      "loss": 0.5711,
+      "step": 25176
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.045296365763962e-07,
+      "loss": 0.6541,
+      "step": 25177
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0419942729368445e-07,
+      "loss": 0.639,
+      "step": 25178
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0386935005762074e-07,
+      "loss": 0.5595,
+      "step": 25179
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.03539404872747e-07,
+      "loss": 0.5883,
+      "step": 25180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.03209591743603e-07,
+      "loss": 0.557,
+      "step": 25181
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0287991067472607e-07,
+      "loss": 0.5537,
+      "step": 25182
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0255036167065607e-07,
+      "loss": 0.5491,
+      "step": 25183
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.02220944735926e-07,
+      "loss": 0.5427,
+      "step": 25184
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0189165987506883e-07,
+      "loss": 0.5837,
+      "step": 25185
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.015625070926155e-07,
+      "loss": 0.5478,
+      "step": 25186
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0123348639309665e-07,
+      "loss": 0.602,
+      "step": 25187
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0090459778103997e-07,
+      "loss": 0.5537,
+      "step": 25188
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0057584126096837e-07,
+      "loss": 0.5803,
+      "step": 25189
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.002472168374094e-07,
+      "loss": 0.6271,
+      "step": 25190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.999187245148839e-07,
+      "loss": 0.5629,
+      "step": 25191
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.995903642979104e-07,
+      "loss": 0.5537,
+      "step": 25192
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9926213619101097e-07,
+      "loss": 0.6414,
+      "step": 25193
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.989340401987007e-07,
+      "loss": 0.5634,
+      "step": 25194
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9860607632549396e-07,
+      "loss": 0.6251,
+      "step": 25195
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.982782445759026e-07,
+      "loss": 0.5982,
+      "step": 25196
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9795054495444075e-07,
+      "loss": 0.5392,
+      "step": 25197
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9762297746561706e-07,
+      "loss": 0.5335,
+      "step": 25198
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9729554211393686e-07,
+      "loss": 0.5055,
+      "step": 25199
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9696823890390867e-07,
+      "loss": 0.6235,
+      "step": 25200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9664106784003565e-07,
+      "loss": 0.4832,
+      "step": 25201
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.963140289268208e-07,
+      "loss": 0.4941,
+      "step": 25202
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9598712216876054e-07,
+      "loss": 0.5667,
+      "step": 25203
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.956603475703591e-07,
+      "loss": 0.5282,
+      "step": 25204
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.953337051361095e-07,
+      "loss": 0.6082,
+      "step": 25205
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9500719487050696e-07,
+      "loss": 0.5159,
+      "step": 25206
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.946808167780447e-07,
+      "loss": 0.5178,
+      "step": 25207
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.943545708632157e-07,
+      "loss": 0.5712,
+      "step": 25208
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.940284571305075e-07,
+      "loss": 0.5388,
+      "step": 25209
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9370247558440765e-07,
+      "loss": 0.5063,
+      "step": 25210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9337662622940366e-07,
+      "loss": 0.6062,
+      "step": 25211
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.930509090699786e-07,
+      "loss": 0.5302,
+      "step": 25212
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.927253241106144e-07,
+      "loss": 0.6036,
+      "step": 25213
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9239987135578973e-07,
+      "loss": 0.5345,
+      "step": 25214
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9207455080998545e-07,
+      "loss": 0.5803,
+      "step": 25215
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9174936247767916e-07,
+      "loss": 0.5883,
+      "step": 25216
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.914243063633427e-07,
+      "loss": 0.555,
+      "step": 25217
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.910993824714504e-07,
+      "loss": 0.5353,
+      "step": 25218
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.907745908064742e-07,
+      "loss": 0.5703,
+      "step": 25219
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.904499313728838e-07,
+      "loss": 0.4781,
+      "step": 25220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9012540417514457e-07,
+      "loss": 0.5757,
+      "step": 25221
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.898010092177251e-07,
+      "loss": 0.5188,
+      "step": 25222
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.894767465050875e-07,
+      "loss": 0.5392,
+      "step": 25223
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8915261604169475e-07,
+      "loss": 0.5248,
+      "step": 25224
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8882861783200556e-07,
+      "loss": 0.5475,
+      "step": 25225
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8850475188048074e-07,
+      "loss": 0.5717,
+      "step": 25226
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8818101819157684e-07,
+      "loss": 0.5147,
+      "step": 25227
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.878574167697469e-07,
+      "loss": 0.5448,
+      "step": 25228
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.87533947619444e-07,
+      "loss": 0.5887,
+      "step": 25229
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8721061074512125e-07,
+      "loss": 0.5966,
+      "step": 25230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.868874061512273e-07,
+      "loss": 0.6354,
+      "step": 25231
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.865643338422076e-07,
+      "loss": 0.5391,
+      "step": 25232
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.862413938225118e-07,
+      "loss": 0.5929,
+      "step": 25233
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.859185860965808e-07,
+      "loss": 0.5573,
+      "step": 25234
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8559591066885893e-07,
+      "loss": 0.5118,
+      "step": 25235
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8527336754378366e-07,
+      "loss": 0.523,
+      "step": 25236
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8495095672579584e-07,
+      "loss": 0.6114,
+      "step": 25237
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8462867821933203e-07,
+      "loss": 0.5195,
+      "step": 25238
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.843065320288264e-07,
+      "loss": 0.5479,
+      "step": 25239
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8398451815870984e-07,
+      "loss": 0.5369,
+      "step": 25240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8366263661341773e-07,
+      "loss": 0.5692,
+      "step": 25241
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8334088739737654e-07,
+      "loss": 0.507,
+      "step": 25242
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.830192705150149e-07,
+      "loss": 0.5948,
+      "step": 25243
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.826977859707581e-07,
+      "loss": 0.5573,
+      "step": 25244
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.823764337690317e-07,
+      "loss": 0.5809,
+      "step": 25245
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8205521391425525e-07,
+      "loss": 0.6056,
+      "step": 25246
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8173412641084873e-07,
+      "loss": 0.5126,
+      "step": 25247
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8141317126323406e-07,
+      "loss": 0.5407,
+      "step": 25248
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8109234847582555e-07,
+      "loss": 0.4825,
+      "step": 25249
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.807716580530374e-07,
+      "loss": 0.512,
+      "step": 25250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8045109999928274e-07,
+      "loss": 0.5723,
+      "step": 25251
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.801306743189759e-07,
+      "loss": 0.5433,
+      "step": 25252
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7981038101652214e-07,
+      "loss": 0.5735,
+      "step": 25253
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7949022009633017e-07,
+      "loss": 0.5359,
+      "step": 25254
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.791701915628065e-07,
+      "loss": 0.5439,
+      "step": 25255
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7885029542035544e-07,
+      "loss": 0.5493,
+      "step": 25256
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.785305316733778e-07,
+      "loss": 0.628,
+      "step": 25257
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.782109003262724e-07,
+      "loss": 0.6206,
+      "step": 25258
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.778914013834412e-07,
+      "loss": 0.5782,
+      "step": 25259
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.775720348492795e-07,
+      "loss": 0.5663,
+      "step": 25260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.772528007281806e-07,
+      "loss": 0.5594,
+      "step": 25261
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.769336990245387e-07,
+      "loss": 0.6116,
+      "step": 25262
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.766147297427447e-07,
+      "loss": 0.5512,
+      "step": 25263
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.762958928871885e-07,
+      "loss": 0.5748,
+      "step": 25264
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.759771884622554e-07,
+      "loss": 0.6098,
+      "step": 25265
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7565861647233413e-07,
+      "loss": 0.5246,
+      "step": 25266
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.753401769218068e-07,
+      "loss": 0.5875,
+      "step": 25267
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7502186981505537e-07,
+      "loss": 0.5544,
+      "step": 25268
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.747036951564587e-07,
+      "loss": 0.5679,
+      "step": 25269
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.743856529503986e-07,
+      "loss": 0.56,
+      "step": 25270
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.740677432012496e-07,
+      "loss": 0.546,
+      "step": 25271
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7374996591338586e-07,
+      "loss": 0.6217,
+      "step": 25272
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7343232109118165e-07,
+      "loss": 0.5666,
+      "step": 25273
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7311480873900687e-07,
+      "loss": 0.5898,
+      "step": 25274
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.727974288612313e-07,
+      "loss": 0.5635,
+      "step": 25275
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.724801814622214e-07,
+      "loss": 0.4882,
+      "step": 25276
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7216306654634496e-07,
+      "loss": 0.5704,
+      "step": 25277
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.71846084117965e-07,
+      "loss": 0.5481,
+      "step": 25278
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7152923418144247e-07,
+      "loss": 0.6159,
+      "step": 25279
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.712125167411362e-07,
+      "loss": 0.5493,
+      "step": 25280
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7089593180140823e-07,
+      "loss": 0.6088,
+      "step": 25281
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.705794793666129e-07,
+      "loss": 0.5845,
+      "step": 25282
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.702631594411044e-07,
+      "loss": 0.5338,
+      "step": 25283
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.699469720292359e-07,
+      "loss": 0.5307,
+      "step": 25284
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.696309171353607e-07,
+      "loss": 0.626,
+      "step": 25285
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.693149947638253e-07,
+      "loss": 0.5614,
+      "step": 25286
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.689992049189761e-07,
+      "loss": 0.592,
+      "step": 25287
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.68683547605162e-07,
+      "loss": 0.6177,
+      "step": 25288
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.68368022826725e-07,
+      "loss": 0.535,
+      "step": 25289
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6805263058800723e-07,
+      "loss": 0.5767,
+      "step": 25290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6773737089334736e-07,
+      "loss": 0.5779,
+      "step": 25291
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.674222437470865e-07,
+      "loss": 0.5807,
+      "step": 25292
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.671072491535588e-07,
+      "loss": 0.5391,
+      "step": 25293
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6679238711709865e-07,
+      "loss": 0.5497,
+      "step": 25294
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6647765764204145e-07,
+      "loss": 0.5846,
+      "step": 25295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6616306073271603e-07,
+      "loss": 0.617,
+      "step": 25296
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6584859639345104e-07,
+      "loss": 0.5715,
+      "step": 25297
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6553426462857423e-07,
+      "loss": 0.6132,
+      "step": 25298
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6522006544241315e-07,
+      "loss": 0.5808,
+      "step": 25299
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.649059988392889e-07,
+      "loss": 0.5483,
+      "step": 25300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.645920648235257e-07,
+      "loss": 0.6284,
+      "step": 25301
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6427826339944017e-07,
+      "loss": 0.5846,
+      "step": 25302
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6396459457135324e-07,
+      "loss": 0.6208,
+      "step": 25303
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6365105834358036e-07,
+      "loss": 0.5516,
+      "step": 25304
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.633376547204348e-07,
+      "loss": 0.5157,
+      "step": 25305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6302438370623304e-07,
+      "loss": 0.5367,
+      "step": 25306
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.627112453052828e-07,
+      "loss": 0.5999,
+      "step": 25307
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.623982395218939e-07,
+      "loss": 0.6142,
+      "step": 25308
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.620853663603718e-07,
+      "loss": 0.5845,
+      "step": 25309
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6177262582502537e-07,
+      "loss": 0.6341,
+      "step": 25310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6146001792015663e-07,
+      "loss": 0.5825,
+      "step": 25311
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6114754265006767e-07,
+      "loss": 0.6196,
+      "step": 25312
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6083520001905623e-07,
+      "loss": 0.6006,
+      "step": 25313
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6052299003142333e-07,
+      "loss": 0.5924,
+      "step": 25314
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6021091269146327e-07,
+      "loss": 0.5857,
+      "step": 25315
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5989896800347146e-07,
+      "loss": 0.537,
+      "step": 25316
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.595871559717412e-07,
+      "loss": 0.5919,
+      "step": 25317
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5927547660056237e-07,
+      "loss": 0.5738,
+      "step": 25318
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.589639298942238e-07,
+      "loss": 0.5899,
+      "step": 25319
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.58652515857012e-07,
+      "loss": 0.6101,
+      "step": 25320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5834123449321357e-07,
+      "loss": 0.5923,
+      "step": 25321
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5803008580711284e-07,
+      "loss": 0.5761,
+      "step": 25322
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.577190698029886e-07,
+      "loss": 0.6234,
+      "step": 25323
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.574081864851242e-07,
+      "loss": 0.5287,
+      "step": 25324
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.57097435857795e-07,
+      "loss": 0.5598,
+      "step": 25325
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.567868179252787e-07,
+      "loss": 0.507,
+      "step": 25326
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5647633269184747e-07,
+      "loss": 0.5605,
+      "step": 25327
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5616598016177675e-07,
+      "loss": 0.4888,
+      "step": 25328
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5585576033933536e-07,
+      "loss": 0.5305,
+      "step": 25329
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.555456732287943e-07,
+      "loss": 0.5326,
+      "step": 25330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.552357188344169e-07,
+      "loss": 0.5693,
+      "step": 25331
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.549258971604719e-07,
+      "loss": 0.596,
+      "step": 25332
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.546162082112214e-07,
+      "loss": 0.5809,
+      "step": 25333
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.543066519909266e-07,
+      "loss": 0.5408,
+      "step": 25334
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.539972285038473e-07,
+      "loss": 0.5703,
+      "step": 25335
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5368793775424347e-07,
+      "loss": 0.6269,
+      "step": 25336
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5337877974636946e-07,
+      "loss": 0.5737,
+      "step": 25337
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5306975448447747e-07,
+      "loss": 0.5694,
+      "step": 25338
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.52760861972824e-07,
+      "loss": 0.6072,
+      "step": 25339
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5245210221565793e-07,
+      "loss": 0.5718,
+      "step": 25340
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5214347521722815e-07,
+      "loss": 0.5703,
+      "step": 25341
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5183498098178e-07,
+      "loss": 0.5686,
+      "step": 25342
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5152661951356136e-07,
+      "loss": 0.5696,
+      "step": 25343
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.512183908168143e-07,
+      "loss": 0.584,
+      "step": 25344
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5091029489577876e-07,
+      "loss": 0.6074,
+      "step": 25345
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5060233175469806e-07,
+      "loss": 0.5289,
+      "step": 25346
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.502945013978076e-07,
+      "loss": 0.5545,
+      "step": 25347
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.499868038293441e-07,
+      "loss": 0.5187,
+      "step": 25348
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.496792390535397e-07,
+      "loss": 0.5521,
+      "step": 25349
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.493718070746299e-07,
+      "loss": 0.5923,
+      "step": 25350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4906450789684353e-07,
+      "loss": 0.5831,
+      "step": 25351
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4875734152441054e-07,
+      "loss": 0.5897,
+      "step": 25352
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.484503079615542e-07,
+      "loss": 0.5623,
+      "step": 25353
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.481434072125045e-07,
+      "loss": 0.465,
+      "step": 25354
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.478366392814825e-07,
+      "loss": 0.5402,
+      "step": 25355
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.47530004172707e-07,
+      "loss": 0.5301,
+      "step": 25356
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.472235018904024e-07,
+      "loss": 0.5415,
+      "step": 25357
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.469171324387843e-07,
+      "loss": 0.6365,
+      "step": 25358
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.466108958220671e-07,
+      "loss": 0.5268,
+      "step": 25359
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.463047920444662e-07,
+      "loss": 0.5936,
+      "step": 25360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4599882111019503e-07,
+      "loss": 0.5469,
+      "step": 25361
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.456929830234623e-07,
+      "loss": 0.5919,
+      "step": 25362
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4538727778847705e-07,
+      "loss": 0.6076,
+      "step": 25363
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.450817054094446e-07,
+      "loss": 0.5398,
+      "step": 25364
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.44776265890574e-07,
+      "loss": 0.6266,
+      "step": 25365
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.44470959236064e-07,
+      "loss": 0.5917,
+      "step": 25366
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.441657854501179e-07,
+      "loss": 0.5969,
+      "step": 25367
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4386074453693573e-07,
+      "loss": 0.547,
+      "step": 25368
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4355583650071413e-07,
+      "loss": 0.5097,
+      "step": 25369
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.432510613456497e-07,
+      "loss": 0.5921,
+      "step": 25370
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.429464190759346e-07,
+      "loss": 0.6199,
+      "step": 25371
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.426419096957623e-07,
+      "loss": 0.5556,
+      "step": 25372
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4233753320932374e-07,
+      "loss": 0.6171,
+      "step": 25373
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.420332896208067e-07,
+      "loss": 0.5327,
+      "step": 25374
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4172917893439684e-07,
+      "loss": 0.584,
+      "step": 25375
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4142520115428066e-07,
+      "loss": 0.5497,
+      "step": 25376
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.411213562846416e-07,
+      "loss": 0.6093,
+      "step": 25377
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4081764432965737e-07,
+      "loss": 0.5281,
+      "step": 25378
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4051406529351126e-07,
+      "loss": 0.536,
+      "step": 25379
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.402106191803789e-07,
+      "loss": 0.5654,
+      "step": 25380
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.399073059944369e-07,
+      "loss": 0.5652,
+      "step": 25381
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3960412573985636e-07,
+      "loss": 0.5581,
+      "step": 25382
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.393010784208128e-07,
+      "loss": 0.5308,
+      "step": 25383
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3899816404147636e-07,
+      "loss": 0.5951,
+      "step": 25384
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.386953826060113e-07,
+      "loss": 0.5546,
+      "step": 25385
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3839273411858887e-07,
+      "loss": 0.5716,
+      "step": 25386
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.380902185833712e-07,
+      "loss": 0.5824,
+      "step": 25387
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3778783600452283e-07,
+      "loss": 0.5672,
+      "step": 25388
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3748558638620146e-07,
+      "loss": 0.5576,
+      "step": 25389
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.371834697325704e-07,
+      "loss": 0.5744,
+      "step": 25390
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.368814860477854e-07,
+      "loss": 0.5165,
+      "step": 25391
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.365796353360018e-07,
+      "loss": 0.6189,
+      "step": 25392
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3627791760137307e-07,
+      "loss": 0.5678,
+      "step": 25393
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.359763328480514e-07,
+      "loss": 0.6138,
+      "step": 25394
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.35674881080188e-07,
+      "loss": 0.631,
+      "step": 25395
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.353735623019294e-07,
+      "loss": 0.5516,
+      "step": 25396
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.350723765174224e-07,
+      "loss": 0.5214,
+      "step": 25397
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3477132373081257e-07,
+      "loss": 0.523,
+      "step": 25398
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3447040394623986e-07,
+      "loss": 0.5266,
+      "step": 25399
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.341696171678499e-07,
+      "loss": 0.5503,
+      "step": 25400
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.338689633997783e-07,
+      "loss": 0.4987,
+      "step": 25401
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3356844264616274e-07,
+      "loss": 0.6126,
+      "step": 25402
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.332680549111378e-07,
+      "loss": 0.6167,
+      "step": 25403
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.329678001988401e-07,
+      "loss": 0.5975,
+      "step": 25404
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3266767851339977e-07,
+      "loss": 0.6199,
+      "step": 25405
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.323676898589467e-07,
+      "loss": 0.5566,
+      "step": 25406
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3206783423960774e-07,
+      "loss": 0.6083,
+      "step": 25407
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3176811165951064e-07,
+      "loss": 0.575,
+      "step": 25408
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.314685221227809e-07,
+      "loss": 0.5386,
+      "step": 25409
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3116906563353757e-07,
+      "loss": 0.4812,
+      "step": 25410
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.308697421959051e-07,
+      "loss": 0.5194,
+      "step": 25411
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.305705518140012e-07,
+      "loss": 0.5532,
+      "step": 25412
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3027149449194273e-07,
+      "loss": 0.5757,
+      "step": 25413
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.299725702338441e-07,
+      "loss": 0.631,
+      "step": 25414
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2967377904382094e-07,
+      "loss": 0.5671,
+      "step": 25415
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2937512092598323e-07,
+      "loss": 0.5854,
+      "step": 25416
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2907659588444106e-07,
+      "loss": 0.5049,
+      "step": 25417
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.287782039233023e-07,
+      "loss": 0.5603,
+      "step": 25418
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.284799450466747e-07,
+      "loss": 0.552,
+      "step": 25419
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2818181925866057e-07,
+      "loss": 0.699,
+      "step": 25420
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2788382656336103e-07,
+      "loss": 0.6284,
+      "step": 25421
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2758596696488066e-07,
+      "loss": 0.4852,
+      "step": 25422
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.272882404673172e-07,
+      "loss": 0.5258,
+      "step": 25423
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.269906470747652e-07,
+      "loss": 0.5598,
+      "step": 25424
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2669318679132146e-07,
+      "loss": 0.5938,
+      "step": 25425
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.263958596210792e-07,
+      "loss": 0.4891,
+      "step": 25426
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2609866556813085e-07,
+      "loss": 0.614,
+      "step": 25427
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2580160463656306e-07,
+      "loss": 0.5782,
+      "step": 25428
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2550467683046707e-07,
+      "loss": 0.6283,
+      "step": 25429
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.252078821539284e-07,
+      "loss": 0.6108,
+      "step": 25430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2491122061102945e-07,
+      "loss": 0.5647,
+      "step": 25431
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.246146922058524e-07,
+      "loss": 0.569,
+      "step": 25432
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2431829694247963e-07,
+      "loss": 0.4726,
+      "step": 25433
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2402203482498895e-07,
+      "loss": 0.6025,
+      "step": 25434
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.237259058574571e-07,
+      "loss": 0.5755,
+      "step": 25435
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2342991004395754e-07,
+      "loss": 0.4838,
+      "step": 25436
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.231340473885669e-07,
+      "loss": 0.559,
+      "step": 25437
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2283831789535425e-07,
+      "loss": 0.5854,
+      "step": 25438
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.225427215683885e-07,
+      "loss": 0.5515,
+      "step": 25439
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.222472584117398e-07,
+      "loss": 0.6299,
+      "step": 25440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2195192842947254e-07,
+      "loss": 0.5945,
+      "step": 25441
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.216567316256491e-07,
+      "loss": 0.5176,
+      "step": 25442
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2136166800433297e-07,
+      "loss": 0.5932,
+      "step": 25443
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.210667375695864e-07,
+      "loss": 0.5246,
+      "step": 25444
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.207719403254661e-07,
+      "loss": 0.5732,
+      "step": 25445
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2047727627602775e-07,
+      "loss": 0.5883,
+      "step": 25446
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.20182745425327e-07,
+      "loss": 0.5874,
+      "step": 25447
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.198883477774184e-07,
+      "loss": 0.542,
+      "step": 25448
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.195940833363509e-07,
+      "loss": 0.5743,
+      "step": 25449
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.192999521061746e-07,
+      "loss": 0.5336,
+      "step": 25450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.190059540909374e-07,
+      "loss": 0.5673,
+      "step": 25451
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.187120892946849e-07,
+      "loss": 0.6077,
+      "step": 25452
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.184183577214617e-07,
+      "loss": 0.6024,
+      "step": 25453
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1812475937530674e-07,
+      "loss": 0.534,
+      "step": 25454
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.178312942602635e-07,
+      "loss": 0.5258,
+      "step": 25455
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.175379623803698e-07,
+      "loss": 0.5663,
+      "step": 25456
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.172447637396614e-07,
+      "loss": 0.5532,
+      "step": 25457
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1695169834217167e-07,
+      "loss": 0.5543,
+      "step": 25458
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1665876619193624e-07,
+      "loss": 0.6067,
+      "step": 25459
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.163659672929842e-07,
+      "loss": 0.5076,
+      "step": 25460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1607330164934446e-07,
+      "loss": 0.5484,
+      "step": 25461
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.157807692650461e-07,
+      "loss": 0.6469,
+      "step": 25462
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.154883701441136e-07,
+      "loss": 0.5561,
+      "step": 25463
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1519610429057044e-07,
+      "loss": 0.5602,
+      "step": 25464
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1490397170843787e-07,
+      "loss": 0.546,
+      "step": 25465
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1461197240173824e-07,
+      "loss": 0.5966,
+      "step": 25466
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.143201063744883e-07,
+      "loss": 0.5448,
+      "step": 25467
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.140283736307037e-07,
+      "loss": 0.5446,
+      "step": 25468
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1373677417439794e-07,
+      "loss": 0.5585,
+      "step": 25469
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1344530800958785e-07,
+      "loss": 0.6053,
+      "step": 25470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1315397514028013e-07,
+      "loss": 0.5456,
+      "step": 25471
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1286277557048497e-07,
+      "loss": 0.6259,
+      "step": 25472
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.125717093042113e-07,
+      "loss": 0.5594,
+      "step": 25473
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.122807763454627e-07,
+      "loss": 0.547,
+      "step": 25474
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.119899766982426e-07,
+      "loss": 0.5341,
+      "step": 25475
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.116993103665522e-07,
+      "loss": 0.6106,
+      "step": 25476
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.114087773543939e-07,
+      "loss": 0.61,
+      "step": 25477
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.111183776657634e-07,
+      "loss": 0.5311,
+      "step": 25478
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1082811130465627e-07,
+      "loss": 0.5675,
+      "step": 25479
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.105379782750695e-07,
+      "loss": 0.5096,
+      "step": 25480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1024797858099307e-07,
+      "loss": 0.6229,
+      "step": 25481
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0995811222641946e-07,
+      "loss": 0.6154,
+      "step": 25482
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0966837921533546e-07,
+      "loss": 0.5668,
+      "step": 25483
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0937877955173e-07,
+      "loss": 0.4933,
+      "step": 25484
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.090893132395878e-07,
+      "loss": 0.6257,
+      "step": 25485
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0879998028289227e-07,
+      "loss": 0.5486,
+      "step": 25486
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.085107806856225e-07,
+      "loss": 0.607,
+      "step": 25487
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0822171445176186e-07,
+      "loss": 0.5361,
+      "step": 25488
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0793278158528506e-07,
+      "loss": 0.6087,
+      "step": 25489
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0764398209017e-07,
+      "loss": 0.5974,
+      "step": 25490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0735531597039013e-07,
+      "loss": 0.5566,
+      "step": 25491
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.070667832299179e-07,
+      "loss": 0.5544,
+      "step": 25492
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.067783838727245e-07,
+      "loss": 0.5812,
+      "step": 25493
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.064901179027757e-07,
+      "loss": 0.5775,
+      "step": 25494
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0620198532404164e-07,
+      "loss": 0.5408,
+      "step": 25495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0591398614048695e-07,
+      "loss": 0.6116,
+      "step": 25496
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.056261203560729e-07,
+      "loss": 0.5892,
+      "step": 25497
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.053383879747607e-07,
+      "loss": 0.5798,
+      "step": 25498
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0505078900051166e-07,
+      "loss": 0.6208,
+      "step": 25499
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0476332343728263e-07,
+      "loss": 0.5709,
+      "step": 25500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.044759912890283e-07,
+      "loss": 0.5305,
+      "step": 25501
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.041887925597042e-07,
+      "loss": 0.5562,
+      "step": 25502
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0390172725326294e-07,
+      "loss": 0.5561,
+      "step": 25503
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0361479537365237e-07,
+      "loss": 0.6496,
+      "step": 25504
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0332799692482153e-07,
+      "loss": 0.6402,
+      "step": 25505
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.030413319107195e-07,
+      "loss": 0.5295,
+      "step": 25506
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.027548003352887e-07,
+      "loss": 0.5833,
+      "step": 25507
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0246840220247153e-07,
+      "loss": 0.5834,
+      "step": 25508
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.021821375162104e-07,
+      "loss": 0.5493,
+      "step": 25509
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0189600628044545e-07,
+      "loss": 0.6208,
+      "step": 25510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0161000849911246e-07,
+      "loss": 0.5146,
+      "step": 25511
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0132414417614606e-07,
+      "loss": 0.6001,
+      "step": 25512
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.010384133154831e-07,
+      "loss": 0.5716,
+      "step": 25513
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0075281592105376e-07,
+      "loss": 0.6379,
+      "step": 25514
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0046735199678823e-07,
+      "loss": 0.5712,
+      "step": 25515
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0018202154661334e-07,
+      "loss": 0.5581,
+      "step": 25516
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.998968245744582e-07,
+      "loss": 0.6178,
+      "step": 25517
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9961176108424526e-07,
+      "loss": 0.5772,
+      "step": 25518
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9932683107989915e-07,
+      "loss": 0.5539,
+      "step": 25519
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9904203456533777e-07,
+      "loss": 0.5797,
+      "step": 25520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.987573715444836e-07,
+      "loss": 0.6126,
+      "step": 25521
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9847284202125236e-07,
+      "loss": 0.6688,
+      "step": 25522
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9818844599955766e-07,
+      "loss": 0.5833,
+      "step": 25523
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9790418348331517e-07,
+      "loss": 0.5653,
+      "step": 25524
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9762005447643737e-07,
+      "loss": 0.5539,
+      "step": 25525
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9733605898283227e-07,
+      "loss": 0.6232,
+      "step": 25526
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9705219700640776e-07,
+      "loss": 0.5251,
+      "step": 25527
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.967684685510719e-07,
+      "loss": 0.5323,
+      "step": 25528
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.964848736207271e-07,
+      "loss": 0.594,
+      "step": 25529
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.96201412219278e-07,
+      "loss": 0.5967,
+      "step": 25530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.959180843506215e-07,
+      "loss": 0.5501,
+      "step": 25531
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.956348900186612e-07,
+      "loss": 0.5928,
+      "step": 25532
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9535182922729057e-07,
+      "loss": 0.653,
+      "step": 25533
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.950689019804065e-07,
+      "loss": 0.5228,
+      "step": 25534
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.947861082819015e-07,
+      "loss": 0.6404,
+      "step": 25535
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.945034481356679e-07,
+      "loss": 0.5785,
+      "step": 25536
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.942209215455949e-07,
+      "loss": 0.575,
+      "step": 25537
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9393852851556937e-07,
+      "loss": 0.5905,
+      "step": 25538
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.936562690494793e-07,
+      "loss": 0.625,
+      "step": 25539
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9337414315120717e-07,
+      "loss": 0.5487,
+      "step": 25540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.930921508246354e-07,
+      "loss": 0.5149,
+      "step": 25541
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.928102920736453e-07,
+      "loss": 0.5691,
+      "step": 25542
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9252856690211605e-07,
+      "loss": 0.5279,
+      "step": 25543
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9224697531392235e-07,
+      "loss": 0.5387,
+      "step": 25544
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9196551731293985e-07,
+      "loss": 0.6304,
+      "step": 25545
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.916841929030423e-07,
+      "loss": 0.5807,
+      "step": 25546
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9140300208810203e-07,
+      "loss": 0.5683,
+      "step": 25547
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.91121944871986e-07,
+      "loss": 0.5098,
+      "step": 25548
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.908410212585622e-07,
+      "loss": 0.5607,
+      "step": 25549
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.905602312516986e-07,
+      "loss": 0.6399,
+      "step": 25550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.902795748552578e-07,
+      "loss": 0.5434,
+      "step": 25551
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.899990520731e-07,
+      "loss": 0.5893,
+      "step": 25552
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.897186629090887e-07,
+      "loss": 0.5724,
+      "step": 25553
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8943840736708086e-07,
+      "loss": 0.5579,
+      "step": 25554
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.891582854509323e-07,
+      "loss": 0.5953,
+      "step": 25555
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8887829716449877e-07,
+      "loss": 0.6276,
+      "step": 25556
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.885984425116328e-07,
+      "loss": 0.5599,
+      "step": 25557
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.883187214961858e-07,
+      "loss": 0.5446,
+      "step": 25558
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.880391341220057e-07,
+      "loss": 0.5389,
+      "step": 25559
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.877596803929406e-07,
+      "loss": 0.6575,
+      "step": 25560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8748036031283734e-07,
+      "loss": 0.5609,
+      "step": 25561
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.872011738855385e-07,
+      "loss": 0.6124,
+      "step": 25562
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.869221211148843e-07,
+      "loss": 0.5612,
+      "step": 25563
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.866432020047183e-07,
+      "loss": 0.5911,
+      "step": 25564
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8636441655887637e-07,
+      "loss": 0.5186,
+      "step": 25565
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8608576478119543e-07,
+      "loss": 0.4639,
+      "step": 25566
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.85807246675508e-07,
+      "loss": 0.5566,
+      "step": 25567
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.855288622456498e-07,
+      "loss": 0.552,
+      "step": 25568
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.852506114954501e-07,
+      "loss": 0.55,
+      "step": 25569
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.849724944287391e-07,
+      "loss": 0.6009,
+      "step": 25570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.846945110493415e-07,
+      "loss": 0.6196,
+      "step": 25571
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.844166613610844e-07,
+      "loss": 0.5937,
+      "step": 25572
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8413894536779117e-07,
+      "loss": 0.5613,
+      "step": 25573
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.838613630732823e-07,
+      "loss": 0.497,
+      "step": 25574
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.835839144813801e-07,
+      "loss": 0.5478,
+      "step": 25575
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8330659959589944e-07,
+      "loss": 0.5855,
+      "step": 25576
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8302941842065945e-07,
+      "loss": 0.5576,
+      "step": 25577
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.827523709594704e-07,
+      "loss": 0.5548,
+      "step": 25578
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.824754572161481e-07,
+      "loss": 0.5618,
+      "step": 25579
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.821986771945029e-07,
+      "loss": 0.5638,
+      "step": 25580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.819220308983428e-07,
+      "loss": 0.6099,
+      "step": 25581
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8164551833147256e-07,
+      "loss": 0.5488,
+      "step": 25582
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.813691394977014e-07,
+      "loss": 0.554,
+      "step": 25583
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8109289440082954e-07,
+      "loss": 0.5309,
+      "step": 25584
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.808167830446584e-07,
+      "loss": 0.602,
+      "step": 25585
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8054080543298943e-07,
+      "loss": 0.6393,
+      "step": 25586
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.802649615696196e-07,
+      "loss": 0.5845,
+      "step": 25587
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7998925145834355e-07,
+      "loss": 0.5948,
+      "step": 25588
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.797136751029561e-07,
+      "loss": 0.5779,
+      "step": 25589
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7943823250724977e-07,
+      "loss": 0.5692,
+      "step": 25590
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7916292367501483e-07,
+      "loss": 0.5355,
+      "step": 25591
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7888774861003944e-07,
+      "loss": 0.5426,
+      "step": 25592
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7861270731610936e-07,
+      "loss": 0.5305,
+      "step": 25593
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.783377997970116e-07,
+      "loss": 0.6083,
+      "step": 25594
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.780630260565287e-07,
+      "loss": 0.5828,
+      "step": 25595
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.777883860984387e-07,
+      "loss": 0.5247,
+      "step": 25596
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7751387992652423e-07,
+      "loss": 0.5409,
+      "step": 25597
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.772395075445622e-07,
+      "loss": 0.5415,
+      "step": 25598
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7696526895632737e-07,
+      "loss": 0.5714,
+      "step": 25599
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7669116416559336e-07,
+      "loss": 0.5281,
+      "step": 25600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.764171931761339e-07,
+      "loss": 0.6252,
+      "step": 25601
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7614335599171706e-07,
+      "loss": 0.5828,
+      "step": 25602
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7586965261611086e-07,
+      "loss": 0.5499,
+      "step": 25603
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.755960830530835e-07,
+      "loss": 0.4959,
+      "step": 25604
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.753226473063986e-07,
+      "loss": 0.5189,
+      "step": 25605
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7504934537981865e-07,
+      "loss": 0.5554,
+      "step": 25606
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.747761772771063e-07,
+      "loss": 0.5959,
+      "step": 25607
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.745031430020173e-07,
+      "loss": 0.5465,
+      "step": 25608
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.74230242558311e-07,
+      "loss": 0.5597,
+      "step": 25609
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.739574759497421e-07,
+      "loss": 0.5232,
+      "step": 25610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.736848431800654e-07,
+      "loss": 0.5195,
+      "step": 25611
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.734123442530312e-07,
+      "loss": 0.5397,
+      "step": 25612
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.731399791723888e-07,
+      "loss": 0.5721,
+      "step": 25613
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.728677479418873e-07,
+      "loss": 0.5626,
+      "step": 25614
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.725956505652727e-07,
+      "loss": 0.634,
+      "step": 25615
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7232368704628974e-07,
+      "loss": 0.5765,
+      "step": 25616
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.720518573886799e-07,
+      "loss": 0.5695,
+      "step": 25617
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.717801615961846e-07,
+      "loss": 0.554,
+      "step": 25618
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7150859967254196e-07,
+      "loss": 0.5833,
+      "step": 25619
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.71237171621489e-07,
+      "loss": 0.5424,
+      "step": 25620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.709658774467605e-07,
+      "loss": 0.5615,
+      "step": 25621
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.706947171520913e-07,
+      "loss": 0.5208,
+      "step": 25622
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7042369074121164e-07,
+      "loss": 0.5221,
+      "step": 25623
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.701527982178509e-07,
+      "loss": 0.5774,
+      "step": 25624
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.69882039585736e-07,
+      "loss": 0.5811,
+      "step": 25625
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.696114148485951e-07,
+      "loss": 0.6369,
+      "step": 25626
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6934092401015187e-07,
+      "loss": 0.5404,
+      "step": 25627
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.690705670741256e-07,
+      "loss": 0.5711,
+      "step": 25628
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6880034404423995e-07,
+      "loss": 0.5559,
+      "step": 25629
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.685302549242119e-07,
+      "loss": 0.5454,
+      "step": 25630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.682602997177586e-07,
+      "loss": 0.5918,
+      "step": 25631
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6799047842859247e-07,
+      "loss": 0.4745,
+      "step": 25632
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.677207910604307e-07,
+      "loss": 0.5974,
+      "step": 25633
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.674512376169824e-07,
+      "loss": 0.6008,
+      "step": 25634
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.67181818101957e-07,
+      "loss": 0.5547,
+      "step": 25635
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6691253251906023e-07,
+      "loss": 0.604,
+      "step": 25636
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6664338087200146e-07,
+      "loss": 0.6011,
+      "step": 25637
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6637436316448106e-07,
+      "loss": 0.5538,
+      "step": 25638
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.661054794002016e-07,
+      "loss": 0.5036,
+      "step": 25639
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6583672958286454e-07,
+      "loss": 0.5315,
+      "step": 25640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.65568113716167e-07,
+      "loss": 0.5686,
+      "step": 25641
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.652996318038059e-07,
+      "loss": 0.5292,
+      "step": 25642
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.650312838494751e-07,
+      "loss": 0.5618,
+      "step": 25643
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.647630698568682e-07,
+      "loss": 0.5483,
+      "step": 25644
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.644949898296756e-07,
+      "loss": 0.5583,
+      "step": 25645
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.642270437715844e-07,
+      "loss": 0.5509,
+      "step": 25646
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.63959231686286e-07,
+      "loss": 0.5258,
+      "step": 25647
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6369155357746315e-07,
+      "loss": 0.6052,
+      "step": 25648
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.634240094487994e-07,
+      "loss": 0.5773,
+      "step": 25649
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.631565993039753e-07,
+      "loss": 0.5908,
+      "step": 25650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6288932314667336e-07,
+      "loss": 0.5679,
+      "step": 25651
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6262218098056957e-07,
+      "loss": 0.5872,
+      "step": 25652
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6235517280934097e-07,
+      "loss": 0.5781,
+      "step": 25653
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.620882986366602e-07,
+      "loss": 0.5217,
+      "step": 25654
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6182155846620205e-07,
+      "loss": 0.5909,
+      "step": 25655
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.61554952301637e-07,
+      "loss": 0.663,
+      "step": 25656
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.612884801466298e-07,
+      "loss": 0.5314,
+      "step": 25657
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.610221420048531e-07,
+      "loss": 0.5951,
+      "step": 25658
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6075593787996844e-07,
+      "loss": 0.5446,
+      "step": 25659
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.604898677756396e-07,
+      "loss": 0.5836,
+      "step": 25660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6022393169552687e-07,
+      "loss": 0.5673,
+      "step": 25661
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.599581296432918e-07,
+      "loss": 0.5857,
+      "step": 25662
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5969246162259156e-07,
+      "loss": 0.6547,
+      "step": 25663
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5942692763708197e-07,
+      "loss": 0.5961,
+      "step": 25664
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.591615276904158e-07,
+      "loss": 0.5287,
+      "step": 25665
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.588962617862467e-07,
+      "loss": 0.5543,
+      "step": 25666
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.58631129928224e-07,
+      "loss": 0.5742,
+      "step": 25667
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5836613211999593e-07,
+      "loss": 0.6534,
+      "step": 25668
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5810126836520954e-07,
+      "loss": 0.5217,
+      "step": 25669
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.578365386675108e-07,
+      "loss": 0.545,
+      "step": 25670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.575719430305412e-07,
+      "loss": 0.597,
+      "step": 25671
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5730748145794126e-07,
+      "loss": 0.5492,
+      "step": 25672
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5704315395335245e-07,
+      "loss": 0.5849,
+      "step": 25673
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.567789605204096e-07,
+      "loss": 0.5576,
+      "step": 25674
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.565149011627499e-07,
+      "loss": 0.594,
+      "step": 25675
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5625097588400596e-07,
+      "loss": 0.5126,
+      "step": 25676
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5598718468781146e-07,
+      "loss": 0.6432,
+      "step": 25677
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5572352757779474e-07,
+      "loss": 0.5732,
+      "step": 25678
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.554600045575828e-07,
+      "loss": 0.5407,
+      "step": 25679
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.55196615630805e-07,
+      "loss": 0.5201,
+      "step": 25680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.54933360801084e-07,
+      "loss": 0.5196,
+      "step": 25681
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5467024007204355e-07,
+      "loss": 0.5919,
+      "step": 25682
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.544072534473019e-07,
+      "loss": 0.6263,
+      "step": 25683
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.541444009304805e-07,
+      "loss": 0.6143,
+      "step": 25684
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5388168252519553e-07,
+      "loss": 0.5418,
+      "step": 25685
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.536190982350628e-07,
+      "loss": 0.6138,
+      "step": 25686
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5335664806369397e-07,
+      "loss": 0.5905,
+      "step": 25687
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.530943320147028e-07,
+      "loss": 0.4887,
+      "step": 25688
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.528321500916986e-07,
+      "loss": 0.6248,
+      "step": 25689
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5257010229828735e-07,
+      "loss": 0.6055,
+      "step": 25690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.523081886380763e-07,
+      "loss": 0.5445,
+      "step": 25691
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5204640911467125e-07,
+      "loss": 0.6082,
+      "step": 25692
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5178476373167173e-07,
+      "loss": 0.5399,
+      "step": 25693
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5152325249267916e-07,
+      "loss": 0.5573,
+      "step": 25694
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.51261875401293e-07,
+      "loss": 0.5392,
+      "step": 25695
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5100063246110915e-07,
+      "loss": 0.5678,
+      "step": 25696
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5073952367572265e-07,
+      "loss": 0.5678,
+      "step": 25697
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5047854904872603e-07,
+      "loss": 0.5619,
+      "step": 25698
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.502177085837132e-07,
+      "loss": 0.6121,
+      "step": 25699
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.499570022842701e-07,
+      "loss": 0.5636,
+      "step": 25700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4969643015398613e-07,
+      "loss": 0.5761,
+      "step": 25701
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4943599219644733e-07,
+      "loss": 0.631,
+      "step": 25702
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.491756884152363e-07,
+      "loss": 0.5367,
+      "step": 25703
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4891551881393585e-07,
+      "loss": 0.5532,
+      "step": 25704
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.486554833961252e-07,
+      "loss": 0.5571,
+      "step": 25705
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4839558216538386e-07,
+      "loss": 0.5605,
+      "step": 25706
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.481358151252877e-07,
+      "loss": 0.5598,
+      "step": 25707
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.478761822794118e-07,
+      "loss": 0.6136,
+      "step": 25708
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4761668363132873e-07,
+      "loss": 0.546,
+      "step": 25709
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4735731918460906e-07,
+      "loss": 0.5238,
+      "step": 25710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4709808894282203e-07,
+      "loss": 0.5582,
+      "step": 25711
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.468389929095338e-07,
+      "loss": 0.5304,
+      "step": 25712
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4658003108831264e-07,
+      "loss": 0.5417,
+      "step": 25713
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4632120348272e-07,
+      "loss": 0.5636,
+      "step": 25714
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4606251009631767e-07,
+      "loss": 0.5616,
+      "step": 25715
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4580395093266483e-07,
+      "loss": 0.5538,
+      "step": 25716
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.455455259953221e-07,
+      "loss": 0.5298,
+      "step": 25717
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4528723528784326e-07,
+      "loss": 0.5798,
+      "step": 25718
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4502907881378326e-07,
+      "loss": 0.5649,
+      "step": 25719
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4477105657669477e-07,
+      "loss": 0.5225,
+      "step": 25720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4451316858012833e-07,
+      "loss": 0.5883,
+      "step": 25721
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4425541482763217e-07,
+      "loss": 0.6079,
+      "step": 25722
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.439977953227535e-07,
+      "loss": 0.5385,
+      "step": 25723
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.437403100690372e-07,
+      "loss": 0.5235,
+      "step": 25724
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.434829590700283e-07,
+      "loss": 0.5786,
+      "step": 25725
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.432257423292661e-07,
+      "loss": 0.6459,
+      "step": 25726
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4296865985029007e-07,
+      "loss": 0.5674,
+      "step": 25727
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4271171163663954e-07,
+      "loss": 0.5493,
+      "step": 25728
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.424548976918484e-07,
+      "loss": 0.5671,
+      "step": 25729
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4219821801945156e-07,
+      "loss": 0.6085,
+      "step": 25730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4194167262298176e-07,
+      "loss": 0.5751,
+      "step": 25731
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4168526150596947e-07,
+      "loss": 0.6095,
+      "step": 25732
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.41428984671942e-07,
+      "loss": 0.5842,
+      "step": 25733
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4117284212442413e-07,
+      "loss": 0.5475,
+      "step": 25734
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4091683386694544e-07,
+      "loss": 0.5224,
+      "step": 25735
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.406609599030252e-07,
+      "loss": 0.5573,
+      "step": 25736
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.404052202361862e-07,
+      "loss": 0.6178,
+      "step": 25737
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.401496148699456e-07,
+      "loss": 0.6057,
+      "step": 25738
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3989414380782285e-07,
+      "loss": 0.5966,
+      "step": 25739
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3963880705333285e-07,
+      "loss": 0.5216,
+      "step": 25740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3938360460998846e-07,
+      "loss": 0.6006,
+      "step": 25741
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.391285364813023e-07,
+      "loss": 0.5773,
+      "step": 25742
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.38873602670785e-07,
+      "loss": 0.5434,
+      "step": 25743
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.386188031819425e-07,
+      "loss": 0.5225,
+      "step": 25744
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3836413801828217e-07,
+      "loss": 0.6106,
+      "step": 25745
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3810960718331e-07,
+      "loss": 0.5159,
+      "step": 25746
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3785521068052652e-07,
+      "loss": 0.5598,
+      "step": 25747
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.376009485134334e-07,
+      "loss": 0.5759,
+      "step": 25748
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3734682068552893e-07,
+      "loss": 0.5922,
+      "step": 25749
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3709282720031035e-07,
+      "loss": 0.6235,
+      "step": 25750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.368389680612737e-07,
+      "loss": 0.5309,
+      "step": 25751
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3658524327190957e-07,
+      "loss": 0.569,
+      "step": 25752
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3633165283571292e-07,
+      "loss": 0.5668,
+      "step": 25753
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3607819675617204e-07,
+      "loss": 0.5775,
+      "step": 25754
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.358248750367742e-07,
+      "loss": 0.5556,
+      "step": 25755
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3557168768100548e-07,
+      "loss": 0.6467,
+      "step": 25756
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3531863469234974e-07,
+      "loss": 0.5277,
+      "step": 25757
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3506571607429085e-07,
+      "loss": 0.5986,
+      "step": 25758
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3481293183030718e-07,
+      "loss": 0.5146,
+      "step": 25759
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3456028196387703e-07,
+      "loss": 0.5745,
+      "step": 25760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3430776647847876e-07,
+      "loss": 0.5425,
+      "step": 25761
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.340553853775873e-07,
+      "loss": 0.5931,
+      "step": 25762
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.338031386646733e-07,
+      "loss": 0.5673,
+      "step": 25763
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.335510263432106e-07,
+      "loss": 0.5682,
+      "step": 25764
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3329904841666752e-07,
+      "loss": 0.5959,
+      "step": 25765
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3304720488851128e-07,
+      "loss": 0.5914,
+      "step": 25766
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3279549576220584e-07,
+      "loss": 0.5193,
+      "step": 25767
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3254392104121837e-07,
+      "loss": 0.5603,
+      "step": 25768
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3229248072900835e-07,
+      "loss": 0.5788,
+      "step": 25769
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3204117482903633e-07,
+      "loss": 0.5943,
+      "step": 25770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3179000334476064e-07,
+      "loss": 0.5275,
+      "step": 25771
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3153896627963856e-07,
+      "loss": 0.6126,
+      "step": 25772
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3128806363712287e-07,
+      "loss": 0.5987,
+      "step": 25773
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3103729542066633e-07,
+      "loss": 0.5542,
+      "step": 25774
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3078666163372176e-07,
+      "loss": 0.5199,
+      "step": 25775
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.305361622797353e-07,
+      "loss": 0.5619,
+      "step": 25776
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3028579736215638e-07,
+      "loss": 0.5535,
+      "step": 25777
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3003556688442897e-07,
+      "loss": 0.6215,
+      "step": 25778
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2978547084999693e-07,
+      "loss": 0.5846,
+      "step": 25779
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2953550926230195e-07,
+      "loss": 0.6147,
+      "step": 25780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2928568212478242e-07,
+      "loss": 0.5858,
+      "step": 25781
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.290359894408778e-07,
+      "loss": 0.6404,
+      "step": 25782
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.287864312140242e-07,
+      "loss": 0.5601,
+      "step": 25783
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.285370074476545e-07,
+      "loss": 0.5669,
+      "step": 25784
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2828771814520145e-07,
+      "loss": 0.5779,
+      "step": 25785
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2803856331009566e-07,
+      "loss": 0.5792,
+      "step": 25786
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.277895429457655e-07,
+      "loss": 0.5859,
+      "step": 25787
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2754065705563822e-07,
+      "loss": 0.5345,
+      "step": 25788
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2729190564313774e-07,
+      "loss": 0.595,
+      "step": 25789
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.27043288711688e-07,
+      "loss": 0.5893,
+      "step": 25790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2679480626470963e-07,
+      "loss": 0.5337,
+      "step": 25791
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2654645830562205e-07,
+      "loss": 0.5694,
+      "step": 25792
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.262982448378437e-07,
+      "loss": 0.5582,
+      "step": 25793
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2605016586478957e-07,
+      "loss": 0.614,
+      "step": 25794
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2580222138987252e-07,
+      "loss": 0.5674,
+      "step": 25795
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2555441141650536e-07,
+      "loss": 0.5474,
+      "step": 25796
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.253067359480987e-07,
+      "loss": 0.5212,
+      "step": 25797
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2505919498805985e-07,
+      "loss": 0.5068,
+      "step": 25798
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.248117885397949e-07,
+      "loss": 0.5523,
+      "step": 25799
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2456451660670897e-07,
+      "loss": 0.6021,
+      "step": 25800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2431737919220597e-07,
+      "loss": 0.5548,
+      "step": 25801
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2407037629968431e-07,
+      "loss": 0.5094,
+      "step": 25802
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.238235079325446e-07,
+      "loss": 0.561,
+      "step": 25803
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.23576774094183e-07,
+      "loss": 0.6214,
+      "step": 25804
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.233301747879968e-07,
+      "loss": 0.5623,
+      "step": 25805
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.230837100173766e-07,
+      "loss": 0.5446,
+      "step": 25806
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.228373797857153e-07,
+      "loss": 0.5295,
+      "step": 25807
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2259118409640235e-07,
+      "loss": 0.5812,
+      "step": 25808
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.223451229528273e-07,
+      "loss": 0.5857,
+      "step": 25809
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2209919635837406e-07,
+      "loss": 0.5771,
+      "step": 25810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2185340431642666e-07,
+      "loss": 0.5332,
+      "step": 25811
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2160774683036902e-07,
+      "loss": 0.5305,
+      "step": 25812
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2136222390358175e-07,
+      "loss": 0.5293,
+      "step": 25813
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2111683553943996e-07,
+      "loss": 0.6067,
+      "step": 25814
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.208715817413254e-07,
+      "loss": 0.6024,
+      "step": 25815
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2062646251260977e-07,
+      "loss": 0.5939,
+      "step": 25816
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2038147785666708e-07,
+      "loss": 0.5492,
+      "step": 25817
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2013662777686906e-07,
+      "loss": 0.5642,
+      "step": 25818
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.19891912276583e-07,
+      "loss": 0.5866,
+      "step": 25819
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.196473313591796e-07,
+      "loss": 0.5447,
+      "step": 25820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1940288502802054e-07,
+      "loss": 0.5483,
+      "step": 25821
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1915857328647205e-07,
+      "loss": 0.6124,
+      "step": 25822
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.18914396137897e-07,
+      "loss": 0.5521,
+      "step": 25823
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1867035358565268e-07,
+      "loss": 0.5603,
+      "step": 25824
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1842644563310088e-07,
+      "loss": 0.5811,
+      "step": 25825
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1818267228359557e-07,
+      "loss": 0.5598,
+      "step": 25826
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1793903354049074e-07,
+      "loss": 0.6307,
+      "step": 25827
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1769552940714034e-07,
+      "loss": 0.5159,
+      "step": 25828
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1745215988689506e-07,
+      "loss": 0.5377,
+      "step": 25829
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.172089249831044e-07,
+      "loss": 0.5262,
+      "step": 25830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1696582469911464e-07,
+      "loss": 0.5832,
+      "step": 25831
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1672285903826974e-07,
+      "loss": 0.4728,
+      "step": 25832
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1648002800391587e-07,
+      "loss": 0.5481,
+      "step": 25833
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1623733159939375e-07,
+      "loss": 0.5937,
+      "step": 25834
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1599476982804179e-07,
+      "loss": 0.5398,
+      "step": 25835
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1575234269319957e-07,
+      "loss": 0.5923,
+      "step": 25836
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1551005019820213e-07,
+      "loss": 0.58,
+      "step": 25837
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1526789234638467e-07,
+      "loss": 0.5423,
+      "step": 25838
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.150258691410767e-07,
+      "loss": 0.6374,
+      "step": 25839
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1478398058561222e-07,
+      "loss": 0.6138,
+      "step": 25840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1454222668331748e-07,
+      "loss": 0.5793,
+      "step": 25841
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.143006074375198e-07,
+      "loss": 0.5968,
+      "step": 25842
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.140591228515443e-07,
+      "loss": 0.64,
+      "step": 25843
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1381777292871385e-07,
+      "loss": 0.5752,
+      "step": 25844
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.135765576723503e-07,
+      "loss": 0.5522,
+      "step": 25845
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1333547708577207e-07,
+      "loss": 0.5347,
+      "step": 25846
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1309453117229652e-07,
+      "loss": 0.5602,
+      "step": 25847
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1285371993524097e-07,
+      "loss": 0.5651,
+      "step": 25848
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1261304337791722e-07,
+      "loss": 0.609,
+      "step": 25849
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.123725015036371e-07,
+      "loss": 0.5465,
+      "step": 25850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1213209431571346e-07,
+      "loss": 0.554,
+      "step": 25851
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.118918218174515e-07,
+      "loss": 0.6061,
+      "step": 25852
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.116516840121585e-07,
+      "loss": 0.5802,
+      "step": 25853
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1141168090313857e-07,
+      "loss": 0.5912,
+      "step": 25854
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.111718124936968e-07,
+      "loss": 0.5854,
+      "step": 25855
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.109320787871305e-07,
+      "loss": 0.6287,
+      "step": 25856
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1069247978674046e-07,
+      "loss": 0.4687,
+      "step": 25857
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1045301549582398e-07,
+      "loss": 0.5403,
+      "step": 25858
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.102136859176762e-07,
+      "loss": 0.5999,
+      "step": 25859
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0997449105558898e-07,
+      "loss": 0.5483,
+      "step": 25860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.097354309128552e-07,
+      "loss": 0.5435,
+      "step": 25861
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0949650549276556e-07,
+      "loss": 0.5468,
+      "step": 25862
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0925771479860635e-07,
+      "loss": 0.5373,
+      "step": 25863
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.090190588336627e-07,
+      "loss": 0.5703,
+      "step": 25864
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0878053760122085e-07,
+      "loss": 0.5972,
+      "step": 25865
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0854215110456265e-07,
+      "loss": 0.5364,
+      "step": 25866
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.083038993469666e-07,
+      "loss": 0.6072,
+      "step": 25867
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0806578233171338e-07,
+      "loss": 0.6139,
+      "step": 25868
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0782780006207925e-07,
+      "loss": 0.549,
+      "step": 25869
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0758995254133829e-07,
+      "loss": 0.6091,
+      "step": 25870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0735223977276453e-07,
+      "loss": 0.607,
+      "step": 25871
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.071146617596276e-07,
+      "loss": 0.5422,
+      "step": 25872
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0687721850519814e-07,
+      "loss": 0.6023,
+      "step": 25873
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0663991001274363e-07,
+      "loss": 0.5946,
+      "step": 25874
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.064027362855281e-07,
+      "loss": 0.5114,
+      "step": 25875
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0616569732681669e-07,
+      "loss": 0.5414,
+      "step": 25876
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0592879313987125e-07,
+      "loss": 0.582,
+      "step": 25877
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0569202372795138e-07,
+      "loss": 0.5886,
+      "step": 25878
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0545538909431452e-07,
+      "loss": 0.5511,
+      "step": 25879
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.052188892422191e-07,
+      "loss": 0.5644,
+      "step": 25880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0498252417491815e-07,
+      "loss": 0.572,
+      "step": 25881
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0474629389566347e-07,
+      "loss": 0.5738,
+      "step": 25882
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0451019840770692e-07,
+      "loss": 0.5593,
+      "step": 25883
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.04274237714297e-07,
+      "loss": 0.5047,
+      "step": 25884
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.040384118186811e-07,
+      "loss": 0.5504,
+      "step": 25885
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0380272072410335e-07,
+      "loss": 0.615,
+      "step": 25886
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0356716443380885e-07,
+      "loss": 0.5704,
+      "step": 25887
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.033317429510384e-07,
+      "loss": 0.5777,
+      "step": 25888
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0309645627903051e-07,
+      "loss": 0.5638,
+      "step": 25889
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.028613044210237e-07,
+      "loss": 0.5601,
+      "step": 25890
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0262628738025314e-07,
+      "loss": 0.6065,
+      "step": 25891
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0239140515995515e-07,
+      "loss": 0.5801,
+      "step": 25892
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0215665776335935e-07,
+      "loss": 0.5584,
+      "step": 25893
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0192204519369762e-07,
+      "loss": 0.5687,
+      "step": 25894
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0168756745419737e-07,
+      "loss": 0.5446,
+      "step": 25895
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0145322454808603e-07,
+      "loss": 0.5618,
+      "step": 25896
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0121901647858765e-07,
+      "loss": 0.5935,
+      "step": 25897
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0098494324892636e-07,
+      "loss": 0.6382,
+      "step": 25898
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.007510048623218e-07,
+      "loss": 0.5673,
+      "step": 25899
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0051720132199471e-07,
+      "loss": 0.5131,
+      "step": 25900
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.002835326311603e-07,
+      "loss": 0.5545,
+      "step": 25901
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.00049998793036e-07,
+      "loss": 0.5734,
+      "step": 25902
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9981659981083478e-07,
+      "loss": 0.5808,
+      "step": 25903
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9958333568776745e-07,
+      "loss": 0.5783,
+      "step": 25904
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.993502064270447e-07,
+      "loss": 0.5368,
+      "step": 25905
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9911721203187628e-07,
+      "loss": 0.561,
+      "step": 25906
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9888435250546623e-07,
+      "loss": 0.6014,
+      "step": 25907
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9865162785101866e-07,
+      "loss": 0.5364,
+      "step": 25908
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9841903807173769e-07,
+      "loss": 0.5834,
+      "step": 25909
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9818658317082296e-07,
+      "loss": 0.5984,
+      "step": 25910
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9795426315147414e-07,
+      "loss": 0.5708,
+      "step": 25911
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9772207801688647e-07,
+      "loss": 0.6133,
+      "step": 25912
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.974900277702563e-07,
+      "loss": 0.5565,
+      "step": 25913
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9725811241477654e-07,
+      "loss": 0.5582,
+      "step": 25914
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9702633195363918e-07,
+      "loss": 0.6086,
+      "step": 25915
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9679468639003274e-07,
+      "loss": 0.5865,
+      "step": 25916
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9656317572714468e-07,
+      "loss": 0.6019,
+      "step": 25917
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9633179996816242e-07,
+      "loss": 0.568,
+      "step": 25918
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.961005591162679e-07,
+      "loss": 0.5266,
+      "step": 25919
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9586945317464522e-07,
+      "loss": 0.5977,
+      "step": 25920
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9563848214647296e-07,
+      "loss": 0.4936,
+      "step": 25921
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9540764603492966e-07,
+      "loss": 0.5803,
+      "step": 25922
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9517694484319283e-07,
+      "loss": 0.5615,
+      "step": 25923
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9494637857443543e-07,
+      "loss": 0.5539,
+      "step": 25924
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9471594723183273e-07,
+      "loss": 0.5772,
+      "step": 25925
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.944856508185533e-07,
+      "loss": 0.5271,
+      "step": 25926
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9425548933776796e-07,
+      "loss": 0.6442,
+      "step": 25927
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.94025462792643e-07,
+      "loss": 0.5432,
+      "step": 25928
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9379557118634373e-07,
+      "loss": 0.5077,
+      "step": 25929
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9356581452203317e-07,
+      "loss": 0.6,
+      "step": 25930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.933361928028754e-07,
+      "loss": 0.5588,
+      "step": 25931
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9310670603202796e-07,
+      "loss": 0.5502,
+      "step": 25932
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.928773542126483e-07,
+      "loss": 0.5455,
+      "step": 25933
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.926481373478939e-07,
+      "loss": 0.6494,
+      "step": 25934
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9241905544091888e-07,
+      "loss": 0.5308,
+      "step": 25935
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.921901084948763e-07,
+      "loss": 0.5451,
+      "step": 25936
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9196129651291362e-07,
+      "loss": 0.5475,
+      "step": 25937
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9173261949818278e-07,
+      "loss": 0.5428,
+      "step": 25938
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9150407745383015e-07,
+      "loss": 0.5372,
+      "step": 25939
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9127567038299878e-07,
+      "loss": 0.6009,
+      "step": 25940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9104739828883277e-07,
+      "loss": 0.5175,
+      "step": 25941
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9081926117447413e-07,
+      "loss": 0.5565,
+      "step": 25942
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.905912590430614e-07,
+      "loss": 0.595,
+      "step": 25943
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9036339189773323e-07,
+      "loss": 0.5342,
+      "step": 25944
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9013565974162262e-07,
+      "loss": 0.6035,
+      "step": 25945
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.89908062577866e-07,
+      "loss": 0.5215,
+      "step": 25946
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8968060040959413e-07,
+      "loss": 0.5343,
+      "step": 25947
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8945327323993678e-07,
+      "loss": 0.5616,
+      "step": 25948
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8922608107202256e-07,
+      "loss": 0.5518,
+      "step": 25949
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8899902390897894e-07,
+      "loss": 0.6348,
+      "step": 25950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.88772101753929e-07,
+      "loss": 0.5938,
+      "step": 25951
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.885453146099958e-07,
+      "loss": 0.5535,
+      "step": 25952
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8831866248030018e-07,
+      "loss": 0.6258,
+      "step": 25953
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8809214536796184e-07,
+      "loss": 0.5891,
+      "step": 25954
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8786576327609606e-07,
+      "loss": 0.5215,
+      "step": 25955
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.876395162078193e-07,
+      "loss": 0.6272,
+      "step": 25956
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8741340416624453e-07,
+      "loss": 0.6328,
+      "step": 25957
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8718742715448378e-07,
+      "loss": 0.549,
+      "step": 25958
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8696158517564566e-07,
+      "loss": 0.5375,
+      "step": 25959
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.867358782328399e-07,
+      "loss": 0.577,
+      "step": 25960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8651030632917065e-07,
+      "loss": 0.6257,
+      "step": 25961
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8628486946774216e-07,
+      "loss": 0.6131,
+      "step": 25962
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8605956765165633e-07,
+      "loss": 0.535,
+      "step": 25963
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8583440088401516e-07,
+      "loss": 0.5203,
+      "step": 25964
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8560936916791618e-07,
+      "loss": 0.5112,
+      "step": 25965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8538447250645575e-07,
+      "loss": 0.6351,
+      "step": 25966
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.851597109027281e-07,
+      "loss": 0.5444,
+      "step": 25967
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8493508435982743e-07,
+      "loss": 0.5201,
+      "step": 25968
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.847105928808446e-07,
+      "loss": 0.608,
+      "step": 25969
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.844862364688671e-07,
+      "loss": 0.564,
+      "step": 25970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8426201512698582e-07,
+      "loss": 0.5589,
+      "step": 25971
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8403792885828276e-07,
+      "loss": 0.5413,
+      "step": 25972
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8381397766584318e-07,
+      "loss": 0.5892,
+      "step": 25973
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.83590161552748e-07,
+      "loss": 0.5834,
+      "step": 25974
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8336648052207807e-07,
+      "loss": 0.5428,
+      "step": 25975
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.831429345769109e-07,
+      "loss": 0.5757,
+      "step": 25976
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8291952372032406e-07,
+      "loss": 0.5806,
+      "step": 25977
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8269624795538843e-07,
+      "loss": 0.5454,
+      "step": 25978
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8247310728518042e-07,
+      "loss": 0.5665,
+      "step": 25979
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.822501017127676e-07,
+      "loss": 0.5632,
+      "step": 25980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8202723124122078e-07,
+      "loss": 0.5884,
+      "step": 25981
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8180449587360538e-07,
+      "loss": 0.6578,
+      "step": 25982
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8158189561298777e-07,
+      "loss": 0.5706,
+      "step": 25983
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.813594304624311e-07,
+      "loss": 0.5609,
+      "step": 25984
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.811371004249951e-07,
+      "loss": 0.5698,
+      "step": 25985
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8091490550374069e-07,
+      "loss": 0.5981,
+      "step": 25986
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.806928457017243e-07,
+      "loss": 0.5832,
+      "step": 25987
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8047092102200238e-07,
+      "loss": 0.543,
+      "step": 25988
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8024913146762912e-07,
+      "loss": 0.5378,
+      "step": 25989
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8002747704165656e-07,
+      "loss": 0.5833,
+      "step": 25990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.798059577471345e-07,
+      "loss": 0.5674,
+      "step": 25991
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7958457358711045e-07,
+      "loss": 0.5883,
+      "step": 25992
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.79363324564632e-07,
+      "loss": 0.633,
+      "step": 25993
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.791422106827445e-07,
+      "loss": 0.5535,
+      "step": 25994
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7892123194448886e-07,
+      "loss": 0.5718,
+      "step": 25995
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7870038835290594e-07,
+      "loss": 0.5521,
+      "step": 25996
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7847967991103555e-07,
+      "loss": 0.5869,
+      "step": 25997
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7825910662191525e-07,
+      "loss": 0.5652,
+      "step": 25998
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.780386684885793e-07,
+      "loss": 0.6072,
+      "step": 25999
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7781836551406195e-07,
+      "loss": 0.6098,
+      "step": 26000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7759819770139408e-07,
+      "loss": 0.5728,
+      "step": 26001
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.773781650536066e-07,
+      "loss": 0.5602,
+      "step": 26002
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.771582675737249e-07,
+      "loss": 0.4803,
+      "step": 26003
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7693850526477763e-07,
+      "loss": 0.5788,
+      "step": 26004
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7671887812978794e-07,
+      "loss": 0.5728,
+      "step": 26005
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7649938617177785e-07,
+      "loss": 0.5612,
+      "step": 26006
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.762800293937672e-07,
+      "loss": 0.5049,
+      "step": 26007
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7606080779877578e-07,
+      "loss": 0.5792,
+      "step": 26008
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7584172138982003e-07,
+      "loss": 0.5266,
+      "step": 26009
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7562277016991314e-07,
+      "loss": 0.5706,
+      "step": 26010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7540395414207045e-07,
+      "loss": 0.5525,
+      "step": 26011
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.751852733093029e-07,
+      "loss": 0.5584,
+      "step": 26012
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.749667276746181e-07,
+      "loss": 0.5766,
+      "step": 26013
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.747483172410225e-07,
+      "loss": 0.5514,
+      "step": 26014
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.745300420115259e-07,
+      "loss": 0.5795,
+      "step": 26015
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7431190198912817e-07,
+      "loss": 0.5383,
+      "step": 26016
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7409389717683244e-07,
+      "loss": 0.5178,
+      "step": 26017
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7387602757763743e-07,
+      "loss": 0.6189,
+      "step": 26018
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7365829319454296e-07,
+      "loss": 0.496,
+      "step": 26019
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7344069403054554e-07,
+      "loss": 0.6353,
+      "step": 26020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7322323008863607e-07,
+      "loss": 0.5988,
+      "step": 26021
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7300590137181217e-07,
+      "loss": 0.5769,
+      "step": 26022
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7278870788306035e-07,
+      "loss": 0.5464,
+      "step": 26023
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7257164962537153e-07,
+      "loss": 0.5628,
+      "step": 26024
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7235472660173114e-07,
+      "loss": 0.5645,
+      "step": 26025
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7213793881512563e-07,
+      "loss": 0.6054,
+      "step": 26026
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7192128626853822e-07,
+      "loss": 0.6305,
+      "step": 26027
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.717047689649487e-07,
+      "loss": 0.5937,
+      "step": 26028
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7148838690733804e-07,
+      "loss": 0.5591,
+      "step": 26029
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7127214009868387e-07,
+      "loss": 0.5335,
+      "step": 26030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7105602854195936e-07,
+      "loss": 0.5805,
+      "step": 26031
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7084005224014212e-07,
+      "loss": 0.5841,
+      "step": 26032
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7062421119620198e-07,
+      "loss": 0.613,
+      "step": 26033
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7040850541310883e-07,
+      "loss": 0.5449,
+      "step": 26034
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7019293489383138e-07,
+      "loss": 0.6064,
+      "step": 26035
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6997749964133724e-07,
+      "loss": 0.5856,
+      "step": 26036
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6976219965859076e-07,
+      "loss": 0.5544,
+      "step": 26037
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6954703494855286e-07,
+      "loss": 0.5209,
+      "step": 26038
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.693320055141856e-07,
+      "loss": 0.6022,
+      "step": 26039
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6911711135844888e-07,
+      "loss": 0.5975,
+      "step": 26040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.689023524842981e-07,
+      "loss": 0.5993,
+      "step": 26041
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6868772889468866e-07,
+      "loss": 0.5279,
+      "step": 26042
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6847324059257596e-07,
+      "loss": 0.5516,
+      "step": 26043
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6825888758090992e-07,
+      "loss": 0.578,
+      "step": 26044
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6804466986264033e-07,
+      "loss": 0.5604,
+      "step": 26045
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.678305874407149e-07,
+      "loss": 0.636,
+      "step": 26046
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6761664031808012e-07,
+      "loss": 0.5673,
+      "step": 26047
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6740282849768031e-07,
+      "loss": 0.5456,
+      "step": 26048
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6718915198245644e-07,
+      "loss": 0.5589,
+      "step": 26049
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6697561077534952e-07,
+      "loss": 0.5733,
+      "step": 26050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.667622048792994e-07,
+      "loss": 0.5404,
+      "step": 26051
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.665489342972415e-07,
+      "loss": 0.5548,
+      "step": 26052
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6633579903210906e-07,
+      "loss": 0.5648,
+      "step": 26053
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6612279908683858e-07,
+      "loss": 0.5399,
+      "step": 26054
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6590993446435776e-07,
+      "loss": 0.5756,
+      "step": 26055
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6569720516759868e-07,
+      "loss": 0.6032,
+      "step": 26056
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6548461119948568e-07,
+      "loss": 0.574,
+      "step": 26057
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6527215256294638e-07,
+      "loss": 0.5418,
+      "step": 26058
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6505982926090403e-07,
+      "loss": 0.586,
+      "step": 26059
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6484764129628073e-07,
+      "loss": 0.5214,
+      "step": 26060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6463558867199415e-07,
+      "loss": 0.5492,
+      "step": 26061
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6442367139096528e-07,
+      "loss": 0.6215,
+      "step": 26062
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6421188945610843e-07,
+      "loss": 0.6103,
+      "step": 26063
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.64000242870338e-07,
+      "loss": 0.5543,
+      "step": 26064
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6378873163656828e-07,
+      "loss": 0.5475,
+      "step": 26065
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6357735575770806e-07,
+      "loss": 0.6077,
+      "step": 26066
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6336611523666612e-07,
+      "loss": 0.5262,
+      "step": 26067
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6315501007634905e-07,
+      "loss": 0.5585,
+      "step": 26068
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6294404027966338e-07,
+      "loss": 0.5591,
+      "step": 26069
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6273320584951124e-07,
+      "loss": 0.5409,
+      "step": 26070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6252250678879368e-07,
+      "loss": 0.4757,
+      "step": 26071
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6231194310041055e-07,
+      "loss": 0.532,
+      "step": 26072
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6210151478725954e-07,
+      "loss": 0.6074,
+      "step": 26073
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6189122185223504e-07,
+      "loss": 0.5993,
+      "step": 26074
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.616810642982325e-07,
+      "loss": 0.5988,
+      "step": 26075
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.614710421281429e-07,
+      "loss": 0.5962,
+      "step": 26076
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6126115534485733e-07,
+      "loss": 0.5673,
+      "step": 26077
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.610514039512634e-07,
+      "loss": 0.5362,
+      "step": 26078
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6084178795024664e-07,
+      "loss": 0.6126,
+      "step": 26079
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.606323073446925e-07,
+      "loss": 0.5686,
+      "step": 26080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6042296213748308e-07,
+      "loss": 0.6078,
+      "step": 26081
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.602137523315006e-07,
+      "loss": 0.5078,
+      "step": 26082
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.600046779296216e-07,
+      "loss": 0.5498,
+      "step": 26083
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5979573893472488e-07,
+      "loss": 0.5654,
+      "step": 26084
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5958693534968595e-07,
+      "loss": 0.6117,
+      "step": 26085
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5937826717737582e-07,
+      "loss": 0.555,
+      "step": 26086
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5916973442066886e-07,
+      "loss": 0.5471,
+      "step": 26087
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5896133708243276e-07,
+      "loss": 0.6059,
+      "step": 26088
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.587530751655364e-07,
+      "loss": 0.5134,
+      "step": 26089
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5854494867284408e-07,
+      "loss": 0.5588,
+      "step": 26090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5833695760722135e-07,
+      "loss": 0.487,
+      "step": 26091
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.581291019715292e-07,
+      "loss": 0.5728,
+      "step": 26092
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.579213817686276e-07,
+      "loss": 0.5758,
+      "step": 26093
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5771379700137758e-07,
+      "loss": 0.6379,
+      "step": 26094
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5750634767263352e-07,
+      "loss": 0.6298,
+      "step": 26095
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.572990337852498e-07,
+      "loss": 0.5899,
+      "step": 26096
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.570918553420797e-07,
+      "loss": 0.6019,
+      "step": 26097
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5688481234597542e-07,
+      "loss": 0.4872,
+      "step": 26098
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5667790479978462e-07,
+      "loss": 0.5746,
+      "step": 26099
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5647113270635504e-07,
+      "loss": 0.5716,
+      "step": 26100
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.562644960685311e-07,
+      "loss": 0.6015,
+      "step": 26101
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.560579948891583e-07,
+      "loss": 0.5246,
+      "step": 26102
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5585162917107656e-07,
+      "loss": 0.5371,
+      "step": 26103
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5564539891712583e-07,
+      "loss": 0.5311,
+      "step": 26104
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5543930413014496e-07,
+      "loss": 0.5149,
+      "step": 26105
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5523334481297058e-07,
+      "loss": 0.5498,
+      "step": 26106
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5502752096843488e-07,
+      "loss": 0.5992,
+      "step": 26107
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5482183259937e-07,
+      "loss": 0.5778,
+      "step": 26108
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5461627970860814e-07,
+      "loss": 0.5365,
+      "step": 26109
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5441086229897706e-07,
+      "loss": 0.6204,
+      "step": 26110
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5420558037330448e-07,
+      "loss": 0.6172,
+      "step": 26111
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5400043393441256e-07,
+      "loss": 0.5441,
+      "step": 26112
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5379542298512796e-07,
+      "loss": 0.581,
+      "step": 26113
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5359054752826953e-07,
+      "loss": 0.5496,
+      "step": 26114
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.533858075666561e-07,
+      "loss": 0.5948,
+      "step": 26115
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.531812031031066e-07,
+      "loss": 0.6128,
+      "step": 26116
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5297673414043644e-07,
+      "loss": 0.5507,
+      "step": 26117
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5277240068145795e-07,
+      "loss": 0.5385,
+      "step": 26118
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.525682027289843e-07,
+      "loss": 0.5634,
+      "step": 26119
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5236414028582447e-07,
+      "loss": 0.5494,
+      "step": 26120
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5216021335478725e-07,
+      "loss": 0.622,
+      "step": 26121
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.519564219386782e-07,
+      "loss": 0.5461,
+      "step": 26122
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5175276604030177e-07,
+      "loss": 0.5513,
+      "step": 26123
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5154924566246232e-07,
+      "loss": 0.5816,
+      "step": 26124
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5134586080795766e-07,
+      "loss": 0.5724,
+      "step": 26125
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5114261147958665e-07,
+      "loss": 0.5566,
+      "step": 26126
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5093949768014925e-07,
+      "loss": 0.6232,
+      "step": 26127
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5073651941243772e-07,
+      "loss": 0.5895,
+      "step": 26128
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5053367667924645e-07,
+      "loss": 0.5514,
+      "step": 26129
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5033096948336546e-07,
+      "loss": 0.5384,
+      "step": 26130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5012839782758472e-07,
+      "loss": 0.5963,
+      "step": 26131
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4992596171469308e-07,
+      "loss": 0.5198,
+      "step": 26132
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4972366114747504e-07,
+      "loss": 0.5775,
+      "step": 26133
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4952149612871391e-07,
+      "loss": 0.5985,
+      "step": 26134
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.49319466661193e-07,
+      "loss": 0.5877,
+      "step": 26135
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4911757274769235e-07,
+      "loss": 0.5032,
+      "step": 26136
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4891581439098858e-07,
+      "loss": 0.6491,
+      "step": 26137
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4871419159385948e-07,
+      "loss": 0.5672,
+      "step": 26138
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.485127043590795e-07,
+      "loss": 0.566,
+      "step": 26139
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4831135268942088e-07,
+      "loss": 0.5999,
+      "step": 26140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4811013658765471e-07,
+      "loss": 0.6234,
+      "step": 26141
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4790905605654882e-07,
+      "loss": 0.5217,
+      "step": 26142
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4770811109887206e-07,
+      "loss": 0.6026,
+      "step": 26143
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4750730171738892e-07,
+      "loss": 0.5135,
+      "step": 26144
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.473066279148616e-07,
+      "loss": 0.5844,
+      "step": 26145
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.471060896940524e-07,
+      "loss": 0.6226,
+      "step": 26146
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4690568705772123e-07,
+      "loss": 0.5682,
+      "step": 26147
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4670542000862486e-07,
+      "loss": 0.6047,
+      "step": 26148
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4650528854952105e-07,
+      "loss": 0.5878,
+      "step": 26149
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4630529268316207e-07,
+      "loss": 0.5749,
+      "step": 26150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4610543241230012e-07,
+      "loss": 0.6196,
+      "step": 26151
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4590570773968526e-07,
+      "loss": 0.5665,
+      "step": 26152
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.457061186680675e-07,
+      "loss": 0.5848,
+      "step": 26153
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.455066652001924e-07,
+      "loss": 0.5683,
+      "step": 26154
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4530734733880337e-07,
+      "loss": 0.6171,
+      "step": 26155
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4510816508664482e-07,
+      "loss": 0.5606,
+      "step": 26156
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4490911844645793e-07,
+      "loss": 0.5429,
+      "step": 26157
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4471020742098052e-07,
+      "loss": 0.5416,
+      "step": 26158
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4451143201294925e-07,
+      "loss": 0.6442,
+      "step": 26159
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4431279222510196e-07,
+      "loss": 0.5906,
+      "step": 26160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.441142880601698e-07,
+      "loss": 0.5374,
+      "step": 26161
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4391591952088503e-07,
+      "loss": 0.5481,
+      "step": 26162
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4371768660997764e-07,
+      "loss": 0.523,
+      "step": 26163
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.435195893301755e-07,
+      "loss": 0.6505,
+      "step": 26164
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4332162768420533e-07,
+      "loss": 0.5221,
+      "step": 26165
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4312380167478935e-07,
+      "loss": 0.571,
+      "step": 26166
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4292611130465096e-07,
+      "loss": 0.6112,
+      "step": 26167
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.427285565765113e-07,
+      "loss": 0.6072,
+      "step": 26168
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4253113749308712e-07,
+      "loss": 0.5619,
+      "step": 26169
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.423338540570962e-07,
+      "loss": 0.5167,
+      "step": 26170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4213670627125308e-07,
+      "loss": 0.5361,
+      "step": 26171
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4193969413827113e-07,
+      "loss": 0.5542,
+      "step": 26172
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.417428176608615e-07,
+      "loss": 0.5445,
+      "step": 26173
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4154607684173094e-07,
+      "loss": 0.5055,
+      "step": 26174
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.413494716835906e-07,
+      "loss": 0.5263,
+      "step": 26175
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4115300218914386e-07,
+      "loss": 0.5762,
+      "step": 26176
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.409566683610941e-07,
+      "loss": 0.6186,
+      "step": 26177
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4076047020214368e-07,
+      "loss": 0.5875,
+      "step": 26178
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.405644077149926e-07,
+      "loss": 0.569,
+      "step": 26179
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.403684809023387e-07,
+      "loss": 0.5693,
+      "step": 26180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4017268976687648e-07,
+      "loss": 0.5116,
+      "step": 26181
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3997703431130271e-07,
+      "loss": 0.6293,
+      "step": 26182
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3978151453830856e-07,
+      "loss": 0.5728,
+      "step": 26183
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3958613045058522e-07,
+      "loss": 0.5672,
+      "step": 26184
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3939088205082052e-07,
+      "loss": 0.5296,
+      "step": 26185
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.391957693417012e-07,
+      "loss": 0.5452,
+      "step": 26186
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.390007923259129e-07,
+      "loss": 0.5329,
+      "step": 26187
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.388059510061379e-07,
+      "loss": 0.5005,
+      "step": 26188
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3861124538505854e-07,
+      "loss": 0.5802,
+      "step": 26189
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3841667546535265e-07,
+      "loss": 0.6061,
+      "step": 26190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3822224124969918e-07,
+      "loss": 0.6112,
+      "step": 26191
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.380279427407727e-07,
+      "loss": 0.5807,
+      "step": 26192
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.378337799412477e-07,
+      "loss": 0.5516,
+      "step": 26193
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3763975285379426e-07,
+      "loss": 0.5414,
+      "step": 26194
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3744586148108475e-07,
+      "loss": 0.5353,
+      "step": 26195
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3725210582578584e-07,
+      "loss": 0.5969,
+      "step": 26196
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3705848589056437e-07,
+      "loss": 0.5994,
+      "step": 26197
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368650016780837e-07,
+      "loss": 0.5892,
+      "step": 26198
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3667165319100729e-07,
+      "loss": 0.5843,
+      "step": 26199
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3647844043199633e-07,
+      "loss": 0.5334,
+      "step": 26200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3628536340370869e-07,
+      "loss": 0.5776,
+      "step": 26201
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3609242210880114e-07,
+      "loss": 0.5687,
+      "step": 26202
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3589961654992934e-07,
+      "loss": 0.524,
+      "step": 26203
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.357069467297456e-07,
+      "loss": 0.5816,
+      "step": 26204
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3551441265090336e-07,
+      "loss": 0.6298,
+      "step": 26205
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.353220143160483e-07,
+      "loss": 0.6137,
+      "step": 26206
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.351297517278316e-07,
+      "loss": 0.6384,
+      "step": 26207
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3493762488889784e-07,
+      "loss": 0.584,
+      "step": 26208
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3474563380188932e-07,
+      "loss": 0.5232,
+      "step": 26209
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.345537784694495e-07,
+      "loss": 0.519,
+      "step": 26210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.343620588942196e-07,
+      "loss": 0.6027,
+      "step": 26211
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3417047507883528e-07,
+      "loss": 0.5889,
+      "step": 26212
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3397902702593446e-07,
+      "loss": 0.583,
+      "step": 26213
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3378771473815054e-07,
+      "loss": 0.5555,
+      "step": 26214
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3359653821811812e-07,
+      "loss": 0.5798,
+      "step": 26215
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.334054974684662e-07,
+      "loss": 0.5698,
+      "step": 26216
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3321459249182266e-07,
+      "loss": 0.5779,
+      "step": 26217
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3302382329081765e-07,
+      "loss": 0.5799,
+      "step": 26218
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3283318986807458e-07,
+      "loss": 0.5496,
+      "step": 26219
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3264269222621696e-07,
+      "loss": 0.5815,
+      "step": 26220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3245233036786597e-07,
+      "loss": 0.5674,
+      "step": 26221
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.322621042956418e-07,
+      "loss": 0.5232,
+      "step": 26222
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.320720140121612e-07,
+      "loss": 0.5579,
+      "step": 26223
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3188205952003984e-07,
+      "loss": 0.5186,
+      "step": 26224
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3169224082189236e-07,
+      "loss": 0.5628,
+      "step": 26225
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3150255792033107e-07,
+      "loss": 0.5764,
+      "step": 26226
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.313130108179661e-07,
+      "loss": 0.5069,
+      "step": 26227
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3112359951740428e-07,
+      "loss": 0.5713,
+      "step": 26228
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.309343240212546e-07,
+      "loss": 0.6252,
+      "step": 26229
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.307451843321206e-07,
+      "loss": 0.5621,
+      "step": 26230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3055618045260344e-07,
+      "loss": 0.5334,
+      "step": 26231
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3036731238530553e-07,
+      "loss": 0.6005,
+      "step": 26232
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3017858013282702e-07,
+      "loss": 0.5507,
+      "step": 26233
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.299899836977625e-07,
+      "loss": 0.5841,
+      "step": 26234
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2980152308270877e-07,
+      "loss": 0.6123,
+      "step": 26235
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2961319829025708e-07,
+      "loss": 0.5432,
+      "step": 26236
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2942500932300317e-07,
+      "loss": 0.5951,
+      "step": 26237
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2923695618353272e-07,
+      "loss": 0.643,
+      "step": 26238
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2904903887443477e-07,
+      "loss": 0.529,
+      "step": 26239
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2886125739829724e-07,
+      "loss": 0.5692,
+      "step": 26240
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2867361175770143e-07,
+      "loss": 0.6359,
+      "step": 26241
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.284861019552297e-07,
+      "loss": 0.5885,
+      "step": 26242
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2829872799346334e-07,
+      "loss": 0.5364,
+      "step": 26243
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2811148987498025e-07,
+      "loss": 0.6053,
+      "step": 26244
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2792438760235838e-07,
+      "loss": 0.5197,
+      "step": 26245
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2773742117817012e-07,
+      "loss": 0.579,
+      "step": 26246
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2755059060499008e-07,
+      "loss": 0.566,
+      "step": 26247
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2736389588538844e-07,
+      "loss": 0.5695,
+      "step": 26248
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2717733702193424e-07,
+      "loss": 0.5318,
+      "step": 26249
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2699091401719544e-07,
+      "loss": 0.5848,
+      "step": 26250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2680462687373664e-07,
+      "loss": 0.5672,
+      "step": 26251
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2661847559412133e-07,
+      "loss": 0.5534,
+      "step": 26252
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.264324601809108e-07,
+      "loss": 0.5357,
+      "step": 26253
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.262465806366664e-07,
+      "loss": 0.5446,
+      "step": 26254
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2606083696394488e-07,
+      "loss": 0.5579,
+      "step": 26255
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2587522916530093e-07,
+      "loss": 0.5611,
+      "step": 26256
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2568975724329134e-07,
+      "loss": 0.4684,
+      "step": 26257
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2550442120046636e-07,
+      "loss": 0.5774,
+      "step": 26258
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.253192210393772e-07,
+      "loss": 0.5971,
+      "step": 26259
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.251341567625719e-07,
+      "loss": 0.5514,
+      "step": 26260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2494922837259727e-07,
+      "loss": 0.5082,
+      "step": 26261
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2476443587199904e-07,
+      "loss": 0.5804,
+      "step": 26262
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2457977926331966e-07,
+      "loss": 0.5593,
+      "step": 26263
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2439525854909817e-07,
+      "loss": 0.5071,
+      "step": 26264
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.24210873731877e-07,
+      "loss": 0.5567,
+      "step": 26265
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2402662481419082e-07,
+      "loss": 0.5481,
+      "step": 26266
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2384251179857642e-07,
+      "loss": 0.5495,
+      "step": 26267
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2365853468756628e-07,
+      "loss": 0.5878,
+      "step": 26268
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2347469348369389e-07,
+      "loss": 0.5943,
+      "step": 26269
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2329098818948836e-07,
+      "loss": 0.4926,
+      "step": 26270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2310741880747545e-07,
+      "loss": 0.5741,
+      "step": 26271
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2292398534018423e-07,
+      "loss": 0.5955,
+      "step": 26272
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2274068779013825e-07,
+      "loss": 0.5797,
+      "step": 26273
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2255752615985884e-07,
+      "loss": 0.5381,
+      "step": 26274
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2237450045186616e-07,
+      "loss": 0.5629,
+      "step": 26275
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2219161066868047e-07,
+      "loss": 0.5814,
+      "step": 26276
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.220088568128175e-07,
+      "loss": 0.623,
+      "step": 26277
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.21826238886793e-07,
+      "loss": 0.6049,
+      "step": 26278
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2164375689311725e-07,
+      "loss": 0.5025,
+      "step": 26279
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2146141083430486e-07,
+      "loss": 0.5634,
+      "step": 26280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.212792007128638e-07,
+      "loss": 0.5567,
+      "step": 26281
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.210971265312988e-07,
+      "loss": 0.5553,
+      "step": 26282
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2091518829212e-07,
+      "loss": 0.5713,
+      "step": 26283
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.207333859978277e-07,
+      "loss": 0.5941,
+      "step": 26284
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2055171965092538e-07,
+      "loss": 0.5369,
+      "step": 26285
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2037018925391108e-07,
+      "loss": 0.5395,
+      "step": 26286
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2018879480928502e-07,
+      "loss": 0.5533,
+      "step": 26287
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2000753631954298e-07,
+      "loss": 0.6224,
+      "step": 26288
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.198264137871774e-07,
+      "loss": 0.5571,
+      "step": 26289
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1964542721468187e-07,
+      "loss": 0.5538,
+      "step": 26290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1946457660454657e-07,
+      "loss": 0.5723,
+      "step": 26291
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.192838619592618e-07,
+      "loss": 0.5129,
+      "step": 26292
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1910328328131105e-07,
+      "loss": 0.5772,
+      "step": 26293
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1892284057318349e-07,
+      "loss": 0.5612,
+      "step": 26294
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1874253383735824e-07,
+      "loss": 0.521,
+      "step": 26295
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1856236307631996e-07,
+      "loss": 0.5733,
+      "step": 26296
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1838232829254447e-07,
+      "loss": 0.573,
+      "step": 26297
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1820242948851091e-07,
+      "loss": 0.5442,
+      "step": 26298
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1802266666669616e-07,
+      "loss": 0.5541,
+      "step": 26299
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1784303982957158e-07,
+      "loss": 0.633,
+      "step": 26300
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1766354897960963e-07,
+      "loss": 0.6008,
+      "step": 26301
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1748419411928169e-07,
+      "loss": 0.5471,
+      "step": 26302
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1730497525105356e-07,
+      "loss": 0.5777,
+      "step": 26303
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1712589237739324e-07,
+      "loss": 0.6054,
+      "step": 26304
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1694694550076435e-07,
+      "loss": 0.6086,
+      "step": 26305
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1676813462363046e-07,
+      "loss": 0.5912,
+      "step": 26306
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1658945974845071e-07,
+      "loss": 0.6111,
+      "step": 26307
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1641092087768314e-07,
+      "loss": 0.4905,
+      "step": 26308
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.16232518013788e-07,
+      "loss": 0.6014,
+      "step": 26309
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1605425115921665e-07,
+      "loss": 0.5361,
+      "step": 26310
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1587612031642271e-07,
+      "loss": 0.5558,
+      "step": 26311
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1569812548785974e-07,
+      "loss": 0.5579,
+      "step": 26312
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1552026667597582e-07,
+      "loss": 0.5636,
+      "step": 26313
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1534254388321786e-07,
+      "loss": 0.5992,
+      "step": 26314
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1516495711203169e-07,
+      "loss": 0.537,
+      "step": 26315
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1498750636486089e-07,
+      "loss": 0.5449,
+      "step": 26316
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1481019164414909e-07,
+      "loss": 0.5706,
+      "step": 26317
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.146330129523343e-07,
+      "loss": 0.6451,
+      "step": 26318
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1445597029185462e-07,
+      "loss": 0.584,
+      "step": 26319
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1427906366514696e-07,
+      "loss": 0.5161,
+      "step": 26320
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1410229307464715e-07,
+      "loss": 0.5396,
+      "step": 26321
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1392565852278437e-07,
+      "loss": 0.6103,
+      "step": 26322
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.137491600119922e-07,
+      "loss": 0.5737,
+      "step": 26323
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1357279754469874e-07,
+      "loss": 0.6267,
+      "step": 26324
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1339657112333091e-07,
+      "loss": 0.5639,
+      "step": 26325
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1322048075031233e-07,
+      "loss": 0.6067,
+      "step": 26326
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1304452642806773e-07,
+      "loss": 0.5937,
+      "step": 26327
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.128687081590174e-07,
+      "loss": 0.6259,
+      "step": 26328
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1269302594558163e-07,
+      "loss": 0.5232,
+      "step": 26329
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1251747979017735e-07,
+      "loss": 0.5955,
+      "step": 26330
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1234206969522155e-07,
+      "loss": 0.5352,
+      "step": 26331
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.121667956631256e-07,
+      "loss": 0.5437,
+      "step": 26332
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1199165769630316e-07,
+      "loss": 0.5333,
+      "step": 26333
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1181665579716339e-07,
+      "loss": 0.5502,
+      "step": 26334
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1164178996811659e-07,
+      "loss": 0.5683,
+      "step": 26335
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1146706021156639e-07,
+      "loss": 0.6174,
+      "step": 26336
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1129246652991755e-07,
+      "loss": 0.5948,
+      "step": 26337
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1111800892557367e-07,
+      "loss": 0.5849,
+      "step": 26338
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.109436874009362e-07,
+      "loss": 0.5251,
+      "step": 26339
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1076950195840208e-07,
+      "loss": 0.5203,
+      "step": 26340
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.105954526003683e-07,
+      "loss": 0.567,
+      "step": 26341
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1042153932923184e-07,
+      "loss": 0.5834,
+      "step": 26342
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.102477621473852e-07,
+      "loss": 0.5818,
+      "step": 26343
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.100741210572176e-07,
+      "loss": 0.5961,
+      "step": 26344
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0990061606112156e-07,
+      "loss": 0.5421,
+      "step": 26345
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0972724716148186e-07,
+      "loss": 0.5782,
+      "step": 26346
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0955401436068657e-07,
+      "loss": 0.6008,
+      "step": 26347
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0938091766111825e-07,
+      "loss": 0.6029,
+      "step": 26348
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0920795706515941e-07,
+      "loss": 0.5368,
+      "step": 26349
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0903513257519039e-07,
+      "loss": 0.5164,
+      "step": 26350
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0886244419358816e-07,
+      "loss": 0.5779,
+      "step": 26351
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0868989192272972e-07,
+      "loss": 0.5444,
+      "step": 26352
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0851747576498983e-07,
+      "loss": 0.5745,
+      "step": 26353
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0834519572274105e-07,
+      "loss": 0.6161,
+      "step": 26354
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0817305179835368e-07,
+      "loss": 0.5646,
+      "step": 26355
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0800104399419698e-07,
+      "loss": 0.6291,
+      "step": 26356
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.078291723126379e-07,
+      "loss": 0.52,
+      "step": 26357
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0765743675604123e-07,
+      "loss": 0.5234,
+      "step": 26358
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0748583732677065e-07,
+      "loss": 0.5959,
+      "step": 26359
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0731437402718647e-07,
+      "loss": 0.5438,
+      "step": 26360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0714304685965015e-07,
+      "loss": 0.5554,
+      "step": 26361
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0697185582651759e-07,
+      "loss": 0.5659,
+      "step": 26362
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0680080093014466e-07,
+      "loss": 0.5357,
+      "step": 26363
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0662988217288617e-07,
+      "loss": 0.6125,
+      "step": 26364
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0645909955709355e-07,
+      "loss": 0.5251,
+      "step": 26365
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0628845308511714e-07,
+      "loss": 0.6451,
+      "step": 26366
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0611794275930398e-07,
+      "loss": 0.573,
+      "step": 26367
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0594756858200328e-07,
+      "loss": 0.5907,
+      "step": 26368
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.057773305555565e-07,
+      "loss": 0.5526,
+      "step": 26369
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0560722868230733e-07,
+      "loss": 0.4682,
+      "step": 26370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0543726296459722e-07,
+      "loss": 0.5556,
+      "step": 26371
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0526743340476431e-07,
+      "loss": 0.6633,
+      "step": 26372
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.050977400051445e-07,
+      "loss": 0.5512,
+      "step": 26373
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0492818276807592e-07,
+      "loss": 0.529,
+      "step": 26374
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0475876169589005e-07,
+      "loss": 0.5853,
+      "step": 26375
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0458947679091724e-07,
+      "loss": 0.5179,
+      "step": 26376
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0442032805548896e-07,
+      "loss": 0.598,
+      "step": 26377
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0425131549193224e-07,
+      "loss": 0.6191,
+      "step": 26378
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0408243910257187e-07,
+      "loss": 0.6028,
+      "step": 26379
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0391369888973269e-07,
+      "loss": 0.5104,
+      "step": 26380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0374509485573614e-07,
+      "loss": 0.5861,
+      "step": 26381
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0357662700290372e-07,
+      "loss": 0.5514,
+      "step": 26382
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0340829533355135e-07,
+      "loss": 0.5742,
+      "step": 26383
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0324009984999717e-07,
+      "loss": 0.5551,
+      "step": 26384
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.03072040554556e-07,
+      "loss": 0.5362,
+      "step": 26385
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0290411744953932e-07,
+      "loss": 0.4924,
+      "step": 26386
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0273633053725862e-07,
+      "loss": 0.5431,
+      "step": 26387
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0256867982002094e-07,
+      "loss": 0.5409,
+      "step": 26388
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0240116530013666e-07,
+      "loss": 0.539,
+      "step": 26389
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0223378697990837e-07,
+      "loss": 0.6012,
+      "step": 26390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0206654486164091e-07,
+      "loss": 0.5603,
+      "step": 26391
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0189943894763355e-07,
+      "loss": 0.4874,
+      "step": 26392
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0173246924018775e-07,
+      "loss": 0.5816,
+      "step": 26393
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0156563574160172e-07,
+      "loss": 0.6085,
+      "step": 26394
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0139893845416804e-07,
+      "loss": 0.5781,
+      "step": 26395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0123237738018377e-07,
+      "loss": 0.5671,
+      "step": 26396
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0106595252193929e-07,
+      "loss": 0.6018,
+      "step": 26397
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0089966388172611e-07,
+      "loss": 0.56,
+      "step": 26398
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.007335114618313e-07,
+      "loss": 0.5856,
+      "step": 26399
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0056749526454079e-07,
+      "loss": 0.5309,
+      "step": 26400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0040161529214166e-07,
+      "loss": 0.529,
+      "step": 26401
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0023587154691317e-07,
+      "loss": 0.5331,
+      "step": 26402
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0007026403113906e-07,
+      "loss": 0.5937,
+      "step": 26403
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.990479274709641e-08,
+      "loss": 0.5894,
+      "step": 26404
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.973945769706228e-08,
+      "loss": 0.5514,
+      "step": 26405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.957425888331262e-08,
+      "loss": 0.5672,
+      "step": 26406
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.940919630812117e-08,
+      "loss": 0.5539,
+      "step": 26407
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.924426997375836e-08,
+      "loss": 0.5925,
+      "step": 26408
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.907947988249344e-08,
+      "loss": 0.5316,
+      "step": 26409
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.891482603659463e-08,
+      "loss": 0.6265,
+      "step": 26410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.875030843832789e-08,
+      "loss": 0.55,
+      "step": 26411
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.858592708995696e-08,
+      "loss": 0.5826,
+      "step": 26412
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.842168199374336e-08,
+      "loss": 0.6067,
+      "step": 26413
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.825757315194751e-08,
+      "loss": 0.6283,
+      "step": 26414
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.80936005668287e-08,
+      "loss": 0.5929,
+      "step": 26415
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.792976424064293e-08,
+      "loss": 0.5827,
+      "step": 26416
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.776606417564283e-08,
+      "loss": 0.4725,
+      "step": 26417
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.760250037408325e-08,
+      "loss": 0.5435,
+      "step": 26418
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.743907283821463e-08,
+      "loss": 0.5371,
+      "step": 26419
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.727578157028627e-08,
+      "loss": 0.578,
+      "step": 26420
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.711262657254305e-08,
+      "loss": 0.6027,
+      "step": 26421
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.694960784723206e-08,
+      "loss": 0.5374,
+      "step": 26422
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.678672539659595e-08,
+      "loss": 0.5274,
+      "step": 26423
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.662397922287625e-08,
+      "loss": 0.581,
+      "step": 26424
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.64613693283123e-08,
+      "loss": 0.6053,
+      "step": 26425
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.629889571514228e-08,
+      "loss": 0.6034,
+      "step": 26426
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.61365583856011e-08,
+      "loss": 0.6065,
+      "step": 26427
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.597435734192251e-08,
+      "loss": 0.6016,
+      "step": 26428
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.581229258634028e-08,
+      "loss": 0.5902,
+      "step": 26429
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.565036412108375e-08,
+      "loss": 0.5619,
+      "step": 26430
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.548857194838112e-08,
+      "loss": 0.5464,
+      "step": 26431
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.53269160704573e-08,
+      "loss": 0.574,
+      "step": 26432
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.516539648953937e-08,
+      "loss": 0.5727,
+      "step": 26433
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.50040132078478e-08,
+      "loss": 0.5849,
+      "step": 26434
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.484276622760413e-08,
+      "loss": 0.597,
+      "step": 26435
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.468165555102882e-08,
+      "loss": 0.5124,
+      "step": 26436
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.452068118033675e-08,
+      "loss": 0.5171,
+      "step": 26437
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.435984311774394e-08,
+      "loss": 0.5991,
+      "step": 26438
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.419914136546304e-08,
+      "loss": 0.6497,
+      "step": 26439
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.403857592570675e-08,
+      "loss": 0.5338,
+      "step": 26440
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.38781468006833e-08,
+      "loss": 0.5583,
+      "step": 26441
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.371785399260092e-08,
+      "loss": 0.6014,
+      "step": 26442
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.35576975036645e-08,
+      "loss": 0.6002,
+      "step": 26443
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.339767733608007e-08,
+      "loss": 0.5478,
+      "step": 26444
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.323779349204698e-08,
+      "loss": 0.4906,
+      "step": 26445
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.307804597376791e-08,
+      "loss": 0.5332,
+      "step": 26446
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.291843478343888e-08,
+      "loss": 0.5684,
+      "step": 26447
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.275895992325812e-08,
+      "loss": 0.6239,
+      "step": 26448
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.259962139541945e-08,
+      "loss": 0.6314,
+      "step": 26449
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.244041920211445e-08,
+      "loss": 0.5867,
+      "step": 26450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.228135334553579e-08,
+      "loss": 0.5488,
+      "step": 26451
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.212242382787173e-08,
+      "loss": 0.5345,
+      "step": 26452
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.19636306513072e-08,
+      "loss": 0.6113,
+      "step": 26453
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.180497381803044e-08,
+      "loss": 0.5841,
+      "step": 26454
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.164645333022303e-08,
+      "loss": 0.5293,
+      "step": 26455
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.148806919006659e-08,
+      "loss": 0.5135,
+      "step": 26456
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.132982139974045e-08,
+      "loss": 0.4833,
+      "step": 26457
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.117170996142289e-08,
+      "loss": 0.5468,
+      "step": 26458
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.101373487728993e-08,
+      "loss": 0.6083,
+      "step": 26459
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.085589614951318e-08,
+      "loss": 0.582,
+      "step": 26460
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.069819378026757e-08,
+      "loss": 0.62,
+      "step": 26461
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.054062777172134e-08,
+      "loss": 0.5694,
+      "step": 26462
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.038319812604279e-08,
+      "loss": 0.5658,
+      "step": 26463
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.022590484539906e-08,
+      "loss": 0.5268,
+      "step": 26464
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.006874793195509e-08,
+      "loss": 0.5629,
+      "step": 26465
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.991172738787246e-08,
+      "loss": 0.5316,
+      "step": 26466
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.97548432153117e-08,
+      "loss": 0.5474,
+      "step": 26467
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.959809541643216e-08,
+      "loss": 0.5775,
+      "step": 26468
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.944148399339103e-08,
+      "loss": 0.5234,
+      "step": 26469
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.928500894834436e-08,
+      "loss": 0.5978,
+      "step": 26470
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.912867028344263e-08,
+      "loss": 0.5391,
+      "step": 26471
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.89724680008397e-08,
+      "loss": 0.5668,
+      "step": 26472
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.881640210268494e-08,
+      "loss": 0.5471,
+      "step": 26473
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.866047259112442e-08,
+      "loss": 0.6076,
+      "step": 26474
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.850467946830421e-08,
+      "loss": 0.5512,
+      "step": 26475
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.834902273636925e-08,
+      "loss": 0.536,
+      "step": 26476
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.819350239746005e-08,
+      "loss": 0.5434,
+      "step": 26477
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.803811845371823e-08,
+      "loss": 0.5564,
+      "step": 26478
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.788287090728098e-08,
+      "loss": 0.5367,
+      "step": 26479
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.772775976028547e-08,
+      "loss": 0.5157,
+      "step": 26480
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.757278501486555e-08,
+      "loss": 0.589,
+      "step": 26481
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.741794667315396e-08,
+      "loss": 0.5718,
+      "step": 26482
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.726324473728121e-08,
+      "loss": 0.5097,
+      "step": 26483
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.710867920937671e-08,
+      "loss": 0.6015,
+      "step": 26484
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.695425009156654e-08,
+      "loss": 0.6519,
+      "step": 26485
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.679995738597568e-08,
+      "loss": 0.6092,
+      "step": 26486
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.664580109472908e-08,
+      "loss": 0.5231,
+      "step": 26487
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.649178121994616e-08,
+      "loss": 0.5707,
+      "step": 26488
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.633789776374745e-08,
+      "loss": 0.5814,
+      "step": 26489
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.618415072825014e-08,
+      "loss": 0.5849,
+      "step": 26490
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.603054011557032e-08,
+      "loss": 0.5635,
+      "step": 26491
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.587706592782075e-08,
+      "loss": 0.5955,
+      "step": 26492
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.572372816711416e-08,
+      "loss": 0.5863,
+      "step": 26493
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.557052683556e-08,
+      "loss": 0.5763,
+      "step": 26494
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.541746193526879e-08,
+      "loss": 0.5174,
+      "step": 26495
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.526453346834329e-08,
+      "loss": 0.526,
+      "step": 26496
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.51117414368896e-08,
+      "loss": 0.5818,
+      "step": 26497
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.49590858430116e-08,
+      "loss": 0.608,
+      "step": 26498
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.480656668880871e-08,
+      "loss": 0.6349,
+      "step": 26499
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.465418397637815e-08,
+      "loss": 0.5603,
+      "step": 26500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.450193770781933e-08,
+      "loss": 0.5322,
+      "step": 26501
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.434982788522617e-08,
+      "loss": 0.5556,
+      "step": 26502
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.41978545106925e-08,
+      "loss": 0.569,
+      "step": 26503
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.404601758630892e-08,
+      "loss": 0.5966,
+      "step": 26504
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.389431711416484e-08,
+      "loss": 0.6159,
+      "step": 26505
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.374275309634861e-08,
+      "loss": 0.4956,
+      "step": 26506
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.35913255349452e-08,
+      "loss": 0.5918,
+      "step": 26507
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.34400344320374e-08,
+      "loss": 0.5727,
+      "step": 26508
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.328887978970912e-08,
+      "loss": 0.594,
+      "step": 26509
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.313786161003978e-08,
+      "loss": 0.5709,
+      "step": 26510
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.298697989510551e-08,
+      "loss": 0.5431,
+      "step": 26511
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.283623464698576e-08,
+      "loss": 0.5197,
+      "step": 26512
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.268562586775331e-08,
+      "loss": 0.5901,
+      "step": 26513
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.253515355948094e-08,
+      "loss": 0.5147,
+      "step": 26514
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.238481772423812e-08,
+      "loss": 0.6064,
+      "step": 26515
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.223461836409429e-08,
+      "loss": 0.5634,
+      "step": 26516
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.208455548111782e-08,
+      "loss": 0.4733,
+      "step": 26517
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.19346290773726e-08,
+      "loss": 0.6513,
+      "step": 26518
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.17848391549203e-08,
+      "loss": 0.5836,
+      "step": 26519
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.163518571582486e-08,
+      "loss": 0.5771,
+      "step": 26520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.148566876214348e-08,
+      "loss": 0.5837,
+      "step": 26521
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.133628829593454e-08,
+      "loss": 0.5699,
+      "step": 26522
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.118704431925417e-08,
+      "loss": 0.5434,
+      "step": 26523
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.103793683415517e-08,
+      "loss": 0.6231,
+      "step": 26524
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.088896584269035e-08,
+      "loss": 0.5568,
+      "step": 26525
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.074013134690805e-08,
+      "loss": 0.602,
+      "step": 26526
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.059143334885778e-08,
+      "loss": 0.5798,
+      "step": 26527
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.044287185058564e-08,
+      "loss": 0.6025,
+      "step": 26528
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.029444685413445e-08,
+      "loss": 0.5809,
+      "step": 26529
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.014615836154926e-08,
+      "loss": 0.5039,
+      "step": 26530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.999800637486954e-08,
+      "loss": 0.5395,
+      "step": 26531
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.984999089613254e-08,
+      "loss": 0.5516,
+      "step": 26532
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.970211192737664e-08,
+      "loss": 0.5145,
+      "step": 26533
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.955436947063688e-08,
+      "loss": 0.6059,
+      "step": 26534
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.940676352794496e-08,
+      "loss": 0.5726,
+      "step": 26535
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.92592941013337e-08,
+      "loss": 0.5514,
+      "step": 26536
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.911196119283149e-08,
+      "loss": 0.621,
+      "step": 26537
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.896476480446558e-08,
+      "loss": 0.5624,
+      "step": 26538
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.881770493826324e-08,
+      "loss": 0.5838,
+      "step": 26539
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.867078159624508e-08,
+      "loss": 0.5604,
+      "step": 26540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.852399478043505e-08,
+      "loss": 0.5061,
+      "step": 26541
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.837734449285372e-08,
+      "loss": 0.6358,
+      "step": 26542
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.823083073551729e-08,
+      "loss": 0.6297,
+      "step": 26543
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.808445351044192e-08,
+      "loss": 0.576,
+      "step": 26544
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.793821281964376e-08,
+      "loss": 0.6164,
+      "step": 26545
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.779210866513342e-08,
+      "loss": 0.5609,
+      "step": 26546
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.764614104892265e-08,
+      "loss": 0.5951,
+      "step": 26547
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.750030997301872e-08,
+      "loss": 0.5598,
+      "step": 26548
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.735461543943001e-08,
+      "loss": 0.5548,
+      "step": 26549
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.720905745015939e-08,
+      "loss": 0.5376,
+      "step": 26550
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.70636360072119e-08,
+      "loss": 0.5249,
+      "step": 26551
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.691835111258817e-08,
+      "loss": 0.5435,
+      "step": 26552
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.67732027682866e-08,
+      "loss": 0.5936,
+      "step": 26553
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.662819097630558e-08,
+      "loss": 0.5404,
+      "step": 26554
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.64833157386391e-08,
+      "loss": 0.5266,
+      "step": 26555
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.633857705728331e-08,
+      "loss": 0.5742,
+      "step": 26556
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.619397493422665e-08,
+      "loss": 0.5606,
+      "step": 26557
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.604950937146193e-08,
+      "loss": 0.6329,
+      "step": 26558
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.590518037097539e-08,
+      "loss": 0.5993,
+      "step": 26559
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.576098793475428e-08,
+      "loss": 0.5881,
+      "step": 26560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.561693206478149e-08,
+      "loss": 0.5707,
+      "step": 26561
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.547301276303987e-08,
+      "loss": 0.6083,
+      "step": 26562
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.532923003151005e-08,
+      "loss": 0.5713,
+      "step": 26563
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.518558387217157e-08,
+      "loss": 0.5673,
+      "step": 26564
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.50420742869995e-08,
+      "loss": 0.5763,
+      "step": 26565
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.489870127796894e-08,
+      "loss": 0.6228,
+      "step": 26566
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.475546484705387e-08,
+      "loss": 0.5506,
+      "step": 26567
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.461236499622493e-08,
+      "loss": 0.625,
+      "step": 26568
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.446940172745055e-08,
+      "loss": 0.5715,
+      "step": 26569
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.432657504269803e-08,
+      "loss": 0.6082,
+      "step": 26570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.418388494393358e-08,
+      "loss": 0.6182,
+      "step": 26571
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.40413314331212e-08,
+      "loss": 0.554,
+      "step": 26572
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.389891451222153e-08,
+      "loss": 0.5778,
+      "step": 26573
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.375663418319411e-08,
+      "loss": 0.5673,
+      "step": 26574
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.36144904479974e-08,
+      "loss": 0.6177,
+      "step": 26575
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.347248330858869e-08,
+      "loss": 0.6069,
+      "step": 26576
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.333061276691866e-08,
+      "loss": 0.5535,
+      "step": 26577
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.318887882494352e-08,
+      "loss": 0.5829,
+      "step": 26578
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.304728148461171e-08,
+      "loss": 0.5725,
+      "step": 26579
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.290582074787056e-08,
+      "loss": 0.5585,
+      "step": 26580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.276449661666851e-08,
+      "loss": 0.6304,
+      "step": 26581
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.262330909295068e-08,
+      "loss": 0.5785,
+      "step": 26582
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.248225817865883e-08,
+      "loss": 0.5693,
+      "step": 26583
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.234134387573367e-08,
+      "loss": 0.5611,
+      "step": 26584
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.220056618611471e-08,
+      "loss": 0.5522,
+      "step": 26585
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.205992511174043e-08,
+      "loss": 0.568,
+      "step": 26586
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.191942065454372e-08,
+      "loss": 0.6218,
+      "step": 26587
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.177905281645969e-08,
+      "loss": 0.5152,
+      "step": 26588
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.163882159942015e-08,
+      "loss": 0.5469,
+      "step": 26589
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.149872700535465e-08,
+      "loss": 0.5661,
+      "step": 26590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.135876903618944e-08,
+      "loss": 0.5723,
+      "step": 26591
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.121894769385185e-08,
+      "loss": 0.5865,
+      "step": 26592
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.10792629802659e-08,
+      "loss": 0.5953,
+      "step": 26593
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.093971489735341e-08,
+      "loss": 0.5563,
+      "step": 26594
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.080030344703503e-08,
+      "loss": 0.5626,
+      "step": 26595
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.066102863122925e-08,
+      "loss": 0.5688,
+      "step": 26596
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.052189045185232e-08,
+      "loss": 0.5437,
+      "step": 26597
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.038288891081935e-08,
+      "loss": 0.5329,
+      "step": 26598
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.024402401004216e-08,
+      "loss": 0.555,
+      "step": 26599
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.010529575143255e-08,
+      "loss": 0.5474,
+      "step": 26600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.996670413689899e-08,
+      "loss": 0.5292,
+      "step": 26601
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.982824916834886e-08,
+      "loss": 0.5889,
+      "step": 26602
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.968993084768728e-08,
+      "loss": 0.5408,
+      "step": 26603
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.955174917681829e-08,
+      "loss": 0.5731,
+      "step": 26604
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.941370415764259e-08,
+      "loss": 0.5556,
+      "step": 26605
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.927579579205978e-08,
+      "loss": 0.5431,
+      "step": 26606
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.913802408196835e-08,
+      "loss": 0.6328,
+      "step": 26607
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.900038902926454e-08,
+      "loss": 0.5468,
+      "step": 26608
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.886289063584017e-08,
+      "loss": 0.5716,
+      "step": 26609
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.87255289035893e-08,
+      "loss": 0.5641,
+      "step": 26610
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.858830383440152e-08,
+      "loss": 0.5811,
+      "step": 26611
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.84512154301653e-08,
+      "loss": 0.5779,
+      "step": 26612
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.831426369276584e-08,
+      "loss": 0.5926,
+      "step": 26613
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.817744862409049e-08,
+      "loss": 0.6119,
+      "step": 26614
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.804077022601996e-08,
+      "loss": 0.5989,
+      "step": 26615
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.790422850043498e-08,
+      "loss": 0.5537,
+      "step": 26616
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.776782344921407e-08,
+      "loss": 0.5833,
+      "step": 26617
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.763155507423679e-08,
+      "loss": 0.5456,
+      "step": 26618
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.749542337737614e-08,
+      "loss": 0.5668,
+      "step": 26619
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.735942836050613e-08,
+      "loss": 0.5757,
+      "step": 26620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.72235700254964e-08,
+      "loss": 0.6061,
+      "step": 26621
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.708784837421989e-08,
+      "loss": 0.5786,
+      "step": 26622
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.695226340854177e-08,
+      "loss": 0.512,
+      "step": 26623
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.681681513032834e-08,
+      "loss": 0.5887,
+      "step": 26624
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.668150354144365e-08,
+      "loss": 0.6213,
+      "step": 26625
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.654632864374954e-08,
+      "loss": 0.5525,
+      "step": 26626
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.641129043910677e-08,
+      "loss": 0.5705,
+      "step": 26627
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.62763889293716e-08,
+      "loss": 0.6052,
+      "step": 26628
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.614162411640368e-08,
+      "loss": 0.6028,
+      "step": 26629
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.600699600205485e-08,
+      "loss": 0.5352,
+      "step": 26630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.587250458817806e-08,
+      "loss": 0.5164,
+      "step": 26631
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.57381498766252e-08,
+      "loss": 0.5579,
+      "step": 26632
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.560393186924363e-08,
+      "loss": 0.5189,
+      "step": 26633
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.546985056788191e-08,
+      "loss": 0.5544,
+      "step": 26634
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.533590597438411e-08,
+      "loss": 0.5559,
+      "step": 26635
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.520209809059319e-08,
+      "loss": 0.5805,
+      "step": 26636
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.506842691835214e-08,
+      "loss": 0.557,
+      "step": 26637
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.493489245949835e-08,
+      "loss": 0.5481,
+      "step": 26638
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.480149471586928e-08,
+      "loss": 0.5819,
+      "step": 26639
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.46682336893023e-08,
+      "loss": 0.5399,
+      "step": 26640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.453510938163043e-08,
+      "loss": 0.5367,
+      "step": 26641
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.440212179468553e-08,
+      "loss": 0.544,
+      "step": 26642
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.426927093029722e-08,
+      "loss": 0.5519,
+      "step": 26643
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.413655679029407e-08,
+      "loss": 0.5595,
+      "step": 26644
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.400397937650238e-08,
+      "loss": 0.5302,
+      "step": 26645
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.387153869074625e-08,
+      "loss": 0.5473,
+      "step": 26646
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.373923473484866e-08,
+      "loss": 0.5909,
+      "step": 26647
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.36070675106304e-08,
+      "loss": 0.6214,
+      "step": 26648
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.347503701990998e-08,
+      "loss": 0.573,
+      "step": 26649
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.334314326450264e-08,
+      "loss": 0.5593,
+      "step": 26650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.32113862462258e-08,
+      "loss": 0.6625,
+      "step": 26651
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.307976596689137e-08,
+      "loss": 0.5522,
+      "step": 26652
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.294828242831008e-08,
+      "loss": 0.5146,
+      "step": 26653
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.281693563229275e-08,
+      "loss": 0.5183,
+      "step": 26654
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.268572558064567e-08,
+      "loss": 0.5856,
+      "step": 26655
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.255465227517521e-08,
+      "loss": 0.5689,
+      "step": 26656
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.242371571768324e-08,
+      "loss": 0.5525,
+      "step": 26657
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.229291590997388e-08,
+      "loss": 0.6322,
+      "step": 26658
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.21622528538457e-08,
+      "loss": 0.5359,
+      "step": 26659
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.203172655109723e-08,
+      "loss": 0.5657,
+      "step": 26660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.190133700352486e-08,
+      "loss": 0.5877,
+      "step": 26661
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.177108421292266e-08,
+      "loss": 0.5712,
+      "step": 26662
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.164096818108145e-08,
+      "loss": 0.5234,
+      "step": 26663
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.151098890979423e-08,
+      "loss": 0.5765,
+      "step": 26664
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.138114640084847e-08,
+      "loss": 0.602,
+      "step": 26665
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.12514406560305e-08,
+      "loss": 0.5471,
+      "step": 26666
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.112187167712558e-08,
+      "loss": 0.5425,
+      "step": 26667
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.09924394659167e-08,
+      "loss": 0.6394,
+      "step": 26668
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.086314402418581e-08,
+      "loss": 0.5785,
+      "step": 26669
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.073398535371034e-08,
+      "loss": 0.5691,
+      "step": 26670
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.060496345626887e-08,
+      "loss": 0.5416,
+      "step": 26671
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.047607833363778e-08,
+      "loss": 0.5658,
+      "step": 26672
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.034732998758785e-08,
+      "loss": 0.5302,
+      "step": 26673
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.021871841989324e-08,
+      "loss": 0.5499,
+      "step": 26674
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.009024363232252e-08,
+      "loss": 0.5376,
+      "step": 26675
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.99619056266454e-08,
+      "loss": 0.5924,
+      "step": 26676
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.983370440462488e-08,
+      "loss": 0.6079,
+      "step": 26677
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.970563996802736e-08,
+      "loss": 0.6161,
+      "step": 26678
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.957771231861476e-08,
+      "loss": 0.5508,
+      "step": 26679
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9449921458146765e-08,
+      "loss": 0.6026,
+      "step": 26680
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.932226738838309e-08,
+      "loss": 0.5815,
+      "step": 26681
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9194750111079e-08,
+      "loss": 0.5742,
+      "step": 26682
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9067369627989755e-08,
+      "loss": 0.5765,
+      "step": 26683
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.894012594086951e-08,
+      "loss": 0.5869,
+      "step": 26684
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8813019051466855e-08,
+      "loss": 0.5977,
+      "step": 26685
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8686048961532627e-08,
+      "loss": 0.5664,
+      "step": 26686
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.85592156728132e-08,
+      "loss": 0.6268,
+      "step": 26687
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.843251918705495e-08,
+      "loss": 0.5399,
+      "step": 26688
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.830595950600093e-08,
+      "loss": 0.5789,
+      "step": 26689
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.817953663139198e-08,
+      "loss": 0.6245,
+      "step": 26690
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8053250564967803e-08,
+      "loss": 0.5992,
+      "step": 26691
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7927101308467015e-08,
+      "loss": 0.5251,
+      "step": 26692
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.78010888636249e-08,
+      "loss": 0.5892,
+      "step": 26693
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.767521323217673e-08,
+      "loss": 0.5305,
+      "step": 26694
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.754947441585335e-08,
+      "loss": 0.5074,
+      "step": 26695
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7423872416384474e-08,
+      "loss": 0.5795,
+      "step": 26696
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7298407235499844e-08,
+      "loss": 0.5754,
+      "step": 26697
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7173078874925845e-08,
+      "loss": 0.5374,
+      "step": 26698
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7047887336386664e-08,
+      "loss": 0.5442,
+      "step": 26699
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6922832621605364e-08,
+      "loss": 0.5645,
+      "step": 26700
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.679791473230278e-08,
+      "loss": 0.5318,
+      "step": 26701
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6673133670197555e-08,
+      "loss": 0.5928,
+      "step": 26702
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.654848943700719e-08,
+      "loss": 0.5807,
+      "step": 26703
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6423982034445876e-08,
+      "loss": 0.5518,
+      "step": 26704
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.629961146422891e-08,
+      "loss": 0.5199,
+      "step": 26705
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.617537772806603e-08,
+      "loss": 0.5954,
+      "step": 26706
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6051280827666975e-08,
+      "loss": 0.6315,
+      "step": 26707
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.592732076474039e-08,
+      "loss": 0.5177,
+      "step": 26708
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.580349754099157e-08,
+      "loss": 0.5315,
+      "step": 26709
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5679811158123597e-08,
+      "loss": 0.6002,
+      "step": 26710
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5556261617839555e-08,
+      "loss": 0.5944,
+      "step": 26711
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.543284892184031e-08,
+      "loss": 0.6005,
+      "step": 26712
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5309573071822276e-08,
+      "loss": 0.5604,
+      "step": 26713
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.518643406948299e-08,
+      "loss": 0.5656,
+      "step": 26714
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.506343191651553e-08,
+      "loss": 0.5042,
+      "step": 26715
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.494056661461522e-08,
+      "loss": 0.5826,
+      "step": 26716
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.48178381654707e-08,
+      "loss": 0.5835,
+      "step": 26717
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.469524657077063e-08,
+      "loss": 0.6045,
+      "step": 26718
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4572791832203655e-08,
+      "loss": 0.5626,
+      "step": 26719
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.445047395145286e-08,
+      "loss": 0.5845,
+      "step": 26720
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4328292930203584e-08,
+      "loss": 0.6133,
+      "step": 26721
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4206248770134474e-08,
+      "loss": 0.5761,
+      "step": 26722
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4084341472927515e-08,
+      "loss": 0.6341,
+      "step": 26723
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.396257104026026e-08,
+      "loss": 0.5816,
+      "step": 26724
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.384093747380581e-08,
+      "loss": 0.5395,
+      "step": 26725
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3719440775240606e-08,
+      "loss": 0.5483,
+      "step": 26726
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.359808094623442e-08,
+      "loss": 0.5808,
+      "step": 26727
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.347685798845925e-08,
+      "loss": 0.5884,
+      "step": 26728
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.335577190358154e-08,
+      "loss": 0.5656,
+      "step": 26729
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.323482269326885e-08,
+      "loss": 0.6038,
+      "step": 26730
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3114010359184286e-08,
+      "loss": 0.616,
+      "step": 26731
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.299333490299208e-08,
+      "loss": 0.5591,
+      "step": 26732
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2872796326350896e-08,
+      "loss": 0.5749,
+      "step": 26733
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2752394630920526e-08,
+      "loss": 0.5244,
+      "step": 26734
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.263212981835852e-08,
+      "loss": 0.5703,
+      "step": 26735
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.251200189031802e-08,
+      "loss": 0.557,
+      "step": 26736
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.239201084845324e-08,
+      "loss": 0.6545,
+      "step": 26737
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2272156694415097e-08,
+      "loss": 0.506,
+      "step": 26738
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.215243942985227e-08,
+      "loss": 0.5454,
+      "step": 26739
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.203285905641342e-08,
+      "loss": 0.5214,
+      "step": 26740
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.191341557574392e-08,
+      "loss": 0.566,
+      "step": 26741
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.179410898948689e-08,
+      "loss": 0.5794,
+      "step": 26742
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1674939299283245e-08,
+      "loss": 0.5533,
+      "step": 26743
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1555906506775e-08,
+      "loss": 0.5106,
+      "step": 26744
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.143701061359752e-08,
+      "loss": 0.4852,
+      "step": 26745
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.131825162138948e-08,
+      "loss": 0.5313,
+      "step": 26746
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.119962953178403e-08,
+      "loss": 0.5844,
+      "step": 26747
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.10811443464132e-08,
+      "loss": 0.562,
+      "step": 26748
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0962796066907906e-08,
+      "loss": 0.5514,
+      "step": 26749
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.084458469489573e-08,
+      "loss": 0.5616,
+      "step": 26750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0726510232004254e-08,
+      "loss": 0.5698,
+      "step": 26751
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.060857267985886e-08,
+      "loss": 0.5684,
+      "step": 26752
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.049077204008157e-08,
+      "loss": 0.6125,
+      "step": 26753
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0373108314294426e-08,
+      "loss": 0.5687,
+      "step": 26754
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0255581504113914e-08,
+      "loss": 0.5926,
+      "step": 26755
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.013819161116096e-08,
+      "loss": 0.5523,
+      "step": 26756
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.002093863704871e-08,
+      "loss": 0.5411,
+      "step": 26757
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.990382258339033e-08,
+      "loss": 0.554,
+      "step": 26758
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9786843451798964e-08,
+      "loss": 0.5482,
+      "step": 26759
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9670001243884434e-08,
+      "loss": 0.5577,
+      "step": 26760
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.955329596125325e-08,
+      "loss": 0.6026,
+      "step": 26761
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.943672760551077e-08,
+      "loss": 0.4836,
+      "step": 26762
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9320296178263506e-08,
+      "loss": 0.5905,
+      "step": 26763
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.920400168111239e-08,
+      "loss": 0.6485,
+      "step": 26764
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.908784411565726e-08,
+      "loss": 0.6089,
+      "step": 26765
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8971823483496825e-08,
+      "loss": 0.536,
+      "step": 26766
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8855939786228715e-08,
+      "loss": 0.5232,
+      "step": 26767
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.874019302544608e-08,
+      "loss": 0.593,
+      "step": 26768
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8624583202742104e-08,
+      "loss": 0.5437,
+      "step": 26769
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8509110319708844e-08,
+      "loss": 0.6299,
+      "step": 26770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.839377437793391e-08,
+      "loss": 0.4873,
+      "step": 26771
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.827857537900493e-08,
+      "loss": 0.5498,
+      "step": 26772
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.81635133245073e-08,
+      "loss": 0.5324,
+      "step": 26773
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.80485882160231e-08,
+      "loss": 0.5254,
+      "step": 26774
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.793380005513659e-08,
+      "loss": 0.5467,
+      "step": 26775
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.781914884342431e-08,
+      "loss": 0.5315,
+      "step": 26776
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.770463458246499e-08,
+      "loss": 0.6003,
+      "step": 26777
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.759025727383626e-08,
+      "loss": 0.5281,
+      "step": 26778
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.747601691910908e-08,
+      "loss": 0.5253,
+      "step": 26779
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.736191351985775e-08,
+      "loss": 0.5671,
+      "step": 26780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.724794707765212e-08,
+      "loss": 0.5867,
+      "step": 26781
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.713411759405873e-08,
+      "loss": 0.5946,
+      "step": 26782
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.70204250706463e-08,
+      "loss": 0.5685,
+      "step": 26783
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6906869508976935e-08,
+      "loss": 0.5441,
+      "step": 26784
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.679345091061604e-08,
+      "loss": 0.609,
+      "step": 26785
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6680169277122376e-08,
+      "loss": 0.5916,
+      "step": 26786
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.65670246100558e-08,
+      "loss": 0.6087,
+      "step": 26787
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.645401691097173e-08,
+      "loss": 0.5147,
+      "step": 26788
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.634114618142782e-08,
+      "loss": 0.5753,
+      "step": 26789
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.622841242297504e-08,
+      "loss": 0.5276,
+      "step": 26790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6115815637164386e-08,
+      "loss": 0.5494,
+      "step": 26791
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.600335582554793e-08,
+      "loss": 0.5843,
+      "step": 26792
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5891032989671126e-08,
+      "loss": 0.5552,
+      "step": 26793
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5778847131079384e-08,
+      "loss": 0.67,
+      "step": 26794
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5666798251318146e-08,
+      "loss": 0.6098,
+      "step": 26795
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5554886351928396e-08,
+      "loss": 0.6113,
+      "step": 26796
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.544311143445002e-08,
+      "loss": 0.6089,
+      "step": 26797
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.533147350042178e-08,
+      "loss": 0.6031,
+      "step": 26798
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5219972551378e-08,
+      "loss": 0.6295,
+      "step": 26799
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.510860858885524e-08,
+      "loss": 0.5325,
+      "step": 26800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.49973816143856e-08,
+      "loss": 0.5491,
+      "step": 26801
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.488629162949787e-08,
+      "loss": 0.5657,
+      "step": 26802
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.477533863572414e-08,
+      "loss": 0.5464,
+      "step": 26803
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.466452263458765e-08,
+      "loss": 0.633,
+      "step": 26804
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.455384362761495e-08,
+      "loss": 0.5893,
+      "step": 26805
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.444330161632926e-08,
+      "loss": 0.5498,
+      "step": 26806
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.433289660225159e-08,
+      "loss": 0.5801,
+      "step": 26807
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.422262858690074e-08,
+      "loss": 0.5191,
+      "step": 26808
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4112497571795475e-08,
+      "loss": 0.5577,
+      "step": 26809
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.400250355844904e-08,
+      "loss": 0.5808,
+      "step": 26810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.389264654837688e-08,
+      "loss": 0.5511,
+      "step": 26811
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.378292654309002e-08,
+      "loss": 0.5861,
+      "step": 26812
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3673343544098354e-08,
+      "loss": 0.5244,
+      "step": 26813
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.356389755290957e-08,
+      "loss": 0.5639,
+      "step": 26814
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.345458857103024e-08,
+      "loss": 0.5792,
+      "step": 26815
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3345416599964715e-08,
+      "loss": 0.5864,
+      "step": 26816
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.323638164121402e-08,
+      "loss": 0.6002,
+      "step": 26817
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.312748369628028e-08,
+      "loss": 0.5076,
+      "step": 26818
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3018722766661194e-08,
+      "loss": 0.5179,
+      "step": 26819
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.291009885385333e-08,
+      "loss": 0.6071,
+      "step": 26820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2801611959351065e-08,
+      "loss": 0.5863,
+      "step": 26821
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.269326208464875e-08,
+      "loss": 0.5685,
+      "step": 26822
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2585049231236295e-08,
+      "loss": 0.5547,
+      "step": 26823
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.247697340060142e-08,
+      "loss": 0.5772,
+      "step": 26824
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.236903459423402e-08,
+      "loss": 0.5318,
+      "step": 26825
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.226123281361849e-08,
+      "loss": 0.5476,
+      "step": 26826
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.215356806023807e-08,
+      "loss": 0.5746,
+      "step": 26827
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.20460403355738e-08,
+      "loss": 0.5444,
+      "step": 26828
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.193864964110561e-08,
+      "loss": 0.6208,
+      "step": 26829
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1831395978312317e-08,
+      "loss": 0.5709,
+      "step": 26830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.17242793486683e-08,
+      "loss": 0.5164,
+      "step": 26831
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1617299753649035e-08,
+      "loss": 0.5625,
+      "step": 26832
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.151045719472557e-08,
+      "loss": 0.5635,
+      "step": 26833
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1403751673367855e-08,
+      "loss": 0.5949,
+      "step": 26834
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.129718319104581e-08,
+      "loss": 0.5724,
+      "step": 26835
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.119075174922493e-08,
+      "loss": 0.5521,
+      "step": 26836
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1084457349369614e-08,
+      "loss": 0.5318,
+      "step": 26837
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.097829999294312e-08,
+      "loss": 0.608,
+      "step": 26838
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.087227968140539e-08,
+      "loss": 0.5387,
+      "step": 26839
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.076639641621638e-08,
+      "loss": 0.6022,
+      "step": 26840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0660650198832696e-08,
+      "loss": 0.6476,
+      "step": 26841
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.055504103070873e-08,
+      "loss": 0.6109,
+      "step": 26842
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.044956891329888e-08,
+      "loss": 0.5652,
+      "step": 26843
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.03442338480553e-08,
+      "loss": 0.5,
+      "step": 26844
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.023903583642574e-08,
+      "loss": 0.5464,
+      "step": 26845
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.013397487985793e-08,
+      "loss": 0.5314,
+      "step": 26846
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0029050979798476e-08,
+      "loss": 0.5546,
+      "step": 26847
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9924264137690684e-08,
+      "loss": 0.5032,
+      "step": 26848
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9819614354977834e-08,
+      "loss": 0.5753,
+      "step": 26849
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.971510163309766e-08,
+      "loss": 0.5836,
+      "step": 26850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.961072597348903e-08,
+      "loss": 0.5601,
+      "step": 26851
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.950648737758966e-08,
+      "loss": 0.5525,
+      "step": 26852
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.940238584683287e-08,
+      "loss": 0.5468,
+      "step": 26853
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.929842138265194e-08,
+      "loss": 0.606,
+      "step": 26854
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9194593986476845e-08,
+      "loss": 0.5523,
+      "step": 26855
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.909090365973645e-08,
+      "loss": 0.5877,
+      "step": 26856
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.898735040385737e-08,
+      "loss": 0.6045,
+      "step": 26857
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.888393422026515e-08,
+      "loss": 0.5972,
+      "step": 26858
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.878065511038198e-08,
+      "loss": 0.6276,
+      "step": 26859
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.867751307563006e-08,
+      "loss": 0.6286,
+      "step": 26860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.857450811742824e-08,
+      "loss": 0.5781,
+      "step": 26861
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.84716402371943e-08,
+      "loss": 0.5631,
+      "step": 26862
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.836890943634375e-08,
+      "loss": 0.5616,
+      "step": 26863
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8266315716288804e-08,
+      "loss": 0.5963,
+      "step": 26864
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.816385907844389e-08,
+      "loss": 0.5857,
+      "step": 26865
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.806153952421676e-08,
+      "loss": 0.6069,
+      "step": 26866
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.795935705501741e-08,
+      "loss": 0.5746,
+      "step": 26867
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.785731167224915e-08,
+      "loss": 0.4879,
+      "step": 26868
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7755403377318644e-08,
+      "loss": 0.6304,
+      "step": 26869
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.765363217162698e-08,
+      "loss": 0.617,
+      "step": 26870
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.755199805657528e-08,
+      "loss": 0.5351,
+      "step": 26871
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.74505010335624e-08,
+      "loss": 0.5548,
+      "step": 26872
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.734914110398391e-08,
+      "loss": 0.5364,
+      "step": 26873
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.724791826923424e-08,
+      "loss": 0.6308,
+      "step": 26874
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7146832530708944e-08,
+      "loss": 0.5327,
+      "step": 26875
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7045883889796907e-08,
+      "loss": 0.5494,
+      "step": 26876
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.694507234788702e-08,
+      "loss": 0.5722,
+      "step": 26877
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.684439790636707e-08,
+      "loss": 0.5927,
+      "step": 26878
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.674386056662371e-08,
+      "loss": 0.5622,
+      "step": 26879
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.664346033003918e-08,
+      "loss": 0.5998,
+      "step": 26880
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.65431971979946e-08,
+      "loss": 0.5826,
+      "step": 26881
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.644307117186996e-08,
+      "loss": 0.5933,
+      "step": 26882
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.634308225304417e-08,
+      "loss": 0.5672,
+      "step": 26883
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.624323044289169e-08,
+      "loss": 0.6295,
+      "step": 26884
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.614351574278807e-08,
+      "loss": 0.5938,
+      "step": 26885
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.604393815410445e-08,
+      "loss": 0.6158,
+      "step": 26886
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.594449767821084e-08,
+      "loss": 0.5673,
+      "step": 26887
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.584519431647726e-08,
+      "loss": 0.5582,
+      "step": 26888
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.574602807026817e-08,
+      "loss": 0.5197,
+      "step": 26889
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.564699894094914e-08,
+      "loss": 0.5894,
+      "step": 26890
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.554810692988353e-08,
+      "loss": 0.6339,
+      "step": 26891
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.544935203843025e-08,
+      "loss": 0.5785,
+      "step": 26892
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.535073426795044e-08,
+      "loss": 0.5406,
+      "step": 26893
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.525225361979967e-08,
+      "loss": 0.5557,
+      "step": 26894
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.515391009533464e-08,
+      "loss": 0.6183,
+      "step": 26895
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5055703695906494e-08,
+      "loss": 0.5853,
+      "step": 26896
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.495763442286859e-08,
+      "loss": 0.5647,
+      "step": 26897
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.485970227756985e-08,
+      "loss": 0.5987,
+      "step": 26898
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4761907261356976e-08,
+      "loss": 0.5192,
+      "step": 26899
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.466424937557667e-08,
+      "loss": 0.5653,
+      "step": 26900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.45667286215734e-08,
+      "loss": 0.6548,
+      "step": 26901
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.446934500068722e-08,
+      "loss": 0.534,
+      "step": 26902
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4372098514260374e-08,
+      "loss": 0.5047,
+      "step": 26903
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4274989163629594e-08,
+      "loss": 0.5801,
+      "step": 26904
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.417801695013156e-08,
+      "loss": 0.6151,
+      "step": 26905
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.408118187510079e-08,
+      "loss": 0.5066,
+      "step": 26906
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3984483939869525e-08,
+      "loss": 0.6103,
+      "step": 26907
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.388792314576894e-08,
+      "loss": 0.5946,
+      "step": 26908
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3791499494127964e-08,
+      "loss": 0.638,
+      "step": 26909
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3695212986272207e-08,
+      "loss": 0.5151,
+      "step": 26910
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3599063623527276e-08,
+      "loss": 0.5608,
+      "step": 26911
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.350305140721766e-08,
+      "loss": 0.6104,
+      "step": 26912
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.340717633866231e-08,
+      "loss": 0.5556,
+      "step": 26913
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3311438419182385e-08,
+      "loss": 0.5856,
+      "step": 26914
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.321583765009462e-08,
+      "loss": 0.534,
+      "step": 26915
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.312037403271351e-08,
+      "loss": 0.5446,
+      "step": 26916
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.302504756835467e-08,
+      "loss": 0.6148,
+      "step": 26917
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.292985825832817e-08,
+      "loss": 0.5128,
+      "step": 26918
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.2834806103945184e-08,
+      "loss": 0.5977,
+      "step": 26919
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.273989110651354e-08,
+      "loss": 0.6344,
+      "step": 26920
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2645113267338876e-08,
+      "loss": 0.5005,
+      "step": 26921
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.255047258772459e-08,
+      "loss": 0.5494,
+      "step": 26922
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.245596906897519e-08,
+      "loss": 0.5789,
+      "step": 26923
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2361602712389637e-08,
+      "loss": 0.6484,
+      "step": 26924
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.226737351926579e-08,
+      "loss": 0.5838,
+      "step": 26925
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2173281490901485e-08,
+      "loss": 0.5849,
+      "step": 26926
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2079326628592365e-08,
+      "loss": 0.5599,
+      "step": 26927
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1985508933629613e-08,
+      "loss": 0.5861,
+      "step": 26928
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.189182840730554e-08,
+      "loss": 0.5832,
+      "step": 26929
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.179828505090798e-08,
+      "loss": 0.5533,
+      "step": 26930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.170487886572482e-08,
+      "loss": 0.5455,
+      "step": 26931
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.161160985304168e-08,
+      "loss": 0.6003,
+      "step": 26932
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.151847801414199e-08,
+      "loss": 0.5768,
+      "step": 26933
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.142548335030693e-08,
+      "loss": 0.6306,
+      "step": 26934
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.13326258628166e-08,
+      "loss": 0.5317,
+      "step": 26935
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.123990555294776e-08,
+      "loss": 0.5352,
+      "step": 26936
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1147322421977154e-08,
+      "loss": 0.5167,
+      "step": 26937
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.105487647117933e-08,
+      "loss": 0.5592,
+      "step": 26938
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0962567701826594e-08,
+      "loss": 0.6363,
+      "step": 26939
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.087039611518683e-08,
+      "loss": 0.5676,
+      "step": 26940
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0778361712531234e-08,
+      "loss": 0.5454,
+      "step": 26941
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.068646449512547e-08,
+      "loss": 0.5346,
+      "step": 26942
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.059470446423296e-08,
+      "loss": 0.5735,
+      "step": 26943
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0503081621117145e-08,
+      "loss": 0.5471,
+      "step": 26944
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.041159596703924e-08,
+      "loss": 0.6346,
+      "step": 26945
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.032024750325824e-08,
+      "loss": 0.5301,
+      "step": 26946
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0229036231029795e-08,
+      "loss": 0.6179,
+      "step": 26947
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.013796215161069e-08,
+      "loss": 0.5525,
+      "step": 26948
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.004702526625325e-08,
+      "loss": 0.5513,
+      "step": 26949
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.99562255762087e-08,
+      "loss": 0.5559,
+      "step": 26950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.986556308272715e-08,
+      "loss": 0.5792,
+      "step": 26951
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9775037787056482e-08,
+      "loss": 0.5468,
+      "step": 26952
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.968464969044238e-08,
+      "loss": 0.5288,
+      "step": 26953
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9594398794127178e-08,
+      "loss": 0.5868,
+      "step": 26954
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9504285099354323e-08,
+      "loss": 0.5518,
+      "step": 26955
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9414308607362828e-08,
+      "loss": 0.5244,
+      "step": 26956
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9324469319391702e-08,
+      "loss": 0.6227,
+      "step": 26957
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.923476723667662e-08,
+      "loss": 0.5202,
+      "step": 26958
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9145202360452153e-08,
+      "loss": 0.5452,
+      "step": 26959
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9055774691951756e-08,
+      "loss": 0.5513,
+      "step": 26960
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8966484232404445e-08,
+      "loss": 0.5292,
+      "step": 26961
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.887733098303924e-08,
+      "loss": 0.5449,
+      "step": 26962
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.878831494508405e-08,
+      "loss": 0.5681,
+      "step": 26963
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8699436119763445e-08,
+      "loss": 0.5318,
+      "step": 26964
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8610694508299785e-08,
+      "loss": 0.5363,
+      "step": 26965
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8522090111914312e-08,
+      "loss": 0.5821,
+      "step": 26966
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8433622931826054e-08,
+      "loss": 0.5023,
+      "step": 26967
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8345292969254035e-08,
+      "loss": 0.6386,
+      "step": 26968
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8257100225411728e-08,
+      "loss": 0.6105,
+      "step": 26969
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8169044701513716e-08,
+      "loss": 0.63,
+      "step": 26970
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8081126398772364e-08,
+      "loss": 0.5544,
+      "step": 26971
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7993345318396702e-08,
+      "loss": 0.6029,
+      "step": 26972
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.790570146159466e-08,
+      "loss": 0.5826,
+      "step": 26973
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7818194829571932e-08,
+      "loss": 0.6222,
+      "step": 26974
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7730825423534224e-08,
+      "loss": 0.6274,
+      "step": 26975
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7643593244681688e-08,
+      "loss": 0.5496,
+      "step": 26976
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7556498294216694e-08,
+      "loss": 0.5477,
+      "step": 26977
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.746954057333606e-08,
+      "loss": 0.6167,
+      "step": 26978
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7382720083237724e-08,
+      "loss": 0.5407,
+      "step": 26979
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.729603682511517e-08,
+      "loss": 0.514,
+      "step": 26980
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7209490800162998e-08,
+      "loss": 0.6275,
+      "step": 26981
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7123082009570255e-08,
+      "loss": 0.4783,
+      "step": 26982
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7036810454527106e-08,
+      "loss": 0.5284,
+      "step": 26983
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6950676136220378e-08,
+      "loss": 0.6204,
+      "step": 26984
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6864679055835786e-08,
+      "loss": 0.6124,
+      "step": 26985
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6778819214555718e-08,
+      "loss": 0.5969,
+      "step": 26986
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6693096613562563e-08,
+      "loss": 0.5661,
+      "step": 26987
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6607511254035377e-08,
+      "loss": 0.6574,
+      "step": 26988
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.652206313715322e-08,
+      "loss": 0.5619,
+      "step": 26989
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6436752264090702e-08,
+      "loss": 0.5544,
+      "step": 26990
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6351578636021335e-08,
+      "loss": 0.5692,
+      "step": 26991
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6266542254118622e-08,
+      "loss": 0.5654,
+      "step": 26992
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6181643119551624e-08,
+      "loss": 0.5939,
+      "step": 26993
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6096881233489413e-08,
+      "loss": 0.5504,
+      "step": 26994
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6012256597096607e-08,
+      "loss": 0.6059,
+      "step": 26995
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5927769211540054e-08,
+      "loss": 0.6211,
+      "step": 26996
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5843419077981046e-08,
+      "loss": 0.499,
+      "step": 26997
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5759206197580876e-08,
+      "loss": 0.5852,
+      "step": 26998
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.567513057149862e-08,
+      "loss": 0.5285,
+      "step": 26999
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5591192200890012e-08,
+      "loss": 0.5716,
+      "step": 27000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5507391086911914e-08,
+      "loss": 0.5114,
+      "step": 27001
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5423727230715624e-08,
+      "loss": 0.5905,
+      "step": 27002
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5340200633453548e-08,
+      "loss": 0.6126,
+      "step": 27003
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5256811296274776e-08,
+      "loss": 0.62,
+      "step": 27004
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.517355922032616e-08,
+      "loss": 0.5314,
+      "step": 27005
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5090444406754566e-08,
+      "loss": 0.5331,
+      "step": 27006
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5007466856703522e-08,
+      "loss": 0.6098,
+      "step": 27007
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4924626571314336e-08,
+      "loss": 0.5249,
+      "step": 27008
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.484192355172721e-08,
+      "loss": 0.6007,
+      "step": 27009
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.475935779908012e-08,
+      "loss": 0.559,
+      "step": 27010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4676929314508823e-08,
+      "loss": 0.5839,
+      "step": 27011
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.459463809914908e-08,
+      "loss": 0.531,
+      "step": 27012
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.45124841541311e-08,
+      "loss": 0.5562,
+      "step": 27013
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4430467480587305e-08,
+      "loss": 0.5624,
+      "step": 27014
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4348588079644575e-08,
+      "loss": 0.5496,
+      "step": 27015
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4266845952432005e-08,
+      "loss": 0.6548,
+      "step": 27016
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.418524110007203e-08,
+      "loss": 0.5539,
+      "step": 27017
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4103773523689312e-08,
+      "loss": 0.6192,
+      "step": 27018
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.402244322440406e-08,
+      "loss": 0.5268,
+      "step": 27019
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3941250203335377e-08,
+      "loss": 0.5721,
+      "step": 27020
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.386019446160126e-08,
+      "loss": 0.6435,
+      "step": 27021
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3779276000317485e-08,
+      "loss": 0.5346,
+      "step": 27022
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3698494820595385e-08,
+      "loss": 0.6223,
+      "step": 27023
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3617850923548512e-08,
+      "loss": 0.6,
+      "step": 27024
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.353734431028598e-08,
+      "loss": 0.6143,
+      "step": 27025
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3456974981915792e-08,
+      "loss": 0.5336,
+      "step": 27026
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3376742939543727e-08,
+      "loss": 0.5845,
+      "step": 27027
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.329664818427335e-08,
+      "loss": 0.5205,
+      "step": 27028
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3216690717207115e-08,
+      "loss": 0.5543,
+      "step": 27029
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3136870539446355e-08,
+      "loss": 0.5229,
+      "step": 27030
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3057187652089085e-08,
+      "loss": 0.5935,
+      "step": 27031
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2977642056231097e-08,
+      "loss": 0.6021,
+      "step": 27032
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.289823375296707e-08,
+      "loss": 0.5691,
+      "step": 27033
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.281896274338946e-08,
+      "loss": 0.5369,
+      "step": 27034
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2739829028590732e-08,
+      "loss": 0.5239,
+      "step": 27035
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.266083260965779e-08,
+      "loss": 0.6317,
+      "step": 27036
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2581973487679763e-08,
+      "loss": 0.5446,
+      "step": 27037
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2503251663740232e-08,
+      "loss": 0.5675,
+      "step": 27038
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2424667138922774e-08,
+      "loss": 0.5596,
+      "step": 27039
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.234621991430874e-08,
+      "loss": 0.5743,
+      "step": 27040
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2267909990978387e-08,
+      "loss": 0.5973,
+      "step": 27041
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2189737370008623e-08,
+      "loss": 0.5262,
+      "step": 27042
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.211170205247526e-08,
+      "loss": 0.5955,
+      "step": 27043
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2033804039451877e-08,
+      "loss": 0.5912,
+      "step": 27044
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1956043332010957e-08,
+      "loss": 0.5662,
+      "step": 27045
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.187841993122275e-08,
+      "loss": 0.5635,
+      "step": 27046
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.180093383815529e-08,
+      "loss": 0.5393,
+      "step": 27047
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1723585053873287e-08,
+      "loss": 0.6269,
+      "step": 27048
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.164637357944366e-08,
+      "loss": 0.6368,
+      "step": 27049
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.156929941592667e-08,
+      "loss": 0.5101,
+      "step": 27050
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1492362564384807e-08,
+      "loss": 0.5279,
+      "step": 27051
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1415563025876106e-08,
+      "loss": 0.5679,
+      "step": 27052
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.13389008014564e-08,
+      "loss": 0.5875,
+      "step": 27053
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1262375892181497e-08,
+      "loss": 0.5231,
+      "step": 27054
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.118598829910501e-08,
+      "loss": 0.5564,
+      "step": 27055
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1109738023277204e-08,
+      "loss": 0.619,
+      "step": 27056
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1033625065747244e-08,
+      "loss": 0.5701,
+      "step": 27057
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0957649427563176e-08,
+      "loss": 0.5859,
+      "step": 27058
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0881811109770832e-08,
+      "loss": 0.5865,
+      "step": 27059
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0806110113411604e-08,
+      "loss": 0.5642,
+      "step": 27060
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0730546439530207e-08,
+      "loss": 0.5775,
+      "step": 27061
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0655120089164704e-08,
+      "loss": 0.5956,
+      "step": 27062
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.057983106335315e-08,
+      "loss": 0.586,
+      "step": 27063
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0504679363132495e-08,
+      "loss": 0.5696,
+      "step": 27064
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0429664989535246e-08,
+      "loss": 0.5607,
+      "step": 27065
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0354787943595022e-08,
+      "loss": 0.5637,
+      "step": 27066
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0280048226342108e-08,
+      "loss": 0.5917,
+      "step": 27067
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0205445838804572e-08,
+      "loss": 0.5852,
+      "step": 27068
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.013098078200826e-08,
+      "loss": 0.5464,
+      "step": 27069
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0056653056980125e-08,
+      "loss": 0.6286,
+      "step": 27070
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9982462664740464e-08,
+      "loss": 0.5899,
+      "step": 27071
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9908409606311795e-08,
+      "loss": 0.5419,
+      "step": 27072
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.983449388271219e-08,
+      "loss": 0.5971,
+      "step": 27073
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9760715494959725e-08,
+      "loss": 0.5866,
+      "step": 27074
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9687074444069142e-08,
+      "loss": 0.663,
+      "step": 27075
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9613570731054077e-08,
+      "loss": 0.5151,
+      "step": 27076
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.954020435692594e-08,
+      "loss": 0.6102,
+      "step": 27077
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9466975322693927e-08,
+      "loss": 0.5224,
+      "step": 27078
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9393883629366117e-08,
+      "loss": 0.5284,
+      "step": 27079
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9320929277948374e-08,
+      "loss": 0.5021,
+      "step": 27080
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.924811226944434e-08,
+      "loss": 0.543,
+      "step": 27081
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9175432604855438e-08,
+      "loss": 0.6527,
+      "step": 27082
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9102890285183083e-08,
+      "loss": 0.5755,
+      "step": 27083
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9030485311425373e-08,
+      "loss": 0.5429,
+      "step": 27084
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8958217684577064e-08,
+      "loss": 0.5275,
+      "step": 27085
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8886087405635134e-08,
+      "loss": 0.5944,
+      "step": 27086
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8814094475589904e-08,
+      "loss": 0.5587,
+      "step": 27087
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8742238895432807e-08,
+      "loss": 0.5533,
+      "step": 27088
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8670520666153046e-08,
+      "loss": 0.5569,
+      "step": 27089
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8598939788737612e-08,
+      "loss": 0.5873,
+      "step": 27090
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8527496264171273e-08,
+      "loss": 0.571,
+      "step": 27091
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8456190093436576e-08,
+      "loss": 0.506,
+      "step": 27092
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8385021277516067e-08,
+      "loss": 0.565,
+      "step": 27093
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8313989817387857e-08,
+      "loss": 0.5736,
+      "step": 27094
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8243095714030044e-08,
+      "loss": 0.6176,
+      "step": 27095
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8172338968417415e-08,
+      "loss": 0.6223,
+      "step": 27096
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8101719581523626e-08,
+      "loss": 0.5731,
+      "step": 27097
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8031237554322346e-08,
+      "loss": 0.568,
+      "step": 27098
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.796089288778169e-08,
+      "loss": 0.5616,
+      "step": 27099
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.789068558286977e-08,
+      "loss": 0.5444,
+      "step": 27100
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7820615640553596e-08,
+      "loss": 0.55,
+      "step": 27101
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7750683061796837e-08,
+      "loss": 0.5533,
+      "step": 27102
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7680887847560945e-08,
+      "loss": 0.5916,
+      "step": 27103
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7611229998808488e-08,
+      "loss": 0.5802,
+      "step": 27104
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7541709516496474e-08,
+      "loss": 0.5143,
+      "step": 27105
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7472326401581917e-08,
+      "loss": 0.589,
+      "step": 27106
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7403080655018502e-08,
+      "loss": 0.5638,
+      "step": 27107
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7333972277761012e-08,
+      "loss": 0.5666,
+      "step": 27108
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7265001270759805e-08,
+      "loss": 0.5522,
+      "step": 27109
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.719616763496412e-08,
+      "loss": 0.5861,
+      "step": 27110
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.712747137131987e-08,
+      "loss": 0.4969,
+      "step": 27111
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.705891248077407e-08,
+      "loss": 0.6451,
+      "step": 27112
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.69904909642693e-08,
+      "loss": 0.5838,
+      "step": 27113
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6922206822747034e-08,
+      "loss": 0.6041,
+      "step": 27114
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6854060057146516e-08,
+      "loss": 0.593,
+      "step": 27115
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6786050668406995e-08,
+      "loss": 0.5665,
+      "step": 27116
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.671817865746328e-08,
+      "loss": 0.6413,
+      "step": 27117
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.665044402524907e-08,
+      "loss": 0.5266,
+      "step": 27118
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6582846772696946e-08,
+      "loss": 0.6052,
+      "step": 27119
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6515386900738396e-08,
+      "loss": 0.563,
+      "step": 27120
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6448064410299335e-08,
+      "loss": 0.5497,
+      "step": 27121
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6380879302306807e-08,
+      "loss": 0.5505,
+      "step": 27122
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6313831577685624e-08,
+      "loss": 0.627,
+      "step": 27123
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.624692123735949e-08,
+      "loss": 0.5275,
+      "step": 27124
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6180148282247678e-08,
+      "loss": 0.6076,
+      "step": 27125
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6113512713269443e-08,
+      "loss": 0.5679,
+      "step": 27126
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.604701453134183e-08,
+      "loss": 0.5316,
+      "step": 27127
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.598065373738078e-08,
+      "loss": 0.5504,
+      "step": 27128
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5914430332297782e-08,
+      "loss": 0.572,
+      "step": 27129
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5848344317005437e-08,
+      "loss": 0.5657,
+      "step": 27130
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.578239569241302e-08,
+      "loss": 0.5998,
+      "step": 27131
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5716584459427588e-08,
+      "loss": 0.59,
+      "step": 27132
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5650910618953964e-08,
+      "loss": 0.5441,
+      "step": 27133
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5585374171898094e-08,
+      "loss": 0.6013,
+      "step": 27134
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.551997511915926e-08,
+      "loss": 0.5725,
+      "step": 27135
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.545471346164007e-08,
+      "loss": 0.5898,
+      "step": 27136
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5389589200236477e-08,
+      "loss": 0.5826,
+      "step": 27137
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5324602335845538e-08,
+      "loss": 0.5574,
+      "step": 27138
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.525975286936099e-08,
+      "loss": 0.5215,
+      "step": 27139
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5195040801675442e-08,
+      "loss": 0.5652,
+      "step": 27140
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.513046613368041e-08,
+      "loss": 0.5957,
+      "step": 27141
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.506602886626296e-08,
+      "loss": 0.5641,
+      "step": 27142
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5001729000311272e-08,
+      "loss": 0.5409,
+      "step": 27143
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.493756653670908e-08,
+      "loss": 0.5806,
+      "step": 27144
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4873541476339016e-08,
+      "loss": 0.6511,
+      "step": 27145
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4809653820083703e-08,
+      "loss": 0.6153,
+      "step": 27146
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4745903568820218e-08,
+      "loss": 0.5008,
+      "step": 27147
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4682290723427861e-08,
+      "loss": 0.545,
+      "step": 27148
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4618815284780374e-08,
+      "loss": 0.5757,
+      "step": 27149
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4555477253752615e-08,
+      "loss": 0.5649,
+      "step": 27150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4492276631214996e-08,
+      "loss": 0.5074,
+      "step": 27151
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4429213418036825e-08,
+      "loss": 0.6098,
+      "step": 27152
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4366287615088514e-08,
+      "loss": 0.5494,
+      "step": 27153
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4303499223232709e-08,
+      "loss": 0.5061,
+      "step": 27154
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4240848243335381e-08,
+      "loss": 0.6085,
+      "step": 27155
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4178334676259176e-08,
+      "loss": 0.6154,
+      "step": 27156
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4115958522862295e-08,
+      "loss": 0.5386,
+      "step": 27157
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.405371978400516e-08,
+      "loss": 0.5602,
+      "step": 27158
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3991618460542644e-08,
+      "loss": 0.6035,
+      "step": 27159
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3929654553329618e-08,
+      "loss": 0.5428,
+      "step": 27160
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3867828063219846e-08,
+      "loss": 0.5784,
+      "step": 27161
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3806138991062645e-08,
+      "loss": 0.5723,
+      "step": 27162
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3744587337708448e-08,
+      "loss": 0.5723,
+      "step": 27163
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3683173104002134e-08,
+      "loss": 0.6069,
+      "step": 27164
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3621896290790804e-08,
+      "loss": 0.5474,
+      "step": 27165
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3560756898917116e-08,
+      "loss": 0.622,
+      "step": 27166
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.349975492922262e-08,
+      "loss": 0.5524,
+      "step": 27167
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3438890382545533e-08,
+      "loss": 0.5169,
+      "step": 27168
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3378163259725185e-08,
+      "loss": 0.5456,
+      "step": 27169
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3317573561595354e-08,
+      "loss": 0.5766,
+      "step": 27170
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3257121288992036e-08,
+      "loss": 0.5199,
+      "step": 27171
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.319680644274457e-08,
+      "loss": 0.4651,
+      "step": 27172
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.313662902368451e-08,
+      "loss": 0.5285,
+      "step": 27173
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3076589032640085e-08,
+      "loss": 0.5983,
+      "step": 27174
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.30166864704373e-08,
+      "loss": 0.5682,
+      "step": 27175
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2956921337899941e-08,
+      "loss": 0.6195,
+      "step": 27176
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2897293635850683e-08,
+      "loss": 0.5735,
+      "step": 27177
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2837803365109979e-08,
+      "loss": 0.5582,
+      "step": 27178
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2778450526496067e-08,
+      "loss": 0.5578,
+      "step": 27179
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2719235120827179e-08,
+      "loss": 0.5215,
+      "step": 27180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2660157148915997e-08,
+      "loss": 0.5607,
+      "step": 27181
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2601216611578538e-08,
+      "loss": 0.5879,
+      "step": 27182
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2542413509623042e-08,
+      "loss": 0.6674,
+      "step": 27183
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2483747843858862e-08,
+      "loss": 0.5735,
+      "step": 27184
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2425219615095352e-08,
+      "loss": 0.5916,
+      "step": 27185
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2366828824136312e-08,
+      "loss": 0.6263,
+      "step": 27186
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2308575471785544e-08,
+      "loss": 0.536,
+      "step": 27187
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.225045955884463e-08,
+      "loss": 0.5548,
+      "step": 27188
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2192481086112928e-08,
+      "loss": 0.5442,
+      "step": 27189
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.21346400543898e-08,
+      "loss": 0.5651,
+      "step": 27190
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2076936464469058e-08,
+      "loss": 0.581,
+      "step": 27191
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2019370317146728e-08,
+      "loss": 0.6271,
+      "step": 27192
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.19619416132144e-08,
+      "loss": 0.5416,
+      "step": 27193
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1904650353461445e-08,
+      "loss": 0.5235,
+      "step": 27194
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1847496538676118e-08,
+      "loss": 0.5948,
+      "step": 27195
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.179048016964668e-08,
+      "loss": 0.5087,
+      "step": 27196
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1733601247155834e-08,
+      "loss": 0.5378,
+      "step": 27197
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1676859771987404e-08,
+      "loss": 0.6132,
+      "step": 27198
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1620255744921871e-08,
+      "loss": 0.5305,
+      "step": 27199
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1563789166738615e-08,
+      "loss": 0.5918,
+      "step": 27200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.150746003821368e-08,
+      "loss": 0.6015,
+      "step": 27201
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1451268360123114e-08,
+      "loss": 0.6076,
+      "step": 27202
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.139521413323963e-08,
+      "loss": 0.5965,
+      "step": 27203
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1339297358334833e-08,
+      "loss": 0.5727,
+      "step": 27204
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1283518036177e-08,
+      "loss": 0.5833,
+      "step": 27205
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1227876167536622e-08,
+      "loss": 0.5832,
+      "step": 27206
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1172371753176426e-08,
+      "loss": 0.5805,
+      "step": 27207
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1117004793861353e-08,
+      "loss": 0.5098,
+      "step": 27208
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1061775290354126e-08,
+      "loss": 0.5813,
+      "step": 27209
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.100668324341303e-08,
+      "loss": 0.5839,
+      "step": 27210
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0951728653797455e-08,
+      "loss": 0.5533,
+      "step": 27211
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0896911522263465e-08,
+      "loss": 0.5978,
+      "step": 27212
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0842231849564899e-08,
+      "loss": 0.6078,
+      "step": 27213
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.078768963645449e-08,
+      "loss": 0.5002,
+      "step": 27214
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0733284883682748e-08,
+      "loss": 0.5606,
+      "step": 27215
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0679017591999074e-08,
+      "loss": 0.5648,
+      "step": 27216
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0624887762148428e-08,
+      "loss": 0.6164,
+      "step": 27217
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0570895394876879e-08,
+      "loss": 0.5907,
+      "step": 27218
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0517040490928276e-08,
+      "loss": 0.5976,
+      "step": 27219
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.046332305104203e-08,
+      "loss": 0.5748,
+      "step": 27220
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0409743075957545e-08,
+      "loss": 0.5284,
+      "step": 27221
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0356300566413124e-08,
+      "loss": 0.5441,
+      "step": 27222
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0302995523143733e-08,
+      "loss": 0.6265,
+      "step": 27223
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0249827946882118e-08,
+      "loss": 0.5571,
+      "step": 27224
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0196797838362137e-08,
+      "loss": 0.5944,
+      "step": 27225
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0143905198310987e-08,
+      "loss": 0.6377,
+      "step": 27226
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0091150027456976e-08,
+      "loss": 0.5437,
+      "step": 27227
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0038532326528405e-08,
+      "loss": 0.5957,
+      "step": 27228
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.986052096245814e-09,
+      "loss": 0.5991,
+      "step": 27229
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.933709337334175e-09,
+      "loss": 0.5673,
+      "step": 27230
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.881504050512913e-09,
+      "loss": 0.5443,
+      "step": 27231
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.829436236499235e-09,
+      "loss": 0.5609,
+      "step": 27232
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.777505896011452e-09,
+      "loss": 0.5693,
+      "step": 27233
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.725713029763439e-09,
+      "loss": 0.5713,
+      "step": 27234
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.67405763846685e-09,
+      "loss": 0.5964,
+      "step": 27235
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.622539722834445e-09,
+      "loss": 0.5534,
+      "step": 27236
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.571159283574548e-09,
+      "loss": 0.6393,
+      "step": 27237
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.519916321392153e-09,
+      "loss": 0.5875,
+      "step": 27238
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.46881083699558e-09,
+      "loss": 0.5639,
+      "step": 27239
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.417842831086488e-09,
+      "loss": 0.4916,
+      "step": 27240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.36701230436654e-09,
+      "loss": 0.6308,
+      "step": 27241
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.316319257534067e-09,
+      "loss": 0.6231,
+      "step": 27242
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.26576369128851e-09,
+      "loss": 0.5731,
+      "step": 27243
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.215345606323756e-09,
+      "loss": 0.5628,
+      "step": 27244
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.165065003335916e-09,
+      "loss": 0.5794,
+      "step": 27245
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.114921883014438e-09,
+      "loss": 0.5685,
+      "step": 27246
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.064916246049881e-09,
+      "loss": 0.5869,
+      "step": 27247
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.015048093131695e-09,
+      "loss": 0.531,
+      "step": 27248
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.965317424945997e-09,
+      "loss": 0.5675,
+      "step": 27249
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.915724242175572e-09,
+      "loss": 0.627,
+      "step": 27250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.86626854550321e-09,
+      "loss": 0.6024,
+      "step": 27251
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.816950335610586e-09,
+      "loss": 0.661,
+      "step": 27252
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.767769613176046e-09,
+      "loss": 0.5495,
+      "step": 27253
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.718726378875719e-09,
+      "loss": 0.5812,
+      "step": 27254
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.669820633384617e-09,
+      "loss": 0.5477,
+      "step": 27255
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.621052377376648e-09,
+      "loss": 0.5659,
+      "step": 27256
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.572421611521275e-09,
+      "loss": 0.5515,
+      "step": 27257
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.523928336489074e-09,
+      "loss": 0.5718,
+      "step": 27258
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.475572552947287e-09,
+      "loss": 0.5719,
+      "step": 27259
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.427354261560939e-09,
+      "loss": 0.6075,
+      "step": 27260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.379273462992832e-09,
+      "loss": 0.533,
+      "step": 27261
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.331330157905771e-09,
+      "loss": 0.5433,
+      "step": 27262
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.283524346959226e-09,
+      "loss": 0.561,
+      "step": 27263
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.235856030810452e-09,
+      "loss": 0.5717,
+      "step": 27264
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.188325210115588e-09,
+      "loss": 0.5849,
+      "step": 27265
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.140931885529668e-09,
+      "loss": 0.6034,
+      "step": 27266
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.093676057703281e-09,
+      "loss": 0.5884,
+      "step": 27267
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.04655772728813e-09,
+      "loss": 0.569,
+      "step": 27268
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.999576894932582e-09,
+      "loss": 0.5462,
+      "step": 27269
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.952733561281679e-09,
+      "loss": 0.5752,
+      "step": 27270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.906027726981568e-09,
+      "loss": 0.574,
+      "step": 27271
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.859459392673963e-09,
+      "loss": 0.5482,
+      "step": 27272
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.813028559000569e-09,
+      "loss": 0.6021,
+      "step": 27273
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.766735226598654e-09,
+      "loss": 0.5328,
+      "step": 27274
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.720579396107708e-09,
+      "loss": 0.5148,
+      "step": 27275
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.674561068160558e-09,
+      "loss": 0.5743,
+      "step": 27276
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.62868024339225e-09,
+      "loss": 0.6724,
+      "step": 27277
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.582936922432282e-09,
+      "loss": 0.5689,
+      "step": 27278
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.53733110591237e-09,
+      "loss": 0.5909,
+      "step": 27279
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.491862794458682e-09,
+      "loss": 0.5588,
+      "step": 27280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.446531988697381e-09,
+      "loss": 0.5433,
+      "step": 27281
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.4013386892513024e-09,
+      "loss": 0.6248,
+      "step": 27282
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.356282896743283e-09,
+      "loss": 0.6172,
+      "step": 27283
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.311364611793937e-09,
+      "loss": 0.5967,
+      "step": 27284
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.266583835019436e-09,
+      "loss": 0.5396,
+      "step": 27285
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.221940567037067e-09,
+      "loss": 0.4905,
+      "step": 27286
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.177434808461891e-09,
+      "loss": 0.591,
+      "step": 27287
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.133066559905644e-09,
+      "loss": 0.5453,
+      "step": 27288
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.088835821978945e-09,
+      "loss": 0.6319,
+      "step": 27289
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.044742595290199e-09,
+      "loss": 0.5482,
+      "step": 27290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.000786880446697e-09,
+      "loss": 0.5547,
+      "step": 27291
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.9569686780524e-09,
+      "loss": 0.5629,
+      "step": 27292
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.9132879887112706e-09,
+      "loss": 0.5258,
+      "step": 27293
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.8697448130239374e-09,
+      "loss": 0.5437,
+      "step": 27294
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.826339151589922e-09,
+      "loss": 0.6338,
+      "step": 27295
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.783071005005415e-09,
+      "loss": 0.5946,
+      "step": 27296
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.7399403738677146e-09,
+      "loss": 0.6099,
+      "step": 27297
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.696947258768571e-09,
+      "loss": 0.5969,
+      "step": 27298
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.654091660299733e-09,
+      "loss": 0.5481,
+      "step": 27299
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.6113735790529484e-09,
+      "loss": 0.5728,
+      "step": 27300
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.568793015613306e-09,
+      "loss": 0.6153,
+      "step": 27301
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.526349970569223e-09,
+      "loss": 0.5818,
+      "step": 27302
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.484044444502457e-09,
+      "loss": 0.564,
+      "step": 27303
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.441876437996986e-09,
+      "loss": 0.5948,
+      "step": 27304
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.399845951633454e-09,
+      "loss": 0.5194,
+      "step": 27305
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.35795298598807e-09,
+      "loss": 0.5394,
+      "step": 27306
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.316197541638147e-09,
+      "loss": 0.559,
+      "step": 27307
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.2745796191587826e-09,
+      "loss": 0.5832,
+      "step": 27308
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.233099219122851e-09,
+      "loss": 0.5797,
+      "step": 27309
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.191756342099897e-09,
+      "loss": 0.5549,
+      "step": 27310
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.1505509886605755e-09,
+      "loss": 0.5744,
+      "step": 27311
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.109483159371099e-09,
+      "loss": 0.5463,
+      "step": 27312
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.068552854795462e-09,
+      "loss": 0.5516,
+      "step": 27313
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.027760075498768e-09,
+      "loss": 0.5491,
+      "step": 27314
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.987104822041678e-09,
+      "loss": 0.5837,
+      "step": 27315
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.946587094983747e-09,
+      "loss": 0.5727,
+      "step": 27316
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.9062068948811945e-09,
+      "loss": 0.6263,
+      "step": 27317
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.865964222291354e-09,
+      "loss": 0.5714,
+      "step": 27318
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.825859077768226e-09,
+      "loss": 0.5161,
+      "step": 27319
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.78589146186248e-09,
+      "loss": 0.5421,
+      "step": 27320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.746061375123679e-09,
+      "loss": 0.5769,
+      "step": 27321
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.706368818101382e-09,
+      "loss": 0.5882,
+      "step": 27322
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.666813791341818e-09,
+      "loss": 0.5479,
+      "step": 27323
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.627396295387888e-09,
+      "loss": 0.589,
+      "step": 27324
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.588116330783599e-09,
+      "loss": 0.5005,
+      "step": 27325
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.548973898068521e-09,
+      "loss": 0.5913,
+      "step": 27326
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.509968997781112e-09,
+      "loss": 0.5922,
+      "step": 27327
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.471101630458719e-09,
+      "loss": 0.6221,
+      "step": 27328
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.432371796635361e-09,
+      "loss": 0.5091,
+      "step": 27329
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.393779496845053e-09,
+      "loss": 0.5726,
+      "step": 27330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.355324731618483e-09,
+      "loss": 0.6121,
+      "step": 27331
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.317007501484117e-09,
+      "loss": 0.5722,
+      "step": 27332
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.278827806970421e-09,
+      "loss": 0.5151,
+      "step": 27333
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.240785648601421e-09,
+      "loss": 0.5423,
+      "step": 27334
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.202881026902251e-09,
+      "loss": 0.5871,
+      "step": 27335
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.165113942393607e-09,
+      "loss": 0.5176,
+      "step": 27336
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.127484395593962e-09,
+      "loss": 0.6404,
+      "step": 27337
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.089992387024012e-09,
+      "loss": 0.5817,
+      "step": 27338
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.052637917197789e-09,
+      "loss": 0.6163,
+      "step": 27339
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.015420986628217e-09,
+      "loss": 0.5897,
+      "step": 27340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.978341595830438e-09,
+      "loss": 0.5113,
+      "step": 27341
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.941399745311826e-09,
+      "loss": 0.577,
+      "step": 27342
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.9045954355830815e-09,
+      "loss": 0.5302,
+      "step": 27343
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.867928667149357e-09,
+      "loss": 0.5489,
+      "step": 27344
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.831399440514695e-09,
+      "loss": 0.5316,
+      "step": 27345
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.795007756183135e-09,
+      "loss": 0.5514,
+      "step": 27346
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.758753614654277e-09,
+      "loss": 0.5114,
+      "step": 27347
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.722637016427722e-09,
+      "loss": 0.6161,
+      "step": 27348
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6866579620008515e-09,
+      "loss": 0.5518,
+      "step": 27349
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.650816451867713e-09,
+      "loss": 0.5774,
+      "step": 27350
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.615112486522355e-09,
+      "loss": 0.5248,
+      "step": 27351
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.579546066455498e-09,
+      "loss": 0.5884,
+      "step": 27352
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.54411719215675e-09,
+      "loss": 0.5451,
+      "step": 27353
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.508825864114608e-09,
+      "loss": 0.5787,
+      "step": 27354
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.47367208281313e-09,
+      "loss": 0.6291,
+      "step": 27355
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.438655848736373e-09,
+      "loss": 0.6247,
+      "step": 27356
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.403777162367284e-09,
+      "loss": 0.5907,
+      "step": 27357
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.36903602418437e-09,
+      "loss": 0.5346,
+      "step": 27358
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.334432434667246e-09,
+      "loss": 0.5285,
+      "step": 27359
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.299966394291089e-09,
+      "loss": 0.5501,
+      "step": 27360
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.265637903529962e-09,
+      "loss": 0.5627,
+      "step": 27361
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.231446962856822e-09,
+      "loss": 0.5994,
+      "step": 27362
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.197393572742403e-09,
+      "loss": 0.56,
+      "step": 27363
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.163477733654109e-09,
+      "loss": 0.5333,
+      "step": 27364
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.129699446060454e-09,
+      "loss": 0.5796,
+      "step": 27365
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0960587104244e-09,
+      "loss": 0.5408,
+      "step": 27366
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.062555527210021e-09,
+      "loss": 0.5154,
+      "step": 27367
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.029189896879171e-09,
+      "loss": 0.5555,
+      "step": 27368
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.99596181988815e-09,
+      "loss": 0.5724,
+      "step": 27369
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.962871296696591e-09,
+      "loss": 0.5235,
+      "step": 27370
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.929918327759685e-09,
+      "loss": 0.513,
+      "step": 27371
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.897102913530404e-09,
+      "loss": 0.5802,
+      "step": 27372
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8644250544594975e-09,
+      "loss": 0.5725,
+      "step": 27373
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.831884750998827e-09,
+      "loss": 0.5931,
+      "step": 27374
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.799482003593591e-09,
+      "loss": 0.543,
+      "step": 27375
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.76721681269121e-09,
+      "loss": 0.5469,
+      "step": 27376
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.735089178735773e-09,
+      "loss": 0.5404,
+      "step": 27377
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.70309910216915e-09,
+      "loss": 0.5776,
+      "step": 27378
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6712465834309874e-09,
+      "loss": 0.6046,
+      "step": 27379
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.639531622959824e-09,
+      "loss": 0.5529,
+      "step": 27380
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6079542211919783e-09,
+      "loss": 0.5025,
+      "step": 27381
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.576514378563767e-09,
+      "loss": 0.6249,
+      "step": 27382
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5452120955048465e-09,
+      "loss": 0.5739,
+      "step": 27383
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.514047372448204e-09,
+      "loss": 0.559,
+      "step": 27384
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.483020209821275e-09,
+      "loss": 0.6584,
+      "step": 27385
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.452130608052606e-09,
+      "loss": 0.5433,
+      "step": 27386
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4213785675651923e-09,
+      "loss": 0.6168,
+      "step": 27387
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3907640887842486e-09,
+      "loss": 0.5518,
+      "step": 27388
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.36028717213055e-09,
+      "loss": 0.5918,
+      "step": 27389
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.32994781802265e-09,
+      "loss": 0.5839,
+      "step": 27390
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2997460268779923e-09,
+      "loss": 0.5187,
+      "step": 27391
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2696817991129116e-09,
+      "loss": 0.5046,
+      "step": 27392
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.23975513514152e-09,
+      "loss": 0.6066,
+      "step": 27393
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2099660353746008e-09,
+      "loss": 0.5886,
+      "step": 27394
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1803145002229363e-09,
+      "loss": 0.5258,
+      "step": 27395
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.150800530093978e-09,
+      "loss": 0.6355,
+      "step": 27396
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1214241253929578e-09,
+      "loss": 0.5542,
+      "step": 27397
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.092185286526217e-09,
+      "loss": 0.5572,
+      "step": 27398
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0630840138945462e-09,
+      "loss": 0.5995,
+      "step": 27399
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.034120307899846e-09,
+      "loss": 0.5922,
+      "step": 27400
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.005294168938466e-09,
+      "loss": 0.4795,
+      "step": 27401
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.976605597408977e-09,
+      "loss": 0.5091,
+      "step": 27402
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9480545937055073e-09,
+      "loss": 0.5637,
+      "step": 27403
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.919641158221076e-09,
+      "loss": 0.555,
+      "step": 27404
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8913652913464817e-09,
+      "loss": 0.544,
+      "step": 27405
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.863226993471413e-09,
+      "loss": 0.5418,
+      "step": 27406
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.835226264982227e-09,
+      "loss": 0.5621,
+      "step": 27407
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8073631062641716e-09,
+      "loss": 0.5271,
+      "step": 27408
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7796375177013836e-09,
+      "loss": 0.6054,
+      "step": 27409
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7520494996757796e-09,
+      "loss": 0.5951,
+      "step": 27410
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.724599052565946e-09,
+      "loss": 0.6254,
+      "step": 27411
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6972861767493587e-09,
+      "loss": 0.5479,
+      "step": 27412
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.670110872603493e-09,
+      "loss": 0.5756,
+      "step": 27413
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6430731405002743e-09,
+      "loss": 0.5935,
+      "step": 27414
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.616172980813847e-09,
+      "loss": 0.5497,
+      "step": 27415
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5894103939128058e-09,
+      "loss": 0.581,
+      "step": 27416
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5627853801657443e-09,
+      "loss": 0.6116,
+      "step": 27417
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.536297939940147e-09,
+      "loss": 0.5503,
+      "step": 27418
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5099480735990557e-09,
+      "loss": 0.5935,
+      "step": 27419
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.483735781505514e-09,
+      "loss": 0.6301,
+      "step": 27420
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.457661064020345e-09,
+      "loss": 0.5491,
+      "step": 27421
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4317239215021493e-09,
+      "loss": 0.5457,
+      "step": 27422
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.40592435430842e-09,
+      "loss": 0.5504,
+      "step": 27423
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.380262362794428e-09,
+      "loss": 0.5674,
+      "step": 27424
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3547379473121137e-09,
+      "loss": 0.625,
+      "step": 27425
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.329351108212308e-09,
+      "loss": 0.5536,
+      "step": 27426
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3041018458469512e-09,
+      "loss": 0.5865,
+      "step": 27427
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.278990160561323e-09,
+      "loss": 0.5761,
+      "step": 27428
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.254016052701813e-09,
+      "loss": 0.5446,
+      "step": 27429
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.22917952261148e-09,
+      "loss": 0.5681,
+      "step": 27430
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2044805706322725e-09,
+      "loss": 0.6304,
+      "step": 27431
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.179919197103919e-09,
+      "loss": 0.5453,
+      "step": 27432
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1554954023661477e-09,
+      "loss": 0.5962,
+      "step": 27433
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1312091867520257e-09,
+      "loss": 0.5588,
+      "step": 27434
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1070605505979504e-09,
+      "loss": 0.6126,
+      "step": 27435
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.083049494235878e-09,
+      "loss": 0.5899,
+      "step": 27436
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0591760179966558e-09,
+      "loss": 0.5478,
+      "step": 27437
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.035440122206689e-09,
+      "loss": 0.4766,
+      "step": 27438
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0118418071946033e-09,
+      "loss": 0.5149,
+      "step": 27439
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.988381073284584e-09,
+      "loss": 0.644,
+      "step": 27440
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9650579207997067e-09,
+      "loss": 0.5552,
+      "step": 27441
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.941872350060825e-09,
+      "loss": 0.6045,
+      "step": 27442
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9188243613865733e-09,
+      "loss": 0.6182,
+      "step": 27443
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.895913955094475e-09,
+      "loss": 0.5333,
+      "step": 27444
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.873141131499834e-09,
+      "loss": 0.588,
+      "step": 27445
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8505058909157325e-09,
+      "loss": 0.6022,
+      "step": 27446
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8280082336530336e-09,
+      "loss": 0.544,
+      "step": 27447
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8056481600237096e-09,
+      "loss": 0.5387,
+      "step": 27448
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.783425670331962e-09,
+      "loss": 0.5979,
+      "step": 27449
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7613407648875426e-09,
+      "loss": 0.5772,
+      "step": 27450
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7393934439913218e-09,
+      "loss": 0.6133,
+      "step": 27451
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7175837079452806e-09,
+      "loss": 0.5403,
+      "step": 27452
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6959115570513996e-09,
+      "loss": 0.5475,
+      "step": 27453
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6743769916072183e-09,
+      "loss": 0.5929,
+      "step": 27454
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6529800119091666e-09,
+      "loss": 0.5781,
+      "step": 27455
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6317206182514534e-09,
+      "loss": 0.5558,
+      "step": 27456
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6105988109260672e-09,
+      "loss": 0.6393,
+      "step": 27457
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5896145902238868e-09,
+      "loss": 0.5947,
+      "step": 27458
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5687679564346803e-09,
+      "loss": 0.6036,
+      "step": 27459
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.548058909844885e-09,
+      "loss": 0.5908,
+      "step": 27460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5274874507387183e-09,
+      "loss": 0.5948,
+      "step": 27461
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5070535793992869e-09,
+      "loss": 0.5813,
+      "step": 27462
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4867572961085874e-09,
+      "loss": 0.567,
+      "step": 27463
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.466598601145286e-09,
+      "loss": 0.5534,
+      "step": 27464
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4465774947880485e-09,
+      "loss": 0.5991,
+      "step": 27465
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4266939773111e-09,
+      "loss": 0.5923,
+      "step": 27466
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4069480489875553e-09,
+      "loss": 0.5604,
+      "step": 27467
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3873397100905296e-09,
+      "loss": 0.5577,
+      "step": 27468
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3678689608898065e-09,
+      "loss": 0.5318,
+      "step": 27469
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3485358016518403e-09,
+      "loss": 0.5929,
+      "step": 27470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.329340232644194e-09,
+      "loss": 0.6424,
+      "step": 27471
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3102822541299908e-09,
+      "loss": 0.5738,
+      "step": 27472
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2913618663712435e-09,
+      "loss": 0.6012,
+      "step": 27473
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2725790696299644e-09,
+      "loss": 0.5503,
+      "step": 27474
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2539338641637256e-09,
+      "loss": 0.5854,
+      "step": 27475
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2354262502289882e-09,
+      "loss": 0.5732,
+      "step": 27476
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2170562280799935e-09,
+      "loss": 0.6819,
+      "step": 27477
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1988237979709826e-09,
+      "loss": 0.5888,
+      "step": 27478
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1807289601506455e-09,
+      "loss": 0.5555,
+      "step": 27479
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1627717148698926e-09,
+      "loss": 0.6453,
+      "step": 27480
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.144952062375193e-09,
+      "loss": 0.6046,
+      "step": 27481
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1272700029119065e-09,
+      "loss": 0.6009,
+      "step": 27482
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1097255367220615e-09,
+      "loss": 0.5343,
+      "step": 27483
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.092318664048797e-09,
+      "loss": 0.6004,
+      "step": 27484
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.075049385131921e-09,
+      "loss": 0.5805,
+      "step": 27485
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0579177002068008e-09,
+      "loss": 0.6465,
+      "step": 27486
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0409236095121344e-09,
+      "loss": 0.5834,
+      "step": 27487
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0240671132788483e-09,
+      "loss": 0.6363,
+      "step": 27488
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.007348211740089e-09,
+      "loss": 0.5752,
+      "step": 27489
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.907669051267833e-10,
+      "loss": 0.5957,
+      "step": 27490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.743231936665264e-10,
+      "loss": 0.5864,
+      "step": 27491
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.580170775858045e-10,
+      "loss": 0.5005,
+      "step": 27492
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.418485571077719e-10,
+      "loss": 0.5896,
+      "step": 27493
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.258176324566936e-10,
+      "loss": 0.5188,
+      "step": 27494
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.099243038523942e-10,
+      "loss": 0.5651,
+      "step": 27495
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.941685715124771e-10,
+      "loss": 0.4925,
+      "step": 27496
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.785504356556562e-10,
+      "loss": 0.5613,
+      "step": 27497
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.630698964950946e-10,
+      "loss": 0.6063,
+      "step": 27498
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.47726954245065e-10,
+      "loss": 0.5519,
+      "step": 27499
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.325216091165101e-10,
+      "loss": 0.5578,
+      "step": 27500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.174538613192617e-10,
+      "loss": 0.5971,
+      "step": 27501
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.025237110587114e-10,
+      "loss": 0.5574,
+      "step": 27502
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.877311585424708e-10,
+      "loss": 0.5221,
+      "step": 27503
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.730762039726004e-10,
+      "loss": 0.5347,
+      "step": 27504
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.585588475511607e-10,
+      "loss": 0.5148,
+      "step": 27505
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.441790894779922e-10,
+      "loss": 0.6002,
+      "step": 27506
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.299369299518244e-10,
+      "loss": 0.5407,
+      "step": 27507
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.158323691669466e-10,
+      "loss": 0.5926,
+      "step": 27508
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.01865407318758e-10,
+      "loss": 0.6449,
+      "step": 27509
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.880360445993273e-10,
+      "loss": 0.5702,
+      "step": 27510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.743442811973921e-10,
+      "loss": 0.5226,
+      "step": 27511
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.607901173039111e-10,
+      "loss": 0.5669,
+      "step": 27512
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.473735531042913e-10,
+      "loss": 0.543,
+      "step": 27513
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.340945887828298e-10,
+      "loss": 0.5593,
+      "step": 27514
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.209532245216032e-10,
+      "loss": 0.5506,
+      "step": 27515
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.079494605026881e-10,
+      "loss": 0.5826,
+      "step": 27516
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.950832969048303e-10,
+      "loss": 0.5322,
+      "step": 27517
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.823547339045555e-10,
+      "loss": 0.6089,
+      "step": 27518
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.697637716783888e-10,
+      "loss": 0.6158,
+      "step": 27519
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.573104103973049e-10,
+      "loss": 0.5354,
+      "step": 27520
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.449946502344983e-10,
+      "loss": 0.5813,
+      "step": 27521
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.328164913587231e-10,
+      "loss": 0.6023,
+      "step": 27522
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.207759339376228e-10,
+      "loss": 0.5967,
+      "step": 27523
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.088729781366208e-10,
+      "loss": 0.6042,
+      "step": 27524
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.971076241211404e-10,
+      "loss": 0.5907,
+      "step": 27525
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.854798720510534e-10,
+      "loss": 0.598,
+      "step": 27526
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.739897220862321e-10,
+      "loss": 0.567,
+      "step": 27527
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.6263717438654874e-10,
+      "loss": 0.5366,
+      "step": 27528
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.5142222910743436e-10,
+      "loss": 0.5798,
+      "step": 27529
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.4034488640321006e-10,
+      "loss": 0.5531,
+      "step": 27530
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.294051464259763e-10,
+      "loss": 0.5541,
+      "step": 27531
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.1860300932561324e-10,
+      "loss": 0.5615,
+      "step": 27532
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.0793847525311124e-10,
+      "loss": 0.523,
+      "step": 27533
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.974115443527993e-10,
+      "loss": 0.5584,
+      "step": 27534
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8702221677122677e-10,
+      "loss": 0.5997,
+      "step": 27535
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.7677049265050226e-10,
+      "loss": 0.6126,
+      "step": 27536
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.666563721316241e-10,
+      "loss": 0.5243,
+      "step": 27537
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.5667985535337014e-10,
+      "loss": 0.5496,
+      "step": 27538
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.468409424545183e-10,
+      "loss": 0.5781,
+      "step": 27539
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.3713963356940547e-10,
+      "loss": 0.5957,
+      "step": 27540
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.275759288323688e-10,
+      "loss": 0.6104,
+      "step": 27541
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.181498283733042e-10,
+      "loss": 0.5395,
+      "step": 27542
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.088613323232181e-10,
+      "loss": 0.6163,
+      "step": 27543
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.997104408097862e-10,
+      "loss": 0.6217,
+      "step": 27544
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9069715395957376e-10,
+      "loss": 0.5818,
+      "step": 27545
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8182147189470543e-10,
+      "loss": 0.5933,
+      "step": 27546
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.730833947395262e-10,
+      "loss": 0.4439,
+      "step": 27547
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.644829226128298e-10,
+      "loss": 0.5753,
+      "step": 27548
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5602005563341024e-10,
+      "loss": 0.5491,
+      "step": 27549
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.476947939178409e-10,
+      "loss": 0.5824,
+      "step": 27550
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.395071375793645e-10,
+      "loss": 0.5002,
+      "step": 27551
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3145708673344425e-10,
+      "loss": 0.5493,
+      "step": 27552
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2354464148888199e-10,
+      "loss": 0.5838,
+      "step": 27553
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.157698019544796e-10,
+      "loss": 0.5801,
+      "step": 27554
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0813256823792872e-10,
+      "loss": 0.5258,
+      "step": 27555
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0063294044359029e-10,
+      "loss": 0.5607,
+      "step": 27556
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9327091867582527e-10,
+      "loss": 0.5546,
+      "step": 27557
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8604650303455374e-10,
+      "loss": 0.6077,
+      "step": 27558
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.78959693620806e-10,
+      "loss": 0.6311,
+      "step": 27559
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7201049053006123e-10,
+      "loss": 0.5334,
+      "step": 27560
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6519889386001907e-10,
+      "loss": 0.6332,
+      "step": 27561
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5852490370282802e-10,
+      "loss": 0.5298,
+      "step": 27562
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.519885201517468e-10,
+      "loss": 0.5731,
+      "step": 27563
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4558974329448304e-10,
+      "loss": 0.5842,
+      "step": 27564
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3932857322207504e-10,
+      "loss": 0.581,
+      "step": 27565
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3320501001889973e-10,
+      "loss": 0.5233,
+      "step": 27566
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2721905376822385e-10,
+      "loss": 0.5006,
+      "step": 27567
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2137070455553456e-10,
+      "loss": 0.5755,
+      "step": 27568
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1565996245854749e-10,
+      "loss": 0.5941,
+      "step": 27569
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1008682755608847e-10,
+      "loss": 0.6078,
+      "step": 27570
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0465129992587308e-10,
+      "loss": 0.6379,
+      "step": 27571
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.935337964339653e-11,
+      "loss": 0.4982,
+      "step": 27572
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.419306677971307e-11,
+      "loss": 0.5808,
+      "step": 27573
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.917036140587699e-11,
+      "loss": 0.6013,
+      "step": 27574
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.428526359294254e-11,
+      "loss": 0.5726,
+      "step": 27575
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.953777340641289e-11,
+      "loss": 0.5929,
+      "step": 27576
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.49278909117912e-11,
+      "loss": 0.5395,
+      "step": 27577
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.045561617347041e-11,
+      "loss": 0.6042,
+      "step": 27578
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.612094925251278e-11,
+      "loss": 0.552,
+      "step": 27579
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.192389020776012e-11,
+      "loss": 0.5691,
+      "step": 27580
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.786443909805428e-11,
+      "loss": 0.5548,
+      "step": 27581
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.3942595978906386e-11,
+      "loss": 0.5407,
+      "step": 27582
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.015836090360715e-11,
+      "loss": 0.5507,
+      "step": 27583
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.651173392544728e-11,
+      "loss": 0.5404,
+      "step": 27584
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3002715094386804e-11,
+      "loss": 0.5081,
+      "step": 27585
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.963130445816532e-11,
+      "loss": 0.6094,
+      "step": 27586
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.639750206230197e-11,
+      "loss": 0.5617,
+      "step": 27587
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.330130795342612e-11,
+      "loss": 0.6317,
+      "step": 27588
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.034272217261602e-11,
+      "loss": 0.5615,
+      "step": 27589
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7521744760949932e-11,
+      "loss": 0.5491,
+      "step": 27590
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4838375758395873e-11,
+      "loss": 0.5633,
+      "step": 27591
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.229261519937076e-11,
+      "loss": 0.5274,
+      "step": 27592
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9884463121622176e-11,
+      "loss": 0.6206,
+      "step": 27593
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.761391955623637e-11,
+      "loss": 0.5818,
+      "step": 27594
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5480984535409803e-11,
+      "loss": 0.5843,
+      "step": 27595
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.348565808800828e-11,
+      "loss": 0.6411,
+      "step": 27596
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1627940241787372e-11,
+      "loss": 0.5286,
+      "step": 27597
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.907831022282211e-12,
+      "loss": 0.5706,
+      "step": 27598
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.325330453917701e-12,
+      "loss": 0.5623,
+      "step": 27599
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.8804385577880825e-12,
+      "loss": 0.5905,
+      "step": 27600
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.573155352767146e-12,
+      "loss": 0.5459,
+      "step": 27601
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.403480858838904e-12,
+      "loss": 0.613,
+      "step": 27602
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.3714150904362587e-12,
+      "loss": 0.6072,
+      "step": 27603
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4769580619921075e-12,
+      "loss": 0.5487,
+      "step": 27604
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7201097868291271e-12,
+      "loss": 0.5402,
+      "step": 27605
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1008702749393251e-12,
+      "loss": 0.5476,
+      "step": 27606
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.192395352044856e-13,
+      "loss": 0.4905,
+      "step": 27607
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7521757206550037e-13,
+      "loss": 0.5432,
+      "step": 27608
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.880439329393085e-14,
+      "loss": 0.549,
+      "step": 27609
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 0.5467,
+      "step": 27610
+    },
+    {
+      "epoch": 1.0,
+      "step": 27610,
+      "total_flos": 3.232099675381338e+20,
+      "train_loss": 0.012050939151360823,
+      "train_runtime": 4549.9797,
+      "train_samples_per_second": 1553.428,
+      "train_steps_per_second": 6.068
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 27610,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 1000,
+  "total_flos": 3.232099675381338e+20,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}